DOI: 10.1126/Science.1182238 , 1480 (2010); 327 Science , Et Al. Joseph Henrich and Punishment Markets, Religion, Community Size

Home , Joseph Henrich

Markets, Religion, Community Size, and the Evolution of Fairness and Punishment Joseph Henrich, et al. Science 327, 1480 (2010); DOI: 10.1126/science.1182238

This copy is for your personal, non-commercial use only.

If you wish to distribute this article to others, you can order high-quality copies for your colleagues, clients, or customers by clicking here. Permission to republish or repurpose articles or portions of articles can be obtained by following the guidelines here.

The following resources related to this article are available online at www.sciencemag.org (this infomation is current as of August 23, 2011 ):

A correction has been published for this article at: http://www.sciencemag.org/content/332/6036/1381.full.html Updated information and services, including high-resolution figures, can be found in the online version of this article at: http://www.sciencemag.org/content/327/5972/1480.full.html

Supporting Online Material can be found at: on August 23, 2011 http://www.sciencemag.org/content/suppl/2010/03/16/327.5972.1480.DC1.html A list of selected additional articles on the Science Web sites related to this article can be found at: http://www.sciencemag.org/content/327/5972/1480.full.html#related This article cites 26 articles, 8 of which can be accessed free: http://www.sciencemag.org/content/327/5972/1480.full.html#ref-list-1

This article has been cited by 1 article(s) on the ISI Web of Science www.sciencemag.org This article has been cited by 15 articles hosted by HighWire Press; see: http://www.sciencemag.org/content/327/5972/1480.full.html#related-urls This article appears in the following subject collections: Economics http://www.sciencemag.org/cgi/collection/economics

Psychology Downloaded from http://www.sciencemag.org/cgi/collection/psychology

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2010 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS. RESEARCH ARTICLE CORRECTED 17 JUNE 2011; SEE LAST PAGE and beliefs to variable circumstances (7, 9). Modeling work shows that when these learning mechanisms are applied to different kinds of Markets, Religion, Community Size, social interactions, such as large-scale cooperation or ephemeral exchange, individually costly behaviors can be sustained by punishment, sig- and the Evolution of Fairness naling, and reputational mechanisms (10–13). By sustaining such behaviors, norms can facil- and Punishment itate trust, fairness, and cooperation in a diverse array of interactions, thereby allowing the most Joseph Henrich,1* Jean Ensminger,2 Richard McElreath,3 Abigail Barr,4 Clark Barrett,5 productive use of unevenly distributed skills, Alexander Bolyanatz,6 Juan Camilo Cardenas,7 Michael Gurven,8 Edwins Gwako,9 knowledge, and resources, as well as increasing Natalie Henrich,10 Carolyn Lesorogol,11 Frank Marlowe,12 David Tracer,13 John Ziker14 cooperation in exchange, public goods, and war- fare. More-effective norms and institutions can Large-scale societies in which strangers regularly engage in mutually beneficial transactions are spread among societies by a variety of theoret- puzzling. The evolutionary mechanisms associated with kinship and reciprocity, which underpin much of ically and empirically grounded mechanisms, primate sociality, do not readily extend to large unrelated groups. Theory suggests that the evolution of including conquest and assimilation, preferential such societies may have required norms and institutions that sustain fairness in ephemeral imitation of more-successful societies, or forward- exchanges. If that is true, then engagement in larger-scale institutions, such as markets and world looking decision making by leaders or high-status religions, should be associated with greater fairness, and larger communities should punish unfairness coalitions (14, 15). more. Using three behavioral experiments administered across 15 diverse populations, we show that Norms that enhance fairness among stran- market integration (measured as the percentage of purchased calories) positively covaries with fairness gers are likely causally interconnected with the while community size positively covaries with punishment. Participation in a world religion is associated diffusion of several kinds of institutions. Here with fairness, although not across all measures. These results suggest that modern prosociality is not we focus on two: (i) the expansion of both solely the product of an innate psychology, but also reflects norms and institutions that have emerged the breadth and intensity of market exchange

overthecourseofhumanhistory. (16, 17), and (ii) the spread of world religions on August 23, 2011 (18). The efficiency of market exchange involv- t the onset of the Holocene, the sta- trusting, and cooperative behavior among ing infrequent or anonymous interactions bilization of global climates created strangers, even in one-shot encounters (3). improves with an increasingly shared set of Apossibilities for the emergence of larger- Two major theoretical approaches have motivations and expectations related to trust, scale sedentary human societies (1). Over the sought to explain both the relatively rapid fairness, and cooperation. This lowers transac- next 10 millennia, the scale of some human expansion of human societal scales and the tion costs, raises the frequency of successful communities dramatically expanded from kin- puzzlingly prosocial behavior observed in ex- transactions, and increases long-term rewards based foraging bands to complex, intensely periments. The first approach proposes that (16). Although frequent and efficient exchanges

cooperative societies in which strangers fre- humans possess an innate social psychology among strangers are now commonplace, studies www.sciencemag.org quently engage in mutually beneficial transactions calibrated to life in the small-scale societies of of nonhuman primates and small-scale societies (2). Consistent with life in these large-scale our Paleolithic ancestors. Rooted in the evolu- suggest that during most of our evolutionary societies, behavioral experiments performed tionary logic of kinship and reciprocity, these history, transactions beyond the local group, and with people from these populations reveal fair, heuristics were mistakenly extended to nonkin certainly beyond the ethnolinguistic unit, were and ephemeral interactants, as societies ex- fraught with danger, mistrust, and exploitation panded with the emergence of agriculture. From (2, 16, 19). Thus, we propose that such “market 1Departments of Psychology and Economics, University of this view, the prosocial behavior observed in norms” may have evolved as part of an overall

British Columbia, 2136 West Mall, Vancouver, British Downloaded from 2 experiments directly reflects the operation of process of societal evolution to sustain mutually Columbia V6T 1Z4, Canada. Division of the Humanities 4 5 and Social Sciences, California Institute of Technology, these ancient heuristics ( , ). beneficial exchanges in contexts where es- Pasadena, CA 91125, USA. 3Department of Anthropology, An alternative approach proposes that a tablished social relationships (for example, kin, University of California, Davis, CA 95616, USA. 4Centre for crucial ingredient in the rise of more-complex reciprocity, and status) were insufficient. If our the Study of African Economies, Department of Economics, societies was the development of new social theory is correct, then measures of fairness in University of Oxford, Manor Road, Oxford OX1 3UQ, UK. 5Department of Anthropology, University of California, 341 norms and informal institutions that are capable situations lacking relationship information (for Haines Hall, Box 951553, Los Angeles, CA 90095, USA. of domesticating our innate psychology for life example, anonymous others) should positively 6Department of Anthropology, College of DuPage, Glen Ellyn, in ever-expanding populations (6). Larger and covary with market integration. 7 IL 60137, USA. Facultad de Economia, Universidad de Los more-complex societies prospered and spread to Recent work has also tentatively proposed Andes, K1 Number 18A-70, Bogotá, Colombia. 8Department of Anthropology, University of California, Santa Barbara, CA the degree that their norms and institutions that certain religious institutions, beliefs, and 93106, USA. 9Department of Sociology and Anthropology, effectively sustained successful interaction in rituals may have coevolved with the norms that Guilford College, 5800 West Friendly Avenue, Greensboro, NC ever-widening socioeconomic spheres, well support large-scale societies and broad exchange 27410, USA. 10Centre for Health Evaluation and Outcome beyond individuals’ local networks of kin and (18, 20, 21). Intersocietal competition may have Sciences, Providence Health Care Research Institute, Vancou- long-term relationships (7). It is these particular favored those religious systems that galvanize ver, British Columbia V6Z 1Y6, Canada. 11George Warren Brown School of Social Work, Washington University, 1 norms and their gradual internalization as prosocial behavior in broader communities, Brookings Drive, St. Louis, MO 63130, USA. 12Department of proximate motivations (8) that recalibrate our perhaps using both supernatural incentives (for Anthropology, Florida State University, Tallahassee, FL 32306, innate psychology for life in small-scale societies example, hell) and recurrent rituals that intensify 13 USA. Department of Anthropology, University of Colorado, in a manner that permits successful larger-scale group solidarity (20, 22). Consistent with this Post Office Box 173364, Campus Box 103, Denver, CO 80217, USA. 14Department of Anthropology, Boise State cooperation and exchange in vast communities. view, analyses of ethnographic data show that University,1910UniversityDrive,Boise,ID83725,USA. Much research suggests that norms arise the emergence of moralizing religions increases *To whom correspondence should be addressed. E-mail: because humans use evolved learning mecha- with greater societal size and complexity (18, 23). [email protected] nisms to calibrate their behavior, motivations, Archaeologically, regular rituals and the con-

1480 19 MARCH 2010 VOL 327 SCIENCE www.sciencemag.org RESEARCH ARTICLE struction of monumental religious architecture effort with 10 new ones (swapping several to accept or reject each of the possible offers (in co-emerge with societal size and complexity researchers as well), while resampling from five 10% increments). Decisions are binding. If (24). In experiments, unconsciously priming the of the same populations used previously (Table 1). Player 2 specified that he or she would accept faithful with religious concepts favors greater This analysis complements prior work from our the amount of the actual offer, then Player 2 fairness toward anonymous others (21). This team, which focused on the distributional receives the offered amount and Player 1 gets suggests that, in contrast to the religions that patterns of punishment (30), by seeking to the remainder. If Player 2 specified that he or likely dominated our evolutionary history, mod- explain the variation in our experimental mea- she would reject the offered amount, then both ern world religions such as Christianity and sures of both fairness and punishment. Our players receive zero. If people are motivated Islam may be unusual in ways that buttress the analysis converges with recent work comparing purely by money maximization, Player 2s will norms and institutions that sustain larger-scale diverse industrialized populations (31). always accept any positive offer; realizing this, interaction [supporting online material (SOM) If markets and world religions are linked to Player 1 should offer the smallest nonzero text]. If this theory is correct, then greater fair- the norms that sustain exchange in large-scale amount. Because this is a one-shot anonymous ness toward anonymous others should be asso- societies, we expect that experimental measures interaction, Player 2’s willingness to reject ciated with adherence to a world religion. of fairness in anonymous interactions will pos- provides a measure of punishment. Player 1’s Evolutionary approaches to norms also itively covary with measures of involvement in offer measures a combination of social motiva- afford a prediction about the willingness of these two institutions. To test this, we used three tions and an assessment of the likelihood of individuals across populations to engage in the experiments that were designed to measure indi- rejection; this gives us a second behavioral costly punishment of norm violations. Theoret- viduals’ propensities for fairness and their willing- measure of fairness. ical modeling reveals at least two different kinds ness to punish unfairness across 15 populations In the Third-Party Punishment Game (TPG), of norm-stabilizing mechanisms: One involves that vary in their degree of market integration two players are again allotted a stake, but now a reputational effects in which norm violators are and their participation in world religions. Our third player also receives the equivalent of one- sanctioned in another interaction by, for exam- three experiments are the Dictator, Ultimatum, half of the stake (33). Player 1 must decide how ple, not receiving aid in a dyadic helping sit- and Third-Party Punishment Games (32). much to allocate to Player 2, who has no uation (10), and a second involves the use of In the Dictator Game (DG), two anonymous choices. Player 3, before hearing the actual diffuse (25) costly punishment (12). Because the players are allotted a sum of money (the stake) amount that Player 1 allocated to Player 2, has effectiveness of reputational systems in sustaining in a one-shot interaction (3). Player 1 must to decide whether to pay 20% of his or her norms degrades rapidly as communities expand decide how to divide this sum between himself allocation to punish Player 1 for each of the on August 23, 2011 (26, 27), fairness in larger communities must or herself and Player 2. Player 2 receives the possible offers (in 10% increments). If punished, increasingly be maintained by diffuse punish- allocation (offer), and the game ends. Player 1’s Player 1 loses triple the amount paid by Player ment; that is, larger communities should punish offer to Player 2 provides a measure of Player 3. Suppose the stake is $100; if Player 1 offers more. 1’s behavioral fairness in this context. $10 to Player 2 (keeping $90), and Player 3 Experiments. The evidence presented below In the Ultimatum Game (UG), two anony- wants to punish this offer amount, then Player derives from a second round of cross-cultural mous players are again allotted a sum in a one- 1 takes home $60 ($90 – $30), Player 2 gets $10, experiments that were designed to illuminate shot interaction (3). Player 1 can offer a portion and Player 3 gets $40 ($50 – $10). If Player 3 findings from our first project (28, 29). This of this to Player 2. Player 2, before hearing the had instead decided not to punish offers of $10,

round replaces 10 populations from our earlier actual offer from Player 1, must decide whether then the take-home amounts would be $90, www.sciencemag.org

Table 1. Summary information for the populations studied. The column sample that reported adhering to a world religion. Mean CS gives the Economic Base classifies the production systems (for example, horticul- average CS for the populations studied. These CSs are for villages, except turalists rely primarily on slash-and-burn agriculture; pastoralists rely on among the Hadza (who live in camps) and Accra. Com Sam (DG/UG/TPG) herding). Residence classifies the nature and frequency of each population’s gives the number of different communities from which participants were movements. Mean MI is the average percentage of total household calories drawn for the DG, UG, and TPG. N gives the number of pairs or trios for each that are purchased in the market. Mean WR gives the percentage of the experiment (table S4). Downloaded from

Mean Mean Mean Com Sam N Population Location Environs Economic base Residence MI WR CS (DG/UG/TPG) (DG/UG/TPG) Accra City Ghana Urban Wage work Sedentary 100 97 44† 3/3/11 30/30/39 Au* Papua New Guinea Mountainous forest Horticulture and foraging Sedentary 1 100 309 3/3/1 30/30/30 Dolgan/ Ng. Siberia Tundra-taiga Hunting, fishing, and wages Semi-sedentary 63 59 612 1/1/0 30/30/0 Gusii Kenya Fertile high plains Farming and wage work Sedentary 28 100 4063 2/2/1 25/25/30 Hadza* Tanzania Savanna-woodlands Foraging Nomadic 0 0 43 4/4/3 31/31/27 Isanga village Tanzania Mountainous forest Farming and wage work Sedentary 70 99 1500 1/1/1 30/30/20 Maragoli Kenya Fertile plains Farming and wage work Sedentary 43 100 3843 2/2/1 25/25/30 Orma* Kenya Semi-arid savanna Pastoralism Semi-nomadic 72 100 125 4/0/0 26/0/0 Samburu Kenya Semi-arid savanna Pastoralism Semi-nomadic 69 66 2000 1/1/1 31/31/30 Sanquianga Colombia Mangrove forest Fisheries Sedentary 82 84 1931 1/1/2 30/30/32 Shuar Ecuador Tropical forest Horticulture Sedentary 22 76 498 2/2/1 21/21/15 Sursurunga Papua New Guinea Island Horticulture Sedentary 24 100 186 3/3/1 30/30/32 Tsimane* Bolivia Tropical forest Horticulture and foraging Semi-nomadic 7 100 314 2/2/1 38/36/27 United States* Missouri Prairie Wage work Sedentary 100 100 1813 1/1/0 15/26/0 Yasawa Fiji Coastal tropical Horticulture and marine Sedentary 21 100 109 2/2/2 35/34/30 pacific foraging *These populations were also sampled from during our first project (28). †Participants were sampled from small enterprises in the city of Accra. The value 44 is the average number of employees in the enterprises. Because of the anomalous nature of this situation with regard to obtaining a CS comparable with our other populations, we did not include Accra in analyses involving CS.

www.sciencemag.org SCIENCE VOL 327 19 MARCH 2010 1481 RESEARCH ARTICLE

$10, and $50, respectively. Because a money- Table 1 provides the location (mapped in fig. S3), associated with an increase in percentage offered maximizing Player 3 would never pay to environment, economic base, residence pattern, ranging from roughly 2 to 3.4. The same punish, a similarly motivated Player 1 should and sampling information for each population, as qualitative results are obtained for MI whether always give zero to Player 2. Thus, Player 3’s well as averages for three key variables. one uses household measures, community aver- willingness-to-pay provides another measure of Results. The theory outlined above predicts ages, or population averages. punishment. Player 1’s offer measures a mixture a positive relationship between our three mea- The coefficients on WR are also large across of his or her social motivations in this context sures of fairness—offers in each game—with all offers and in both the UG and DG, though and an assessment of the punishment threat; this market integration and adherence to a world re- not well estimated in the DG (P < 0.10). How- provides a third behavioral measure of fairness. ligion. Individuals’ offers are measured as a per- ever, WR’s coefficients are significant (P <0.05) These experiments, with their salient contex- centage of the total stake (1 day’s local wage). across all other specifications (tables S5, S8, and tual cues of cash and anonymity, seem well suited Market Integration (MI) is measured at the S11). Participating in a world religion is asso- to tapping the particular norms that support co- household level by calculating the percentage ciated with an increase in percentage offered of operation and exchange among ephemeral inter- of a household’s total calories that were pur- between about 6 and 10 (36). actants in market transactions. Cash is closely chased from the market, as opposed to home- Taken together, these data indicate that going associated with market transactions (34)andoften grown, hunted, or fished, and then averaged to from a fully subsistence-based society (MI = 0) signals a desire to avoid a longer-term nonmarket obtain a community-level measure. We use the with a local religion to a fully market-incorporated relationship (35). Anonymity in our games means community average for MI both to remain society (MI = 100%) with a world religion pre- that players lack the cues or information nec- consistent with our definition of norms (as local dicts an increase in percentage offered of roughly essary to apply the expectations and motivations equilibria) and to remove day-to-day stochastic 23, 20, and 11 in the DG, UG, and TPG, re- associated with other kinds of relationships, variation (32). Table 1 shows that the population spectively. This spans most of the range of var- such as those based on kinship, reciprocity, or means for MI range from 0 to 100%, with a iation across our populations: DG means range status differences, forcing players to default to mean of 57.3%. World Religion (WR) was as- from 26 to 47%, UG from 25 to 51%, and TPG local norms for dealing with people outside sessed by asking participants what religion they from 20 to 43%. durable relationships. However, an important practiced, and coding these as a binary variable, For the seven other socioeconomic variables, concern in interpreting such experiments in- with “1” indicating Islam or Christianity, and none of their coefficients is significant (P < volves the degree to which participants accept “0” marking the practice of a tribal religion or 0.05) across all offers or in the DG (tables S5

the experimental situation, believe the anonym- “no religion.” Table 1 provides the percentage and S8). For UG offers, the coefficient on age is on August 23, 2011 ity, and worry about the experimenter’s judg- of each population that adheres to a world re- also positive (P < 0.05, table S11). For TPG ments of them (SOM text). ligion. The mean value of WR is 89%. offers, Table 2 shows that the coefficients on Our standardized protocols and methods To analyze these data, we regressed offers on income, wealth, and household size are signif- included the following: (i) random adult samples MI and WR, as well as seven control variables: icant (P < 0.05, table S14). from our communities (with little attrition in most age, sex, education, income, wealth, household Figure 1 plots mean offers versus mean MI sites), (ii) stakes set at 1 day’s local wage, (iii) a size, and community size (CS). Except for wealth values for our 15 populations. Population mean show-up fee of 20% of the stake, (iv) back- and household size, which are both measured at MI values account for 52% of the variation in translation of scripts, (v) one-on-one instruction the household level, and CS, these control varia- mean DG offers [correlation r = 0.72, 95% boot-

and pregame testing for comprehension, (vi) bles are individual-level measures. We used 30 strap confidence interval (CI) = 0.4 to 1.0 , P < www.sciencemag.org steps to preclude collusion and contamination, (DG), 26 (UG), and 16 (TPG) community-level 0.01, n = 15 populations]. In designing this and (vii) no deception (32). Our pregame tests means for MI to predict 416 (DG), 398 (UG), and second round of experiments, we sought out an combined with analyses of the relationships 272 (TPG) individual offers with a minimal set of additional New Guinea population (Sursurunga), between experimental decisions and measures controls, or 336 (DG), 319 (UG), and 265 (TPG) because in the first round, the Au of New Guinea of comprehension, indicate no measurable impact individual offers with the full set of controls. revealed highly unusual behavior, including of differences in comprehension on behavior (30). Table 2 shows four regression models using relatively high offers with little market integra- We performed these experiments with 2148 MI and WR to predict all offers together and tion. The Au pattern replicated, and now extends

individuals across populations from Africa, offers from the DG, UG, and TPG, separately. to the Sursurunga. However, because we targeted Downloaded from North and South America, Oceania, New Independent of the other sociodemographic a second population that skews our world sample Guinea, and Asia that included small-scale variables, the coefficients on MI are large and unrepresentatively toward New Guinea, we also societies of hunter-gatherers, marine foragers, significant at conventional levels across all four examined this relationship with either the Au or pastoralists, horticulturalists, and wage laborers. models. A 20–percentage point increase in MI is the Sursurunga dropped. Dropping the Sursurunga

Table 2. Linear regression models for offers. These ordinary least-squares models include four additional control variables (sex, age, community size, and education). Coefficients are followed by standard errors, indicated with T; P values are given in parentheses.

Variables UG TPG All offers* DG offers† offers† offers† MI 0.12 T 0.023 (<0.001) 0.17 T 0.035 (<0.001) 0.098 T 0.035 (0.005) 0.11 T 0.044 (0.044) WR‡ 5.96 T 2.04 (0.0036) 6.4 T 3.61 (0.079) 10.4 T 2.67 (<0.001) 0.45 T 3.06 (0.879) Income (per U.S. $1000) 0.096 T 0.089 (0.28) –0.012 T 0.15 (0.93) 0.16 T 0.10 (0.13) –2.25 T 0.94 (0.017) Wealth (per U.S. $1000) 0.0012 T 0.006 (0.83) 0.0013 T 0.008 (0.88) –0.0056 T 0.008 (0.43) 1.2 T 0.25 (<0.001) Household size –0.24 T 0.21 (0.24) –0.13 T 0.31 (0.67) –0.24 T 0.2 (0.37) –1.0 T 0.43 (0.019) Observations; R2 920; 0.084 336; 0.10 319; 0.14 265; 0.10 *Clustered robust standard errors adjust for repeated observations of the same people, 596 individuals (clusters). Indicator variables for each experiment (UG, TPG) were included to control for differences among offers in the three games (table S5). †Robust standard errors used. See (32) for models with clustered robust standard errors [clustering on populations (Table 1) to control for the nonindependence of individuals], continental-level controls to address any shared culture history (Model 1 in tables S8, S11, and S14), and a variety of other robustness checks, which are summarized in the text. ‡WR is an individual-level dummy variable with “1” indicating Islam or Christianity, and “0” marking the practice of a tribal religion or “no religion.” We have few “no religion” responses, and ethnography suggests that in the wake of missionary activities people associate “religion” with “Christianity.” Thus, responses of “no religion” probably imply belief in the local or traditional religion.

1482 19 MARCH 2010 VOL 327 SCIENCE www.sciencemag.org RESEARCH ARTICLE leads to mean MI accounting for 58% of the and MI, because, for small communities, there sures, new experiments, and additional controls. variation (r = 0.76, 95% bootstrap CI = 0.44 to are numerous reputational mechanisms (not in- Despite swapping in 10 new sites and using a 0.95, P < 0.001). Dropping the Au instead leads volving costly punishment) that can sustain equality different protocol, we have replicated our earlier to mean MI capturing 71% of the variation (r = norms. Similarly, because religion should gal- UG findings at both the level of individual sites 0.84, 95% bootstrap CI = 0.59 to 0.96, P < 0.001). vanize the existing mechanisms for norm stabi- (in four sites) and in obtaining a positive rela- On the punishment side, we looked for a lization, whether participants in a world religion tionship between UG offers and market integra- relationship between our individual-level mea- punish more depends entirely on the local tion with a better measure. This relationship is sures of punishment, from the UG and TPG, stabilization mechanisms, which themselves now demonstrated for DG and TPG offers, each and our theoretically important variable, CS, depend on CS (38). Nevertheless, for MAOs in of which reveals similarly large effects and is while controlling for our other eight variables. the TPG, we find that WR is associated with robust to the inclusion of a range of socio- For both the UG and TPG, we reduced the significantly (P < 0.05) more punishment, economic variables not previously measured and vector of “punish or not punish” responses to a although MI reveals no such relationship. None to a suite of statistical checks (tables S5 to S25). single number called a minimum acceptable of the other predictors is significant (P < 0.05, These findings also delineate two additional offer (MAO). This is the lowest offer (≤50%) table S20A). For MAO in the UG (table S23A), lines of research. First, our inclusion of partic- for which an individual no longer punishes. income, wealth, and MI all significantly predict ipation in a world religion converges with other Thus, if a player accepts all offers, his or her lower MAOs (P < 0.05). The effect for MI, recent findings (21) and tentatively supports the MAO is zero. If the player punishes offers of 0 however, is contingent on having CS in the notion that religion may have coevolved with and 10% but accepts all higher offers, his or her equation. If CS is dropped, the effect of MI complex societies to facilitate larger-scale inter- MAO is set at 20. MAO measures an individ- becomes nonsignificant. Thus, unlike CS, none actions. Second, our analyses (tables S18 and ual’s willingness to punish low offers (32). of these other effects is consistent for both MAO S19) open up the question of why, when a third- To analyze MAOs, we used an ordered measures or across alternative specifications. party punisher is added to a DG to create a logistic regression (OLR) because MAOs are Our analyses for both offers and MAOs are TPG, do mean offers usually decrease, the both discrete and bimodally distributed. This robust to a variety of checks, including alter- predictive effects of WR disappear (Table 2), model assumes that the dependent variable is native model specifications and adjustments to and economic variables emerge as potent discrete and rank ordered, but that the distance our wealth and income variables to account for correlates of offers (tables S14 to S16)? between ranks is not meaningful. Because the local differences in purchasing power (tables S6, On the punishment side, our analyses of output of an OLR is nonintuitive (tables S20A S9, S12, S15, S21, and S24). We also included MAOs are consistent with the idea that as on August 23, 2011 and S23A), we have captured the effects of the continental-level dummies (Africa, Oceania, reputational systems break down in larger highly significant (P < 0.001) coefficients for South America, and Eurasia) to address concerns populations, increasing levels of diffuse costly CS in Fig. 2, A and B. Holding the other eight about shared cultural phylogenies, and we used punishment are required to sustain large harmo- variables at their mean values, the figure shows clustered robust standard errors (clustering on nious communities. This extends earlier bi- how the distribution of MAOs for the TPG and site) to control for the potential nonindepen- variate analyses using population means in the UG shift with increases in CS. Small commu- dence of individual observations within our sites TPG (40) and converges with ethnographic data, nities, the size of foraging bands, are the least (tables S5 to S25). The findings for MI and CS suggesting that large communities, lacking suffi- willing to punish. As CS increases from 50 to are robust to these checks. However, when cient and effective punishing mechanisms, frag-

5000, there is a dramatic shift from a modal continental controls are applied, the effects of ment into smaller groups as their community www.sciencemag.org MAO of 0 to 50. Our communities’ sizes range WR on offers disappears because of the highly sizes reach about 300 people (41). from20to4600people. uneven distribution of populations containing Although our primary interest here is behav- Theoretical arguments suggest that punish- individuals with WR = 0 (tables S5, S8, and ioral fairness, which includes whatever combi- ment (MAO) should be related more directly to S11, Model 1). We reran our models for Africa nation of motivations and expectations yields the natural logarithm of CS, because the effec- only and generally reconfirmed the above more equal divisions, two features of these tiveness of reputational systems decays in rough findings (32). Nevertheless, because of the results suggest that they may stand when proportion to this variable (26, 37). Consistent rather small sample of individuals with WR = considered as measures of intrinsic (internalized) with this, the natural logarithm of CS is a better 0, any conclusions about the effect of WR must motivations. First, because no punishment op- Downloaded from predictor of MAO in both experiments than CS remain quite tentative (39). portunities exist in the DG, this experiment itself (tables S20B to S25B), and this amplifies Discussion. Our results on the relationship likely provides a purer measure of intrinsic the effects illustrated in Fig. 2 (fig. S4). between MI and offers extend the findings of motivations for equal offers. Second, we reran Our theoretical approach makes no predic- our previous project (29) in several ways, the above offer regressions for UG and TPG tions about the relationship between punishment including using fresh samples, better mea- offers including each population’s mean MAO as an additional predictor variable. If individuals Fig. 1. Mean DG offers for 50 are well calibrated to the threat of punishment Sanquianga each population plotted against U.S. posed by their local communities, then adding mean value of MI. Error bars 45 Orma this variable should (at least partially) remove are bootstrapped standard errors Accra the impact of any threat of punishment on offers. (bias corrected and accelerated) Au Sursurunga 40 Samburu The inclusion of mean MAO as a predictor on the population mean. Dolgan && Nganasan Nganasan Shuar variable does not change the coefficients on MI Fiji 35 Maragoli Isanga villagevillage or WR (table S17). Finally, anonymity concerns Gusii regarding our experiments and results are taken up 30 in the SOM text, where we discuss our double- Hadza blind treatments and related analyses, which indi- Tsimane 25 cate that differences in perceived anonymity are

Mean Dictator Game Offer (% Offer (% of stake) Dictator Game Mean unlikely to explain our major findings. 20 These findings indicate that people living in 0 102030405060708090100small communities lacking market integration or Market Integration (percentage of purchased calories in diet) world religions—absences that likely character-

www.sciencemag.org SCIENCE VOL 327 19 MARCH 2010 1483 RESEARCH ARTICLE

Fig. 2. Effect of CS on MAOs. A B Eachsetofbarsshowsthe 0.50 0.80

distribution of MAOs, ranging 0.70 discretely from 0 to 50, for 0.40 Community Size of 50 0.60 Community Size of 50 different CSs. The coefficients Community Size of 500 Community Size of 500 used to create the plot are for 0.30 Community Size of 2500 0.50 Community Size of 2500 thevariableCSinanordered Community Size of 5000 0.40 Community Size of 5000 logistic regression containing all eight of our other varia- 0.20 0.30

bles. All variables except CS 0.20 are held constant at their 0.10 mean values. The coefficients 0.10 Predicted Frequency of MAO in UG of MAO Frequency Predicted on CS represented graphically in TPG of MAO Frequency Predicted 0.00 0.00 here are 0.052 T 0.0085 01020304050 0 1020304050 (robust standard error, P < Minimum Acceptable Offers (MAO) Minimum Acceptable Offers (MAO) 0.001, n =227)fortheTPG (Willingness to engage in third party punishment) (Willingness to punish in the ultimatum) and 0.087 T 0.011 (robust standard error, P <0.001,n =297)fortheUG.(A) TPG (see table S20A for full regression). (B) UG (see table S23A).

ized all societies until the Holocene—display capable of harnessing and extending our evolved 28. J. Henrich et al., Eds., Foundations of Human Sociality: relatively little concern with fairness or punishing social psychology to accommodate life in large, Economic Experiments and Ethnographic Evidence from Fifteen Small-Scale Societies (Oxford Univ. Press, Oxford, unfairness in transactions involving strangers or intensely cooperative communities. Recent exper- 2004). anonymous others. This result challenges the imental work among diverse industrialized popu- 29. J. Henrich et al., Behav. Brain Sci. 28, 795, discussion hypothesis that successful social interaction in lations suggests that the gradual honing of these 815 (2005). large-scale societies—and the corresponding norms and institutions continues in modern soci- 30. J. Henrich et al., Science 312, 1767 (2006). — 31 31. B. Herrmann, C. Thöni, S. Gächter, Science 319, 1362 experimental findings arise directly from an eties ( ). (2008). evolved psychology that mistakenly applies kin 32. Background, methods, and supplemental analyses are

and reciprocity-based heuristics to strangers in References and Notes available as supporting material on Science Online. on August 23, 2011 vast populations (4, 5), without any of the 1. P. J. Richerson, R. Boyd, R. L. Bettinger, Am. Antiq. 66, 33. E. Fehr, U. Fischbacher, Evol. Hum. Behav. 25,63 “psychological workarounds” (42) that are cre- 387 (2001). (2004). 2. J. M. Diamond, Guns, Germs, and Steel: The Fates of 34. Cash has value in all our societies. Everyone wants steel, ated by norms and institutions. Moreover, it is not Human Societies (Norton, New York, 1997). salt, and sugar, if nothing else. clear how this hypothesis can explain why we 3. C. Camerer, Behavior Game Theory: Experiments in 35. K. D. Vohs, N. L. Mead, M. R. Goode, Science 314, 1154 find so much variation among populations in our Strategic Interaction (Princeton Univ. Press, Princeton, (2006). experimental measures and why this variation is NJ, 2003). 36. We found no robust differences among different forms of 4. T. C. Burnham, D. D. Johnson, Anal. Kritik 27, 113 (2005). Christianity (Catholics versus Protestants), or between so strongly related to MI, WR, and CS. The mere 5. M. A. Nowak, K. M. Page, K. Sigmund, Science 289, 1773 Christianity and Islam. fact that the largest and most anonymous com- (2000). 37. R. F. I. Cancho, R. V. Sole, R. Kohler, Phys. Rev. E 69,

munities engage in substantially greater punish- 6. D. C. North, Institutions, Institutional Change, and Economic 051915 (2004). www.sciencemag.org ment relative to the smallest-scale societies, who Performance (Cambridge Univ. Press, New York, 1990). 38. Traditional religions may sometimes galvanize local norms 7. S. Bowles, Microeconomics: Behavior, Institutions, and (SOM text), but they do not galvanize “market norms,” which punish very little, challenges this interpretation. Evolution (Princeton Univ. Press, Princeton, NJ, 2004). are for interacting with strangers or anonymous others. Methodologically, our findings suggest cau- 8. E. Fehr, C. F. Camerer, Trends Cogn. Sci. 11, 419 (2007). When religions do underwrite norms, they must operate in tion in interpreting behavioral experiments from 9. P. Richerson, R. Boyd, Not by Genes Alone: How Culture concert with the local norm-stabilizing mechanisms. industrialized populations as providing direct Transformed Human Evolution (Univ. of Chicago Press, 39. We did not use fixed effects for communities or populations Chicago, 2005). in any of our regressions, because our theoretical focus is on insights into human nature. Combining our 10. K. Panchanathan, R. Boyd, Nature 432, 499 (2004). differences in social norms. Because norms are group-level findings with work on the links between 11. H. Gintis, E. A. Smith, S. Bowles, J. Theor. Biol. 213, 103 properties (different locally stable equilibria), much of the behavioral experiments and real life (31, 43, 44) (2001). relevant norm-related variation occurs between, not within, Downloaded from suggests that such experiments elicit norms, or 12. J. Henrich, R. Boyd, J. Theor. Biol. 208, 79 (2001). communities or populations (SOM text). reflect institutions, that have evolved to facilitate 13. K. Sigmund, C. Hauert, M. A. Nowak, Proc. Natl. Acad. 40. F. W. Marlowe et al., Proc. Biol. Sci. 275, 587 (2008). Sci. U.S.A. 98, 10757 (2001). 41. A. Forge, in Man, Settlement and Urbanisation, P. Ucko, interactions among individuals not engaged in 14. R. Boyd, P. J. Richerson, J. Theor. Biol. 215, 287 (2002). R. Tringham, G. Dimbelby, Eds. (Duckworth, London, durable relationships. Given this, much current 15. J. Ensminger, J. Knight, Curr. Anthropol. 38, 1 (1997). 1972), pp. 363–376. work using behavioral games appears to be 16. S. Bowles, J. Econ. Lit. 36, 75 (1998). 42. P. Richerson, R. Boyd, Hum. Nat. 10, 253 (1999). studying the interaction between a particular set 17. A. Smith, The Theory of Moral Sentiments (Prometheus, 43. M. Poppe, J. Econ. Psychol. 26, 431 (2005). New York, 2000). 44. M. Gurven, J. Winking, Am. Anthropol. 110, 179 (2008). of norms and our evolved psychology, not 18. R. Wright, The Evolution of God (Little, Brown, Boston, 2009). 45. We especially thank the communities that participated tapping this psychology directly. 19. E. Fehr, J. Henrich, in Genetic and Cultural Evolution of in our research. We also thank the National Science Overall, these findings lend support to the idea Cooperation, P. Hammerstein, Ed. (MIT Press, Cambridge, Foundation, the MacArthur Norms and Preferences that the evolution of societal complexity, especially MA, 2003). Network, the Max Planck Institute for Social 20. A. Shariff, A. Norenzayan, J. Henrich, in Evolution, Anthropology, and the Russell Sage Foundation for as it has occurred over the last 10 millennia, in- Culture and the Human Mind, M. Schaller, funding this project. Thanks also to the many audiences volved the selective spread of those norms and A. Norenzayan, S. Heine, T. Yamaguishi, T. Kameda, Eds. and readers who contributed to improving our efforts. institutions that best facilitated the successful (Psychology Press, New York, 2009), pp. 119–136. exchange and interaction in socioeconomic spheres 21. A. Norenzayan, A. F. Shariff, Science 322, 58 (2008). Supporting Online Material 22. J. Henrich, Evol. Hum. Behav. 30, 244 (2009). www.sciencemag.org/cgi/content/full/327/5972/1480/DC1 well beyond local networks of durable kin and 23. F. L. Roes, M. Raymond, Evol.Hum.Behav.24, 126 (2003). Materials and Methods reciprocity-based relationships. Although differ- 24. J. Marcus, K. V. Flannery, Proc. Natl. Acad. Sci. U.S.A. SOM Text ences in environmental affordances probably had 101, 18257 (2004). Figs. S1 to S4 a profound impact on the emergence of complex 25. Diffuse punishment implies that the responsibility for Tables S1 to S25 2 punishing is spread over a large segment of the population. References societies across the globe ( ), the rate-determining 26. K. Panchanathan, R. Boyd, J. Theor. Biol. 224, 115 (2003). step in societal evolution may have involved the 27. A. K. Dixit, Lawlessness and Economics: Alternative Modes of 21 September 2009; accepted 22 January 2010 assembly of the norms and institutions that are Governance (PrincetonUniv.Press,Princeton,NJ,2004). 10.1126/science.1182238

1484 19 MARCH 2010 VOL 327 SCIENCE www.sciencemag.org Corrections & CLarifications

erratum Post date 17 June 2011

Research Articles: “Markets, religion, community size, and the evolution of fairness and punishment” by J. Henrich et al. (19 March 2010, p. 1480). Authors’ note: Since the publication of our article, we have been preparing a book-length treatment of our comparative project for publication, with a chapter describing the work done in each of our small-scale societies. While preparing the chapter on one of the populations in our sample—the Gusii—the researcher for that group (Laban Gwako) uncovered an anomaly in the household size data for this population. After an extensive investigation, it is now clear that the household data from the Maragoli (another population studied by Gwako) were mistakenly entered for the Gusii, and the Gusii household size data were accidentally destroyed. This investigation also confirmed that to the best of our collective knowledge, the rest of the Gusii data, and the project data more generally, were untainted by this data management error. on August 23, 2011 Our baseline regression analyses often used household size as a control variable, so losing the household size data for 1 of our 14 populations necessarily changes the precise coefficient estimates drawn from those regressions. To examine the implications of the loss of these data for our conclusions, we studied both (i) the same baseline regressions used in the main text with our Gusii household size data now missing, and (ii) the same regression with household size dropped as a control variable so that the Gusii reenter the analysis. Overall, for our theoretically relevant variables (market integration, participation in world www.sciencemag.org religion, and community size), the largest change by far occurs in predicting Ultimatum Game offers. Here, the coefficient on market integration increases from 0.098 to 0.14 (a 43% increase) when the Gusii are dropped entirely. This means that for an increase of 20% in calories purchased in the market, there is a predicted increase of 28 percentage points in Ultimatum Game offers, compared to only 20% from the prior analysis. The largest change in coefficients that works against the interpretations defended in our original

article is a drop in the coefficient on world religion from 5.96 to 5.77 (a decrease of 3%). Downloaded from Our reanalyses, which are available upon request, indicate that this loss makes no change to our interpretations (strengthening the effect of market integration) and has no material impact on significance testing. Our data set is available upon request and is in the process of being publically archived with the Inter-University Consortium for Political and Social Research (ICPSR) at the University of Michigan.

www.sciencemag.org SCIENCE erratum post date 17 JUNE 2011 1 COMMENTARY

Ethics and climate change Following the slightest movement 1382 1389 LETTERS I BOOKS I POLICY FORUM I EDUCATION FORUM I PERSPECTIVES LETTERS

edited by Jennifer Sills who might be needed elsewhere. If evacuation is maintained for too long, there can be serious health (including mental health), Disaster Preparation: Lessons from Japan family, and economic consequences. In rapidly evolving situations, scientists ON 11 MARCH 2011, THE EARTHQUAKE AND TSUNAMI IN JAPAN CLAIMED THOUSANDS OF LIVES, who advise decision-makers must think like disrupted the lives of hundreds of thousands, and destroyed regional power and transporta- clinicians and make decisions based on the tion infrastructure. Backup systems intended to maintain cooling for three operating nuclear best information available at the time. reactors and the spent-fuel pools failed. Experts in various technical fi elds from Japan, the C. NORMAN COLEMAN,1* STEVEN L. SIMON,2 United States, and around the world worked to understand and mitigate the consequences MICHAEL A. NOSKA,3 JANA L. TELFER,4 of unplanned releases of radioactivity. As a team of U.S. health and medical subject matter THOMAS BOWMAN5 experts deployed to Japan in the early days after the tsunami, we have identifi ed the elements 1National Cancer Institute and Offi ce of Preparedness and that remain most challenging in preparing for Emergency Operations, Assistant Secretary of Prepared- ness and Response, Department of Health and Human future disasters. Services, 200 Independence Avenue, SW, Washington, DC on August 23, 2011 Given the reality of the uncertainty in 20201, USA. 2Division of Cancer Epidemiology and Genet- the physical state of the Fukushima reac- ics, National Cancer Institute, National Institutes of Health, 3 tors, the ability to make accurate predictions 6120 Executive Boulevard, Bethesda, MD 20892, USA. U.S. Public Health Service, U.S. Food and Drug Administration, based on models was limited. We found that 10903 New Hampshire Avenue, Silver Spring, MD 20993, the absence of such information led to sus- USA. 4National Center for Environmental Health/Agency for picion by the public that data were not being Toxic Substances and Disease Registry, Centers for Disease Control and Prevention, 4770 Buford Highway, Atlanta, GA reported. Information was being provided by 30341, USA. 5U.S. Public Health Service, Division of Strate- a variety of sources, including international gic National Stockpile, Offi ce of Public Health Preparedness news media. Cultural differences and national and Response, Centers for Disease Control and Prevention, www.sciencemag.org sovereignty must be understood and respected 1600 Clifton Road NE, Atlanta, GA 30329, USA. when determining how to present, dissemi- *To whom correspondence should be addressed. E-mail: nate, and discuss information. [email protected] Support. The U.S. Navy delivers relief supplies to Although the public was justifi ably con- Note an evacuation center in northern Japan. cerned about possible radiation health effects, 1. The contents in this manuscript are from the authors and we believe that lack of understanding of radia- do not represent U.S. government policy or opinion.

tion and its effects resulted in unnecessary fear. The misunderstanding extended to the purpose Downloaded from of potassium iodide (KI), which should not be distributed indiscriminately, but rather used to India’s Education protect those with active thyroids, particularly infants, young adults, and teens who are at risk from the exposure resulting from internal contamination. There was also uncertainty about the Commercialization best KI dosage and administration schedule, partly because Japanese dosage guidance differed from that provided by the United States. Fortunately, in this case KI prescriptions were reduced P. BAGLA’S NEWS & ANALYSIS “FACULTIES to very few people, primarily those working close to the reactors. wither as higher education system rapidly We also observed differences in risk tolerance. Compared with those far from the site, expands” (29 April, p. 524) spells out the a local population may choose to accept a modestly greater risk from low-level radioactive problems with India’s expanding higher contamination in order to restore their economy and way of life. Cultural differences also education system, but it does not identify may alter the acceptable level of risk (such as the amount of iodine normally contained in the the root of the problem: India’s neglect of diet) or how to balance risk and benefi t. Less stringent requirements can be acceptable for the primary and secondary school education adults, compared with infants and children, although confl icting protective action guidelines systems. Much of India’s population can create public confusion. only afford to send their children to govern- When evacuation is initiated, it is crucial to identify the criteria that must be met before ment schools, which are mostly in shambles. residents are permitted to return. Each decision requires the assessment of competing risks The children of the wealthiest go to com- at a time when outcomes are uncertain. For example, initially imposing a travel warning and mercial private schools, which are expen- limiting entry of people and removal of food products from a 50-mile zone around the radia- sive yet still not of high quality relative to tion source were sensible. However, there are risks in mass movements of people, including truly private schools, such as nonprofi t pri- transportation accidents, trauma from disrupting lives, separating people from their normal vate schools funded by trusts. Perhaps the

CREDIT: MASS COMMUNICATION SPECIALIST 1ST CLASS BEN FARONE, US NAVY/WIKIMEDIA COMMONS US NAVY/WIKIMEDIA SPECIALIST 1ST CLASSFARONE, BEN MASS COMMUNICATION CREDIT: support structure (including care for chronic medical conditions), and diverting personnel only exceptions to this rule are the nonproﬁ t,

www.sciencemag.org SCIENCE VOL 332 17 JUNE 2011 1379 Published by AAAS LETTERS

trust-run private schools, which are few in Unless India bans commercialization of people’s behavior (1). Ricard et al. (2) con- number and extremely expensive because school education and adopts a public school cluded that personality was just as inﬂ uential they do not receive government grants. The system similar to that of the United States, with regard to behavioral outcomes in social low quality of India’s primary and second- France, and other countries, the problem that psychology experiments. Moreover, Miller’s ary education contributes in turn to the low Bagla describes in the News Focus story will story failed to highlight those individuals in quality of higher education in the country. persist. The present education system cannot Milgram’s study who refused to deliver any The Indian Parliament passed a Right of produce the teachers that India requires. extreme intensity shocks, let alone the maxi- Children to Free and Compulsory Educa- PUSHPA M. BHARGAVA mal shock (3). This percentage (5 out of 40) tion Act in August 2009, but it has already Manorama Ghar, Anveshna, Prashanti Nagar, Hyderabad, is roughly equivalent to the percentage of par- proved to be ineffective, acting only as a way Uppal 500 039, India. E-mail: [email protected] ticipants in public goods paradigms who con- to funnel public funds into private, commer- tinue to contribute when the rest of the group cial schools. (The act requires 25% of all does not (4), suggesting that individual dif- seats of private schools to be reserved for the Personality’s Role ferences and exemplary behavior will play poor and requires the government to pay the in Moral Action a substantial role in understanding morally fees, instead of devoting these government relevant actions. funds to improving public schools.) WE ENJOYED THE NEWS FOCUS STORY Zimbardo claims that there is “almost no “Using the psychology of evil to do good” research” on moral personality or character. Letters to the Editor (G. Miller, 29 April, p. 530) about Philip Yet, Samuel and Pearl Oliner interviewed and Zimbardo’s new Heroic Imagination Project profiled over 400 Holocaust rescuers from Letters (~300 words) discuss material published in (HIP) for promoting moral action. However, World War II (5). Anne Colby and HIP’s own Science in the past 3 months or matters of gen- eral interest. Letters are not acknowledged upon we feel compelled to correct some mislead- William Damon studied exemplars recog- 6 receipt. Whether published in full or in part, Let- ing impressions that could result from the nized by their local communities ( ), and work ters are subject to editing for clarity and space. story’s perspective. by Kristen Monroe highlighted characteris- Letters submitted, published, or posted elsewhere, As noted in the past, Zimbardo often tics shared by Holocaust rescuers (7). Quan-

in print or online, will be disqualifi ed. To submit a strongly emphasizes situationism, down- titative research by Lawrence Walker identi- on August 23, 2011 Letter, go to www.submit2science.org. playing the importance of individual dif- fi ed different kinds of exemplars that differ ferences, such as character, in influencing in personality profi les (8). Our own work in www.sciencemag.org Learn how current events are impacting your work. ScienceInsider, the new policy blog from the journal Science, isyour source for breaking news and instant analysis from the nexus of politics and science. Downloaded from Producedby an international team of science journalists, ScienceInsider offers hard-hitting coverage on a range of issues including climate change,bioterrorism, research funding,and more. Before researchhappens atthe bench, science policy is formulated in the halls of government. Make sure you understand how current events are impacting your work. Read ScienceInsidertoday. www.ScienceInsider.org

Breaking news and analysis from the world of science policy

1380 17 JUNE 2011 VOL 332 SCIENCE www.sciencemag.org Published by AAAS LETTERS

the HABITVS Project (Humane Archetypes: 5. S. P. Oliner, P. M. Oliner, The Altruistic Personality: fi rmed that to the best of our collective knowledge, the rest Biology, Intersubjectivity, and Transcendence Rescuers of Jews in Nazi Europe (The Free Press, New of the Gusii data, and the project data more generally, were York, 1988). untainted by this data management error. in Virtue Science) aims to link exemplary 6. A. Colby, W. Damon, Some Do Care: Contemporary Lives behavior in the real world to behavioral and of Moral Commitment (The Free Press, New York, 1992). Our baseline regression analyses often used household size neural data in the laboratory (9). Ongoing 7. K. Monroe, The Hand of Compassion: Portraits of Moral as a control variable, so losing the household size data for 1 work in the neuroscience of moral action will Choice During the Holocaust (Princeton Univ. Press, of our 14 populations necessarily changes the precise coef- Princeton, NJ, 2004). fi cient estimates drawn from those regressions. To exam- require increased attention to reach the worthy 8. L. J. Walker, J. A. Frimer, J. Personality Soc. Psychol. ine the implications of the loss of these data for our con- goals articulated by HIP (10, 11). 93, 845 (2007). clusions, we studied both (i) the same baseline regressions We are appreciative of Zimbardo’s new 9. M. Spezio et al., Annual Meeting of the Society for used in the main text with our Gusii household size data 16 Human Brain Mapping , 126 (2010). now missing, and (ii) the same regression with household direction, and we look forward to greater 10. J. Moll et al., Proc. Natl. Acad. Sci. U.S.A. 103, 15623 size dropped as a control variable so that the Gusii reen- understanding of the pathways of devel- (2006). ter the analysis. Overall, for our theoretically relevant vari- 11. A. Shenhav, J. D. Greene, Neuron 67, 667 (2010). opment that lead to excellence in moral ables (market integration, participation in world religion, character. MICHAEL L. SPEZIO,1* and community size), the largest change by far occurs in GREGORY R. PETERSON,2 WARREN S. BROWN,3 predicting Ultimatum Game offers. Here, the coeffi cient on KEVIN S. REIMER,4 JAMES VAN SLYKE3 CORRECTIONS AND CLARIFICATIONS market integration increases from 0.098 to 0.14 (a 43% increase) when the Gusii are dropped entirely. This means 1Department of Psychology, Scripps College, Claremont, Research Articles: “Markets, religion, community size, and that for an increase of 20% in calories purchased in the CA 91711, USA. 2Department of Philosophy and Reli- the evolution of fairness and punishment” by J. Henrich et market, there is a predicted increase of 28 percentage gion, South Dakota State University, Brookings, SD 57007, points in Ultimatum Game offers, compared to only 20% USA. 3Fuller Graduate School of Psychology, Pasadena, al. (19 March 2010, p. 1480). Authors’ note: Since the pub- CA 91101, USA. 4Department of Psychology, Azusa Pacifi c lication of our Research Article, we have been preparing a from the prior analysis. The largest change in coeffi cients University, Azusa, CA 91702, USA. book-length treatment of our comparative project for pub- that works against the interpretations defended in our orig- lication, with a chapter describing the work done in each of inal article is a drop in the co effi cient on world religion from *To whom correspondence should be addressed. E-mail: our small-scale societies. While preparing the chapter on one 5.96 to 5.77 (a decrease of 3%). Our reanalyses, which [email protected] of the populations in our sample—the Gusii—the researcher are available upon request, indicate that this loss makes no for that group (Laban Gwako) uncovered an anomaly in the change to our interpretations (strengthening the effect of References household size data for this population. After an extensive market integration) and has no material impact on signifi - investigation, it is now clear that the household data from the cance testing. Our data set is available upon request and is 1. M. B. Donnellan et al., APS Observer 20, 5 (2007). on August 23, 2011 2. F. D. Ricard et al., Rev. Gen. Psychol. 7, 331 (2003). Maragoli (another population studied by Gwako) were mis- in the process of being publically archived with the Inter- 3. S. Milgram, J. Abnormal Soc. Psychol. 67, 371 (1963). takenly entered for the Gusii, and the Gusii household size University Consortium for Political and Social Research 4. D. Houser, R. Kurzban, Am. Econ. Rev. 92, 1062 (2002). data were accidentally destroyed. This investigation also con- (ICPSR) at the University of Michigan. www.sciencemag.org

The complete Science archive 1880–1996 Downloaded from Fully integrated with Science Online (1997–today)

Available to institutional customers through a site license. Contact [email protected] for a quote.

Information: www.sciencemag.org/classic

www.sciencemag.org SCIENCE VOL 332 17 JUNE 2011 1381 Published by AAAS