RICH Economic Games for Networked Relationships ...

Article

RICH Economic Games for Networked Relationships and Communities: Development and Preliminary Validation in Yasawa, Fiji

Field Methods 1-17 ª The Author(s) 2016 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/1525822X16643709 fmx.sagepub.com

Matthew M. Gervais1,2

Abstract Experimental economic games reveal significant population variation in human social behavior. However, most protocols involve anonymous recipients, limiting their validity to fleeting interactions. Understanding human relationship dynamics will require methods with the virtues of economic games that also tap recipient identity-conditioned heuristics (RICHs). This article describes three RICH economic games—an allocation game, a taking game, and a costly reduction game—that involve monetary decisions across photos of one’s social network, integrating recipient identities while maintaining decision confidentiality. I demonstrate the ecological validity of these games in a study of male social relationships in a rural Fijian village. Deciders 1

2

Department of Anthropology, Center for Human Evolutionary Studies, Rutgers University, New Brunswick, NJ, USA School of Human Evolution and Social Change, Arizona State University, Tempe, AZ, USA

Corresponding Author: Matthew M. Gervais, Department of Anthropology, Center for Human Evolutionary Studies, Rutgers University, New Brunswick, NJ 08901, USA. Email: [email protected]

Downloaded from fmx.sagepub.com by guest on July 16, 2016

2

Field Methods

readily map these games onto daily life, and target earnings vary widely; consistent with ethnography, relative need is the primary rationale for decisions across the games, while both punitive and leveling motives drive reduction behavior. Consequently, altruism and spite are both elevated relative to anonymous target games in neighboring villages.

Methodological pluralism is a hallmark of anthropology, alongside holism, multivocality, and reflexivity. Among the newest methods in the anthropologist’s repertoire are experimental economic games. These methods complement interviews and observations by generating incentivized behavioral data for the comparative study of social norms and motives. Economic game studies in anthropology have typically employed two forms of anonymity: decider anonymity, in which decisions remain confidential (though see Hill and Gurven 2004); and target anonymity, in which participants make decisions toward unidentified targets (though see Rucas et al. 2010). Decider anonymity mitigates exogenous incentives, isolating social motives for altruism, fairness, or advantage (Camerer and Fehr 2004). Target anonymity ‘‘forc[es] players to default to local norms for dealing with people outside durable relationships’’ (Henrich et al. 2010:1482). With these design features, economic games have made groundbreaking contributions to the study of one-shot altruism. Using protocols standardized across diverse populations, evolutionary anthropologists have shown that humans ubiquitously engage in altruism and costly punishment in anonymous one-shot encounters, while population variation in such behaviors is predicted by market integration, religion, demography, and ecology (Ensminger and Henrich 2014; Henrich et al. 2004; Henrich et al. 2010; Lamba and Mace 2011; Marlowe et al. 2008). Considering human sociality more broadly, there are compelling reasons to extend economic game methods beyond anonymous interactions while preserving their utility for comparative anthropological research. Most significantly, the external validity of anonymous games—the generalizability of their results to other situations or samples—may be limited, because their ecological validity—their mapping onto naturalistic situations—may vary systematically across populations. Within the stark confines of economic games, participants must apply whatever interpretations and heuristics their own experiences provide (Hagen and Hammerstein 2006). While large societies furnish generalized norms and institutions to facilitate interactions among strangers, most human communities are patterned by diverse relational sentiments and norms conditioned on roles, statuses, and states,


Gervais

3

including cultural kinship (Chibnik 2005; Fiske and Fiske 2007). Accordingly, the most consistent population-level predictor of behavior in anonymous target games is precisely the presence of institutions that facilitate interactions among strangers (Henrich et al. 2010). Results from such games may well generalize to naturalistic behavior in impersonal interactions (Franzen and Pointner 2013; Nettle et al. 2011), but they often fail to generalize to richer social contexts (Gurven and Winking 2008; Wiessner 2009), especially personal relationships (Sønderskov 2011). Even in large societies, schematic identifying information about targets, such as their relative status, age, or social distance, greatly influences economic game decisions (Engel 2011). Games devoid of relational context may tap the generalized norms present in some societies, but they are ill-equipped to capture the diversity of social norms and relational motives across societies. The present study adapts economic game methods to investigate cooperation and punishment within enduring networked relationships, advancing the anthropological project of mapping and explaining behavioral variation across societies. I describe three recipient identity-conditioned heuristic (RICH) economic games designed to tap RICHs—social decision-making processes that integrate interpersonal identities, including endogenous sentiments toward particular people, and norms pertaining to others’ states, roles, and statuses. The games—an allocation game (AcG) similar to an N-recipient dictator game (DG), a taking game (TkG) similar to the ‘‘Social Strategies Game’’ (Rucas et al. 2010), and a costly reduction game (CRG) similar to an N-recipient punishment game—measure behavioral altruism, selfishness, and spite, respectively. These games preserve decision confidentiality, utilize monetary incentives, and are replicable across studies and sites. However, they depart from standard anonymous target games in two significant ways. First, they fully integrate recipient identities by presenting decision makers with photos of known recipients. Second, they entail parallel decisions toward an array of targets in a social network, mirroring the social trade-offs in resource allocations that characterize communities. Together, recipient identification and forced tradeoffs among potential recipients capture critical moderators of decision-making within social networks, improving ecological validity and affording the study of diverse RICH norms and motives. In the metaphor of Camerer and Fehr (2004), RICH games add color and depth to the line drawings of standard games to more fully characterize human social behavior. The data I report here are a descriptive subset of a larger study. I focus on (1) general patterns of decision-making within the three games; (2) general


4

Field Methods

patterns of target outcomes within the three games; (3) experiences of which the games remind deciders; and (4) the reasons deciders give for their game decisions. These data demonstrate the improved ecological validity of RICH games; Fijians readily map them onto their everyday experiences, and they use ethnographically salient decision heuristics to help or harm particular others. Resulting levels of both altruism and spite are higher than in anonymous target games in neighboring villages, and target earnings vary widely—underscoring the limitations of standard economic games for characterizing the RICH norms and sentiments that regulate human communities.

Methods Sample Fifty-four indigenous Fijian males (mean age 48, range 18–88) from one village on Yasawa Island, Fiji, participated in this study as pictured targets. Fifty of these 54 were also deciders in all three games, generating reciprocal dyadic data. These 54 men included members of all clans and households present during the study period. These were the first economic games played in this village, yet other villages in the same population, on the same island, have been well characterized using anonymous target games (Henrich and Henrich 2014), allowing some measure of comparison among their results. See the Supplemental Online Materials (SOMs) for additional ethnographic background and details about sample selection and representativeness.

Protocol All procedures were approved by the UCLA Office for the Protection of Research Subjects. I ran the three games in order over seven weeks, each taking about two weeks to complete—first the AcG, then the TkG, then the CRG. Participants played each game in isolation, usually in their own homes, with only me and a research assistant present. I paid participants FJ$2.00 at the time of participation in each game (FJ$1.00 ¼ US$0.55), making clear that this money was separate from the game stakes and could be kept even if they elected not to participate. Each game had about one day’s wage at stake (*FJ$20.00), either as potential earnings by the decider or as potential reduction of one target’s earnings. After participants made game decisions, we asked them a series of questions about their decision-making and their perception of that game.


Gervais

5

Participants left each interview with only the participation payment in hand; no coins were distributed at the completion of any interview session. A participant’s earnings across the three games—both as decider and as target—were paid out in private in one lump at the end of the study to obscure the decisions of others. Earning distributions were never made public, and we assured participants that their decisions and earnings were confidential. We also implored participants not to discuss the games or their decisions with others, describing this as a rule of each game. Following each game, we asked participants if they had heard about that game before playing, and none reported having heard anything. In the SOM, I discuss safeguards against ‘‘the Three Cs’’ in more detail. In all three games, a 54-cell array composed of three 18-cell boxes displayed standardized target photos (Figure 1). The order of the photos in the array was randomized on each day, and the three boxes were shuffled for each interview. To avoid biasing decisions, we demonstrated example distributions of coins on a model grid drawn to the same scale as the box cells; these demonstrations covered the full range of possible decisions and equity outcomes without making reference to any specific target. See the SOM for additional details about the equipment and protocol. AcG. The allocation game began with the decider’s own photo randomly positioned among those of 53 other men (targets), which we pointed out. Twenty FJ$1.00 coins sat in a separate cup. We instructed deciders that in this tavi (task), they could wasea (divide, distribute) the coins among themselves and the other pictured villagers in any way they wanted, with the money they placed on a photo really going to that person afterward. They were shown a range of possible distributions using the model grid (SOM). We then described the rules, including (1) only the men pictured could be allocated money and (2) they could not break the $1.00 coins into smaller units. TkG. The taking game began with eight FJ$0.05 coins on each photo ($0.40 per photo, $21.20 total) and the decider’s own photo in a separate cup. We instructed deciders that in this task they could taura (take) any number of coins from any of the photos, biuta (put, leave) those coins in their own cup, and maroroya (keep) that money for themselves. Any money they did not take from another’s photo was the amount that person would really receive afterward. Using the model grid and their cup, we demonstrated a number of possible decisions (SOM). We then explained a rule: Coins could not be redistributed across the photos, only taken or left.


6


Figure 1. Schematic representations of the three RICH games. The allocation game involved distributing 20 FJ$1.00 coins across a 54-photo array including the decider’s own photo. The taking game allowed taking up to eight FJ$0.05 coins from any of the 53 targets and placing them into the decider’s own cup. The costly reduction game provided 10 FJ$0.50 coins that could either be kept by decider or be used to buy red coins with which to reduce the total earnings of any target by FJ$2.00.

Gervais

7

CRG. The costly reduction game began with 10 FJ$0.50 coins in a plastic cup containing the decider’s own photo. A second box lid displayed 10 red tokens otherwise identical to the $0.50 coins and marked with ‘‘2.’’ We instructed deciders that in this task, they could keep up to FJ$5.00, or use any of that money to vakalailaitaka (shrink) or musu (break, reduce) other men’s total earnings from the three tasks. Specifically, they could volia (buy) up to 10 red coins and put them on other’s photos; each red coin would sau (cost) $0.50 that they would not get to keep, and each red coin they put on a target would reduce that target’s earnings by $2.00. We recounted the ways money could be earned in the three tasks. We then described an extensive set of examples (SOM) and reiterated a game rule: To reduce a man’s earnings with a red coin, they had to spend a $0.50 coin that we had provided and they could no longer keep this money.

Results Experiential Analogues of the Games for Participants Asked if the AcG reminded (vakananuma) participants of their experiences, the Fijian way of life, or village events, 38 (74.5%) said yes, while 12 (23.5%) said no, and one equivocated. Asked the same questions about the TkG, 38 participants (76%) said yes, while 12 (24%) said no. For the CRG, 35 participants (70%) said yes, while 15 (30%) said no. The SOM gives accounting of specific examples of analogized experiences.

Decisions AcG. In the AcG (N ¼ 51 deciders), the mean amount kept was 12.5% of the FJ$20.00 stake ($2.49 + $4.18 SD). For the modal keep, 43% of the men kept nothing for themselves; 76% kept 10% ($2.00) or less (Figure S1, SOM). Only one individual kept all $20.00 for himself. On average, deciders allocated money to 19% of possible targets (10.1 + 5.4), with a mean allocation to targets of $1.73 + $1.09. The max allocation to one target was $10.00. TkG. In the TkG (N ¼ 50 deciders), the mean amount taken by deciders was 33% of the FJ$21.20 available ($7.03 + $7.20). For the modal take, 20% of deciders took nothing for themselves; 12% took all available coins (Figure S1). On average, deciders took from 59% of the 53 targets (31.42 + 20.93). Among those targets taken from, the mean amount taken was 52.6% (4.21 + 2.61 coins out of eight, or FJ$0.21).


8

Field Methods

CRG. In the CRG (N ¼ 50 deciders), the mean amount spent to reduce targets was 45% of the FJ$5.00 stakes ($2.26 + $2.17). There were two modes: 34% of the deciders (17/50) reduced no targets, while 32% (16/50) spent all $5.00 on reductions. The remaining 34% spent some fraction of the stakes to reduce others. In other words, 66% of the deciders spent some or all of the stakes to reduce others (Figure S1). On average, deciders reduced 7.5% of the 53 targets (3.96 + 3.99). Among those targets reduced, the mean number of $0.50 coins used on each was 1.13 + 0.72. Of the 198 observed reductions, 184 (93%) used one coin, 9 (4.5%) used two coins, and 4 (2%) used three coins. One decider used all 10 coins on one target.

Order Effects There was no evidence of contagion of decider behaviors over the course of the study. Using separate Pearson’s correlations for each game, there was no relationship between the order of participation and the total kept in the AcG (r ¼ .07, p > .6), the total taken in the TkG (r ¼ .09, p > .5), or the total spent in the CRG (r ¼ .16, p > .2).

Target Outcomes AcG. In the AcG (N ¼ 54 targets), 96% of the targets were allocated money by at least one decider (Figure S2). The mean amount targets received from all deciders was FJ$16.54 + $22.71. The minimum received was $0.00, and the maximum was $104.00. Receipt was highly positively skewed, with a median of only $7.00. On average, targets received allocations from 19% of the deciders (9.54 + 10.38); the maximum number of deciders from whom one target received allocations was 41 (80%). TkG. In the TkG (N ¼ 54 targets), the mean amount targets had taken by all deciders was 33% (FJ$6.51 + $1.06). The minimum amount taken was 19% ($3.85), while the maximum taken was 49% ($9.60; Figure S2). On average, targets were taken from by 59% of deciders (29.09 + 4.80). The maximum number of takers was 36 (73%), while the minimum was 16 (32%). CRG. In the CRG (N ¼ 54 targets), 87% of targets were reduced by at least one decider. The mean amount targets were reduced by all deciders was FJ$8.30 + $12.05. The minimum reduction was $0.00, while the maximum was $76.00 (Figure S2). The modal reduction was one token ($2.00). On


Gervais

9

average, targets were reduced by 8% of deciders (4.00 + 4.21). The maximum number of deciders that reduced one target was 22 (44.9%).

Total Earnings The SOM describes intergame correlations for decider earnings and target outcomes. Aggregate participant outcomes from all game sources—keeping and receiving in the AcG, taking and being left in the TkG, and keeping and being reduced in the CRG—varied widely. The mean aggregate earned was FJ$33.13 + $30.84. The maximum earned was $118.15, while the minimum earned was negative $50.55. Four participants finished the games with negative aggregate earnings as a result of being reduced in the CRG more than they kept, took, received, and were left in other games. To help maintain confidentiality and good will, we paid each of these men a positive dollar amount that was less than the smallest total earned by any other participant. Each man who had paid to reduce one of these four was reimbursed in proportion to their contribution to the total of the target’s reduction that was not applied. All preceding descriptives are for prereimbursement totals.

Decision Rationales AcG. Target need was the overwhelming reason given for allocating coins to targets, with 92.2% of the deciders mentioning a target’s weakness (malumalumu), old age, lack of income, financial troubles, many dependents, widower status, general problems, or just wanting to help them. A lack of need was likewise the overwhelming reason given for not allocating to targets, with 72.5% of the deciders mentioning a target’s strength (kaukauwa), sources of income, or support from a large family. As other reasons for allocating, deciders mentioned a target’s goodness of character (25.5%), the quality of their relationship (11.8%), positive feelings toward them (11.8%), or their relatedness (7.8%). Two deciders (3.9%) mentioned an abstract value: doing the right (dodonu) thing. As other reasons for not allocating, five deciders (9.8%) mentioned their bad relationship with a target, and one decider (2%) mentioned their own need. Five out of the six other reasons were deciders saying they could not allocate to everyone for lack of coins. TkG. Target need was also the predominant reason given for not taking. Fortyfour percent of deciders mentioned targets’ needs (e.g., weakness, old age, lack of income) as reasons for not taking from them, while 38% mentioned targets’ ample resources (e.g., strength, youth, high income) as reasons for


10

Field Methods

taking from them. Eight deciders (16%) mentioned an abstract value or rule as a reason for not taking, with seven of these referring to preserving the equal distribution of the money at the start of the game. As other reasons for not taking, four deciders (8%) mentioned good feelings toward a target, including respect (dokai), and three (6%) mentioned their good relationship with a target, such as having been helped by them in the past. A decider’s own lack of need (2%), the good character of a target (2%), and their relatedness to a target (2%) were each mentioned by one decider as reasons for not taking. As other reasons for taking, six deciders (12%) mentioned an abstract reason, with most of these referring to taking some set amount from all so as to preserve equality across them. Other reasons for taking included a decider’s own need (8%), a target’s bad character (6%), their bad relationship with a target (6%), or their lack of relatedness to a target (6%). No deciders mentioned bad feelings as a reason for taking. Seven deciders (14%) gave no reasons for their decisions and simply stated it was their decision (vakatulewa). CRG. The most common reason given for reducing a target was the target’s ample resources (38%). However, unlike in the TkG, only 6% of deciders mentioned a target’s need as a reason for not reducing him. Other reasons for reducing in the CRG included a target’s bad character (16%) and a bad relationship with a target (10%). One decider mentioned reducing crosscousins (a joking relationship), while three (6%) mentioned abstract reasons, including wanting to equalize (vakatautauvatataki) earnings across participants. As reasons for not reducing targets, one decider mentioned their relatedness to a target, while 20 (40%) mentioned an abstract reason. Two of these made reference to not wanting targets to earn different amounts as a result of reduction, but most were referring to why all of their reductions were of the same size—they did not want selected targets to be reduced by different amounts. Four deciders (8%) gave no reason for their decisions other than ‘‘my decision.’’ A number of the other reasons justified not reducing by saying it would reduce the earnings of others (restating the game contingency), while one decider mentioned reducing others so that the money would ‘‘go back to where it came from’’ (i.e., to the author).

Discussion AcG In the AcG, essentially an N-recipient DG, deciders behaved quite altruistically; many allocated themselves nothing, and most kept only a fraction of


Gervais

11

the stakes. A previous DG study in neighboring villages on Yasawa Island (Henrich and Henrich 2014) used comparable stakes and found much less altruism—the mean keep was 65%, compared to 12.5% in the AcG. What explains this difference? Like the standard DG, the AcG maintains confidential decisions. However, the AcG departs from the DG in two important ways: There is more than one recipient, and target identities are known. Previous studies suggest that having multiple recipients does increase the amount allocated to each (Engel 2011). It could be that deciders view the AcG as a multirecipient DG in which fairness is determined by the amount Ego keeps relative to each of the other recipients. In the Yasawan dyadic DG, the modal offer was half the stakes. An analogous pattern in the AcG would involve Ego keeping the same amount as allocated to N recipients. However, only 22% of deciders in the AcG kept the same as they gave to all other recipients; 51% kept less than some or all recipients, and 43% kept nothing. In contrast, only around 5% of deciders in the Yasawan DG gave more than half of the stakes. Instead, target identities apparently drove decision-making in the AcG. Deciders were able to utilize all information associated with pictured targets, such as histories of interaction, reputation, and relational norms. Almost three-fourths (74.5%) of Yasawan deciders mapped the AcG onto their previous experiences, whereas only 45% (9/20) did so for the DG (Henrich and Henrich 2014). Specifically, many Yasawan deciders allocated money according to asymmetric need; over 90% invoked target need as a reason for allocating, and over 70% refrained from allocating to those not in need. Many deciders simultaneously mentioned old age, weakness, low income, and low access to wage labor in their explanations. This fits with previous DG studies showing that recipient need boosts generosity (Engel 2011). Need-based helping also fits with two ethnographically salient aspects of Yasawan life. One is the Christian ideology that pervades daily life in a Fijian village and that emphasizes compassion, helping, forgiveness, and community (Brison 2007). The second is the traditional system of need-based requests (kerekere) that operates among individuals, households, and clans in a Fijian village (Sahlins 1962). Generosity is integral to Fijian village life, and failure to participate in the kerekere system results in reputation damage and likely less support from others. A number of deciders mapped the AcG onto this aspect of village life, including caring for elders and sharing a fish catch.


12

Field Methods

TkG The results of the TkG were essentially the mirror image of the AcG. Among deciders, those who kept more in the AcG took more from others in the TkG (SOM). Target outcomes in the two games were highly negatively correlated: Having more taken in the TkG was associated with receiving less in the AcG. Target need was the predominant explanation for refraining from taking from targets, while a target’s ample resources were used to explain why they were taken from. The consistency of these results provides a measure of the convergent validity of these methods. Of interest, deciders in the TkG were restrained relative to comparable studies. For example, a similar design run among the Tsimane of lowland Bolivia (Rucas et al. 2010) found that Tsimane women took on average 75% of valued beads from other women. Men in Yasawa took only 33% of coins from other men. This difference is stark, hinting at large populationlevel differences in RICH game behavior. However, there are a number of explanations that the present study cannot disentangle, including culture, sex, and currency, which should be explored in future studies. Of note, 76% of Yasawan participants in this study mapped the TkG onto their everyday lives.

CRG Yasawan men were surprisingly willing to spitefully reduce the earnings of other villagers in the CRG. In an ultimatum game (UG), only 29% of Yasawan participants reject an offer of FJ$0.00, while in a third-party punishment game (TPG), only 33% of third parties spend to punish an offer of $0.00 (Henrich and Henrich 2014). In contrast, in the CRG, 66% of deciders reduced at least one target. One decider spent the entire pot to reduce a single target, while 10 men (20%) reduced the maximum of 10 targets. This was despite the absence of any target norm violation within the game protocol. Low rates of punishment in the anonymous target UG and TPG (Henrich and Henrich 2014) may be an artifact of uncertainty regarding whether the instigating acts—a low UG offer or a low TPG offer—do constitute norm violations. Across societies, norm violations are defined relative to the relationships of the involved parties (Edgerton 1985). In Fiji, legitimate status differences are pervasive, and unequal distributions of mundane and sacred resources are routinely conditioned on age, sex, and clan. Yet in anonymous target games, it is unclear how the proposer and receiver are


Gervais

13

related, or whether one might be in greater need. Yasawan second- and third parties in such games may well infer a relationship and its proprietary expectations from game behavior, rather than imposing generalized norms on observed behavior. In other words, given a low offer, antecedent entitlement or need may be inferred, rather than equality assumed (see also Cronk 2007; Gerkey 2013). In general, a higher percentage of Yasawan deciders map the CRG (70%) onto their lived experiences than either the UG (25%) or the TPG (36%; Henrich and Henrich 2014). Unlike these anonymous target games, the CRG potentially captures an observed dynamic of naturalistic punishment in a Fijian village: The secrecy of the CRG allows diverse punitive motives to be enacted without costs. Analogously, physical punishment in a Fijian village is not uncommon when excusable, for example, when a norm violator’s loss of standing undermines their protections from other villagers (Henrich and Henrich 2014), or as long as second-party retribution does not disrupt wider village affairs (Sahlins 1962). Not surprisingly, in the CRG, deciders reduced targets for many punitive reasons, including the target’s character, unpaid debts, and their ‘‘bad’’ relationship state. One decider even spoke of teaching his selected targets a lesson (me vaka nai vakamacala). Even more frequently—38% of the time—deciders spoke of reducing targets that had high income. Sometimes this involved character judgments such as the pejorative ‘‘money head’’ (ulu sede), but more often deciders simply referenced the target’s business or wealth. Several deciders provided specific leveling motives. For example, one man said, ‘‘I want to equalize (vakatautauvatataki) what we earn by putting red coins on those who earn much.’’ These data fit with recent experimental laboratory evidence of costly punishment in the service of egalitarian motives (e.g., Dawes et al. 2007) and may be the first quasi-experimental field data of this phenomenon. They also highlight the perceived illegitimacy of nascent income inequality in a population characterized by legitimate inequalities in traditional currencies of wealth and status such as clan membership and ritual privilege.

Limitations and Future Directions Among the limitations of this study is an all-male sample. I did this to include the entire available social network of one sex, specifically the one to which I had greater access for interviews, time allocation sampling, and participant observation. In such a highly gendered, unilocal society, men


14

Field Methods

and women might well behave differently toward members of their same and opposite sex, even while a previous study in Yasawa found no consistent sex effects on gameplay when recipients were not identified (Henrich and Henrich 2014). Focusing on female social networks (as in Rucas et al. 2010) and expanding the study to intersexual relationships are necessary and worthwhile directions for future research. Additionally, this study employed real Fijian coins as the objects of distribution. Money may have foregrounded particular norms and behaviors, limiting the external validity of the present method beyond economic transactions. However, several lines of evidence suggest otherwise. While currency effects have not been systematically studied, resources other than money have not clearly affected results in other experiments (e.g., Lamba and Mace 2011). Participants themselves also freely mapped the games onto more general norms of generosity and the division of nonmonetary resources such as fish. Of course, the payoffs in these games were monetary, independent of the objects distributed. While participants used many noneconomic rationales for their decisions, the economic leveling evident in the CRG may have hinged on the use of monetary payoffs. Future work can explore leveling behaviors using subsistence foods, purchased goods, or more symbolic capital that is unequally distributed in an ascribed hierarchy. Like most economic games, these RICH games involve stakes that arise as windfalls, without labor investment by participants. This may undercut their external validity (Gurven and Winking 2008). However, participants readily mapped these games onto everyday sharing practices involving both monetary and nonmonetary resources. It may be that windfall acquisition accurately describes the perception of resource acquisition in these villages due to stochastic weather events, frequent injuries, opaque illnesses, and unreliable employment. Future work should explore this possibility. Compared to economic games with anonymous targets, RICH games entail an additional layer of ethical considerations, commensurate with their value added in illuminating community dynamics and intrinsic social motives. By forcing participants to make trade-offs among known recipients, and by allowing them to exploit or reduce the earnings of other villagers, these games could conceivably introduce real-world conflict into a tight-knit community. I took several steps to mitigate this risk, most notably preserving decision confidentiality and obscuring the sources of earnings by making one final aggregate payout to each participant in private. Delaying payments was feasible because of rapport and trust developed over years working in this and neighboring communities; a similar protocol in the absence of such rapport might encounter skepticism and protest.


Gervais

15

Confidential decision-making does compromise ecological validity in one sense: Unlike the very public social decisions of day-to-day village life, there were no reputational or punitive consequences for actions in these games. However, I considered this a feature not a bug, both for ethical reasons and for investigating intrinsic relational motives, a primary goal of the larger study. While introducing public decisions into these games could have a large impact on the results and be of great theoretical interest, the risks would likely outweigh any benefits gained.

Conclusion These RICH economic games tap the norms and motives that regulate enduring social relationships in a Fijian community. The results fit with existing ethnographic accounts of Fiji, while revealing strong motives for economic leveling. In generating novel quantitative data, this study sets the stage for a larger comparative enterprise designed to describe and explain, ultimately and proximately, the patterning of social relationships as they structure and support human adaptation. Author’s Note The opinions expressed in this publication are those of the author and do not necessarily reflect the views of the John Templeton Foundation.

Declaration of Conflicting Interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by NSF DDIG #1061496 to MG and Daniel M. T. Fessler. Alan Fiske, Clark Barrett, Rob Boyd, Joan Silk, and Bailey House provided helpful feedback. Michelle Kline, Joe Henrich, and Joji Savou provided invaluable assistance in the field. The villagers of Yasawa Island, Fiji, were exceedingly generous in their participation. This publication was made possible through the support of a grant from the John Templeton Foundation.

Supplemental Material The online [appendices/data supplements/etc.] are available at http://fmx.sagepub. com/supplemental.


16

Field Methods

References Brison, K. J. 2007. Our wealth is loving each other: Self and society in Fiji. Lanham, MD: Lexington. Camerer, C. F., and E. Fehr. 2004. Measuring social norms and preferences using experimental games: A guide for social scientists. In Foundations of human sociality, eds. J. Henrich, R. Boyd, S. Bowles, H. Gintis, E. Fehr, and C. Camerer, 55–95. Oxford: Oxford University Press. Chibnik, M. 2005. Experimental economics in anthropology: A critical assessment. American Ethnologist 32:198–209. Cronk, L. 2007. The influence of cultural framing on play in the trust game: A Maasai example. Evolution and Human Behavior 28:352–58. Dawes, C. T., J. H. Fowler, T. Johnson, R. McElreath, and O. Smirnov. 2007. Egalitarian motives in humans. Nature 446:794–96. Edgerton, R. B. 1985. Rules, exceptions, and social order. Berkeley: University of California Press. Engel, C. 2011. Dictator games: A meta study. Experimental Economics 14:583–610. Ensminger, J., and J. Henrich, eds. 2014. Experimenting with social norms: Fairness and punishment in cross-cultural perspective. New York: Russell Sage Foundation. Fiske, A. P., and S. T. Fiske. 2007. Social relationships in our species and cultures. In Handbook of cultural psychology, eds. S. Kitayama and D. Cohen, 283–306. New York: Guilford Press. Franzen, A., and S. Pointner. 2013. The external validity of giving in the dictator game. Experimental Economics 16:155–69. Gerkey, D. 2013. Cooperation in context. Current Anthropology 54:144–76. Gurven, M., and J. Winking. 2008. Collective action in action: Prosocial behavior in and out of the laboratory. American Anthropologist 110:179–90. Hagen, E. H., and P. Hammerstein. 2006. Game theory and human evolution: A critique of some recent interpretations of experimental games. Theoretical Population Biology 69:339–48. Henrich, J., R. Boyd, S. Bowles, C. Camerer, E. Fehr, and H. Gintis. 2004. Foundations of human sociality: Economic experiments and ethnographic evidence from fifteen small-scale societies. Oxford: Oxford University Press. Henrich, J., J. Ensminger, R. McElreath, A. Barr, C. Barrett, A. Bolyanatz, J. C. Cardenas, M. Gurven, E. Gwako, N. Henrich, C. Lesorogol, F. Marlowe, F. Tracer, and J. Ziker. 2010. Markets, religion, community size, and the evolution of fairness and punishment. Science 327:1480–84. Henrich, J., and N. Henrich. 2014. Fairness without punishment: Behavioral experiments in the Yasawa Islands, Fiji. In Experimenting with social


Gervais

17

norms, eds. J. Ensminger and J. Henrich, 225–58. New York: Russell Sage Foundation. Hill, K., and M. Gurven. 2004. Economic experiments to examine fairness and cooperation among the Ache Indians of Paraguay. In Foundations of human sociality, eds. J. Henrich, R. Boyd, S. Bowles, H. Gintis, E. Fehr, and C. Camerer, 382–412. Oxford: Oxford University Press. Lamba, S., and R. Mace. 2011. Demography and ecology drive variation in cooperation across human populations. Proceedings of the National Academy of Sciences 108:14426–30. Marlowe, F. W., J. C. Berbesque, A. Barr, C. Barrett, A. Bolyanatz, J. C. Cardenas, J. Ensminger, M. Gurven, E. Gwako, J. Henrich, N. Henrich, C. Lesorogol, R. McElreath, and D. Tracer. 2008. More ‘‘altruistic’’ punishment in larger societies. Proceedings of the Royal Society B 275:587–92. Nettle, D., A. Colle´ony, and M. Cockerill. 2011. Variation in cooperative behaviour within a single city. PloS ONE 6:e26922. Rucas, S. L., M. Gurven, H. Kaplan, and J. Winking. 2010. The social strategy game. Human Nature 21:1–18. Sahlins, M. D. 1962. Moala: Culture and nature on a Fijian Island. Ann Arbor: University of Michigan Press. Sønderskov, K. M. 2011. Explaining large-N cooperation: Generalized social trust and the social exchange heuristic. Rationality and Society 23:51–74. Wiessner, P. 2009. Experimental games and games of life among the Ju/’hoan Bushmen. Current Anthropology 50:133–38.


Supplemental Online Materials (SOM) for RICH Economic Games for Networked Relationships and Communities: Development and Preliminary Validation in Yasawa, Fiji

Matthew M. Gervais Center for Human Evolutionary Studies Department of Anthropology Rutgers University School of Human Evolution and Social Change Arizona State University [email protected]

Ethnographic Background Yasawa Island lies in the northwest corner of the Fiji Islands, at the northern end of the Yasawa group, 65 nautical miles from the town of Lautoka on Fiji’s largest island, Viti Levu. Yasawa Island is 20 km long but rarely more than 2 km wide, with a maximum elevation of 400 m. The island is home to six indigenous Fijian villages averaging around 200 people; the smallest has less than 100 residents, the largest over 300. Sketches of Yasawa Island are given in Raven-Hart (1956) and Henrich and Henrich (2014). Photographs of Yasawa Island taken at the time of this study can be accessed at www.matthewgervais.net/photography. Linguistically, Fiji has some 300 dialects or “communalects” (Geraghty 1983), which fall roughly into two language groups, Eastern and Western, split by the central highlands of Viti Levu (Pawley and Sayaba 1971; Schütz 1962). The communalects of Yasawa Island clearly belong to Western Fijian and are closely related to dialects from the Ba district of Northwestern Viti Levu (Triffit 2000). The Yasawa Island language landscape is quite heterogeneous. There are slightly varying communalects among villages on Yasawa Island and increasingly divergent communalects as one travels away from Yasawa Island down the Yasawa island chain (Triffit 2000). The communalect of Waya, on the southern end of the Yasawa Group, may be the oldest and most distinct communalect in Fiji (Pawley and Sayaba In press). Yasawa, like indigenous Fiji generally, is prescriptively patrilocal, yet perhaps 10% ambilocal (Sahlins 1962), and exogamy sends many people to Yasawa Island from other islands in the Yasawas. An increasing proportion of Yasawa residents come from beyond the Yasawa region as well, as cross-cousin marriage prescriptions relax and individuals attending school or seeking wage labor on Viti Levu return to villages with spouses from as far away as Lau in Eastern Fiji.

To facilitate future economic opportunities, some Yasawan primary schools forbid the use of village communalects in class and require Standard Fijian or English. For these reasons, Standard Fijian is effectively a lingua franca on Yasawa Island, as it is across Fiji (Pawley and Sayaba 1971), and it is heard daily in conversations at all social scales, from family meals to village meetings and church sermons. Only the oldest (those over 80) and youngest (those under five) Yasawa Island villagers speak exclusively in their village communalect, although all villagers routinely deploy theirs in particular contexts. Village-level kinship is the primary mode of Fijian social organization (Sahlins 1962), and Yasawa is no exception. Each village is organized patrilineally and patrilocally, with extended households (itokatoka) composing clans (mataqali) that together constitute a yavusa, or territorial unit. A yavusa tends to have a single chief and is often, though not always, coextensive with a village. Within a village, clans are ranked by an historical division of labor, from the chiefly clan (turaga), and the “face of the chief” (matanivanua), down through priests (bete), kingmakers (sau turaga), warriors (bati), carpenters (matai), fishermen (gonedau), and other commoners (tauvanua). Today, most villagers are generalists, and many historical roles are enacted only at ceremonies. Rank nonetheless pervades village life, dictating terms of address, comportment, obligations, and precedence in seating, eating, and receiving shares of distributed goods (Toren 1990). Within clans, descent, age, and sex determine rank in a strongly patriarchal fashion. Village meetings are run by chiefs or their proxies and are animated by elders and other influential men from all clans. A chief holds the highest inherited rank within a well-defined traditional hierarchy. The office of the chief is greatly respected and is accorded public demonstrations of respect,

including spatial positioning, choice foods, serving order, honorific language, and elaborate funerary rites. The person holding the chiefly office is assumed to embody core Fijian ideals such as generosity, self-control, and respectfulness (Ravuvu 1983). However, an individual chief can lose the respect of the people, or an heir can fail to be installed as a chief, to the extent that he fails to embody these values. Chiefliness (vakaturaga), or village-oriented generosity and noblesse oblige, is a necessary condition for installation as a chief, with chiefly “blood” (drau) being insufficient. A chief has little opportunity to dictate to villagers and depends on prestige and consent for his influence. Yet once officially installed, a chief acquires the sanction of ancestor spirits and a degree of sacredness (Ravuvu 1987). The traditional economic role of a chief was to possess ample resources and distribute them generously based on need, and male villagers in Yasawa still gather once a year to plant their chief’s yam garden, from which all can draw in times of need. A chief was traditionally viewed as possessing immense mana, a kind of force of efficacy that permeated his possessions and was dangerous to the touch. Many Yasawans currently lament the general loss of mana by all Fijians, and it does not appear to play much of a role in legitimating chiefly influence in modern-day Yasawa. That said, the traditional lineage-based power structure of a village is still respected in the villages of Yasawa, and few question the right and responsibility of an exemplary chief to influence village affairs. A village chief is supposed to serve the good of the village, and it can fall to him to adjudicate disputes within the village when other options fail (Sahlins 1962). Each village also has a locally elected representative of the national government (turaga ni koro) who liaises with the government, enforces national laws (albeit loosely), and organizes

some community work. The controversies surrounding the Fijian central government feel very distant in Yasawa, and several water development, road improvement, hall construction, and cyclone relief projects have recently been implemented by the government, though to varying standards. Each village on Yasawa Island hosts two Christian denominations, Methodist and Assemblies of God, whose services give cadence to the week’s activities. The churches variously organize feasts, fundraisers, Bible study sessions, prayer groups, and rotating farm collectives. The congregations tend to cleave according to traditionalist/modernist sentiments. The Methodist church is invariably the older building, located on the central village green, and bound up with the traditional village hierarchy and prestige goods such as whale’s teeth (tabua). The Assemblies of God churches are often built on the fringes of villages and attract younger villagers and commoners through charismatic practices, entreaties to economic advancement, and proscriptions on kava drinking and smoking (see also Brison 2007; Ryle 2010; Tomlinson 2009). There are also three primary schools on the island, and almost all children attend school from around the age of five until 14. Students must leave the island to obtain secondary education, which is increasingly common, and tertiary education, which remains quite rare. Literacy is high and is reinforced by Bible study. Subsistence on Yasawa is primarily root-crop horticulture and fishing with spear, line, and net; littoral gathering, vegetable crops, and fruit trees supplement the diet. Pigs, chickens, goats, and cows are also kept and eaten, especially at ceremonies. Sea turtles are a particularly valued ceremonial food, and caches of eggs are eaten raw when found. Economic activities are highly gendered, with men doing most of the fishing and farming, and women doing much of the gathering and all of the cooking. Around 25% of calories come from purchased goods such as

flour, sugar, and rice (R. Boyd and J. Henrich, personal communication), which are stocked and sold from small canteens run by a handful of families in each village. However, supplies of these goods are unreliable, rendering subsistence activities indispensable, though risky. Cyclones (hurricanes) are common in Yasawa, with a significant tropical storm damaging crops and houses every several years. One village has relocated twice in the last 50 years owing to decimation by cyclones. The eye-wall of a Category 4 cyclone raked Yasawa Island in December 2012 and knocked down almost every traditional house on the island (which accounted for ~30% of dwellings), while destroying root crops for a year. Each village gets its drinking water from several large (~10,000 gallon) concrete tanks that collect rain water from the tin roofs of the church or community hall as well as from many smaller tanks attached to private tin-roofed homes. Yasawa is among the driest locales in Fiji, behind the rain shadow of Viti Levu, and, by the end of the dry season (April–September), many tanks are dry and crops can wither. Several villages also drink water of uncertain quality from wells. Extended households are the primary economic unit on Yasawa, organizing most fishing, farming, and resource pooling. Clans also pool resources for holiday feasts and organize labor for house building, net fishing, yam planting, and life course rituals. Village-level collective action occurs as well, for tasks such as cleaning the island road, planting the chief’s yams, conducting seasonally large fish drives, enacting cultural performances for tourists, and raising funds for schools. Church congregations crosscut kinship and also organize rotating farm work and fundraisers while hosting rotating intervillage services. Beyond the household, resource transfers occur primarily through two mechanisms (Sahlins 1962). First, ritual obligations among clans and villages—involving the exchange of

sacred objects, foodstuffs, woven mats, kerosene, and other valued goods—attend most life course rituals such as births, coming of age ceremonies, marriages, and funerals. Such exchanges are highly ritualized and strictly reciprocal, building social networks that can be tapped for cooperation and support (see Ravuvu 1987). Second, a system of ad hoc need-based requests (kerekere) operates among households, clans, and villages. Any social unit can kerekere any other for virtually anything—food, labor, money, land—on the condition that the initiator is in demonstrable need and that the potential donor has enough to share. Every recipient is obliged to share when they are themselves the target of a kerekere, yet a transfer requires no short-term reciprocity or account keeping, and often the flow of goods carries a net imbalance that follows persistent need. Being a generous provider is a sure path to prestige and influence, and generosity and compassion (components of “chiefliness”) are integral to the “Fijian way of life” (na bula vakaviti). While small canteens are present in each village, other local sources of income vary across villages. A number of villagers from one village and a small number of villagers from other villages work at the one resort on Yasawa Island, a highly insulated five-star operation near the largest village. Another village has a lucrative deal with a cruise ship company, owing to a popular limestone cave nearby. A third village had, at the time of this study, a relatively sophisticated sea cucumber–harvesting operation utilizing scuba gear subsidized by Chinese middlemen in Lautoka. Two villages have little economic alternative to selling fish and other produce to the resort. Most villagers must leave the island to find reliable income, and some villagers receive remittances from relatives working in Fijian towns or cities abroad. Most adult men, and many women, have spent some time working in resorts in the Yasawa group or in odd

jobs on the mainland. Copra is no longer pursued as an economic resource on Yasawa, as smallscale household operations have been excluded by larger-scale operations elsewhere in Fiji. Transport on the island is primarily by foot or horse. There are several trucks for hire, although the trucks are often in disrepair and the rutted dirt road that runs the length of the island is often impassable in the rainy season (November–March). Infrastructure is absent outside of the resort, and electricity comes almost exclusively from privately owned diesel generators that run for only a few hours some nights of the week. Most villages also have a nursing station with solar power, a refrigerator, and basic medical supplies, although the stationed nurse is often visiting the mainland. A cell phone tower was installed by a private company on the high point of the island in 2010, and cell phones are now widely owned, though infrequently charged, while the Internet remains accessible only to temporary residents (e.g., nurses, pastors, anthropologists) who bring computers and wireless modems from the mainland. There are outboard motorboats in all villages and they are the primary means of interisland travel, with indigenous boatbuilding skills having been lost. However, fuel is unreliably available, and larger cargo ships run only intermittently as far north as Yasawa Island, making transport to and from the island unpredictable. A high-speed tourist shuttle that runs daily up the Yasawa group to several dozen backpacker hostels stops short of Yasawa Island, and it is prohibitively expensive for most villagers to use. When used, one must arrange to be picked up by an outboard boat at the southern tip of the island south of Yasawa and then travel for an additional hour north. A small private plane operated out of the resort is also available, but is likewise prohibitively expensive. Trips to or from the mainland can be as quick as four hours, but more often take 15.

Methods Sample The sample for the present study came from one village on Yasawa Island. Seventy-two adult males resided in the village at some point during the study period (September 2011–June 2012) and participated in complete demographic and physical measures interviews. Of these 72 men, 58 were present to participate in an interpersonal attitude ratings task (February–April)—an inclusion criterion for the games—and were included in six months of time allocation sampling. Fifty-four of these men were present at the start of the Allocation Game (May) and were included as Targets. Fifty of these 54 remained available and participated as Deciders in all three games. One Target was unable to participate as a Decider due to vision loss and possible cognitive impairment, while another elected not to participate as a decider while remaining a Target. One Target was available for the Allocation Game but was then unavailable for the other two (hence the N=51 in the AcG), while the fourth took a job as a ship captain shortly after the start of the AcG. The 54 men included as Targets were representative of the starting 72 on all measured variables (e.g., demographics, reputation, attitudes toward), except that they had on average one less year of education. The men in this sample have life-long relationships with cross-cutting bases that include biological relatedness, diverse kinship norms, inherited and achieved status asymmetries, differentiated economic interdependencies, several church congregations, and individual variation in market integration. Appropriately measured and included as covariates, these variables can be pitted against one another as predictors of relational behavior. While a number of such covariates were gathered as part of this study, I do not report them here.

Protocol I administered these games with the assistance of an indigenous Fijian research assistant from elsewhere in Fiji who had been working in Yasawan villages with me for over a year. We displayed target photos in a 54-cell array composed of three 3 x 6-cell Paylak CTNB107 storage boxes with lid hinges notched for easy removal and replacement. This allowed us to transport the apparatus from house to house in a backpack with the photos already in their randomly assigned cells. It also allowed us to have the coins pre-distributed in the Target cells before the start of the Taking Game. We used one of the box lids to hold the model grid during demonstrations and another to display the reduction tokens in the CRG. The photos in each cell were headshots taken during the physical measures interview months before, standardized for scale and size (1.8" x 1.8") and laminated. Each cell was large enough to hold all the relevant coins in each game. In the Taking game, the external cup into which Deciders could place their take was larger than the array cells so as to accommodate the max take of 424 FJ$0.05 coins. For both the TkG and CRG, in which the Decider’s photo was in a separate cup, a white card marked with an “X” held the randomly assigned place of the decider’s photo in the array and was pointed out. We conducted interviews in Standard Fijian supplemented by conversational Yasawan. I first wrote all protocols in English. They were then translated into Standard Fijian by one English–Fijian bilingual research assistant (RA), then back-translated by another English–Fijian bilingual RA. I discussed any discrepancies with both RAs together and finalized the Fijian text. All Fijian translations are available on request. I selected the words to describe key elements of the games—wasea (divide, distribute) in the AcG, taura (take) and maroroya (keep) in the TkG, and vakalailaitaka (shrink) or musu (break, reduce) in the CRG—after extensive discussion with my research assistants and

consultation of several dictionaries (Capell 1991 [1968]; Gatty 2009). The goal was to use words that did not bias decisions toward particular distributive or normative patterns. The Fijian base wase (trans., wasea) is used generally to discuss dividing something into parts—such as a book by chapters, or a room by a wall—and also specifically the division of a resource across recipients, such as in ceremonial distributions of food or valued goods. However, unless additionally modified it implies no particular division, as into equal parts or according to need or status. In Fijian, taura is used generally with reference to taking or receiving something into one’s hands or possession, for example after division of a resource (complementing wase). It is also used metaphorically to speak of grasping the meaning of a statement. It is not, however, generally used with reference to receiving a requested resource, minimizing the mapping of this instruction onto need-based requesting, or kerekere. Maroroya is used to speak of keeping something as one’s possession, with connotations of cherishing, protecting, and being hesitant to relinquish. However, it does not imply selfishness or greed. Vakalailaitaka is a compound Fijian word composed of a base—lailai—meaning “small,” and two modifiers: vaka-, which means “in the manner of” or “possessing the property of,” and –taka, which means “to bring about” or “make happen.” Thus, meaning “to reduce” or “to make smaller,” vakalailaitaka is used generally with reference to any property or quantity. The Fijian base musu, which we used to clarify the object of vakalailaitaka in the course of describing the CRG, means generally to break or cut crosswise (as in a piece of wood), but recently has come to be used specifically in reference to reducing a quantity of money, as in a price or a wage. Like wase for distribution, musu (and the transitive musuka) does not imply a magnitude of reduction unless appropriately modified.

We performed recruitment and obtained informed consent separately for each game, and randomized participant order within each game. While participants made decisions, we moved away and sat with our backs to the participant. While I recorded decisions, my RA asked the debriefing questions. We informed participants that they would receive their earnings at the end of the study. At the time of final payment, after all the games, we also performed a memory check, asking participants to describe the three games they had played. All participants were able to describe the games and mentioned the keywords wasea (distribute) for the Allocation Game, taura (take) for the Taking Game, and musu (break, reduce) for the Costly Reduction Game. Training Examples In explaining each game, we provided each participant with a series of example decisions using the model grid. For the Allocation game, we first showed them a single cell containing first one, then two, then three, then 20 coins, and we explained the payoffs that each allocation would generate. Next, they were shown alternative distributions of all 20 coins: four coins in five cells; five coins in four cells; one, two, three, four, five, and five coins across six cells; one coin in 10 cells, and two coins in five cells; and 10 coins in one cell, nine coins in one cell, and one coin in another cell, with payoffs explained. For the Taking Game, we first used a single cell to explain the payoffs if participants took none, one, two, three, four, or eight coins from one target. Then, we explained the payoffs were they to take one coin from each of six targets (they each earn $.35, Ego earns $.30, everyone else earns $.40), one coin from each of seven targets ($.35, $.35, $.40, respectively), one coin from each of eight targets ($.35, $.40, $.40, respectively), one coin from each of nine targets ($.35, $.45, $.40 respectively), and one coin from every target ($2.65 for self, $.35 for everyone else). We then explained payoffs were they to take different amounts from different targets: one, two,

three, four, five, six, seven, and eight coins from eight different targets (earning Ego $1.80), four from 10 targets and eight from 10 targets (earning Ego $6.00), two from 10 targets, four from 10 targets, six from 10 targets, and eight from 10 targets (earning Ego $10.00), and taking all coins from all 53 targets, earning Ego $21.20. For the Costly Reduction Game, we explained that if a man was going to earn $20 from the previous tasks, and Ego left one red coin on his photo, he would instead earn $18; if he were going to earn $7, he would instead earn $5; and if he were going to earn $3, he would instead earn $1. We then demonstrated additional examples using the model grid: Buying one coin could reduce another target by $2 and leave Ego $4.50; buying two red coins could reduce two targets by $2 each and leave Ego $4; two coins could also reduce one target by $4 and leave Ego $4; three red coins would cost $1.50, reduce one target by $6, and leave Ego $3.50; five red coins would cost $2.50, reduce one target by $10, and leave Ego $2.50; and 10 red coins would cost $5, reduce one target by $20, and leave Ego no money in this task. We also demonstrated leaving different numbers of red coins on different targets: one coin to reduce one target by $2, two to reduce a second target by $4, three to reduce a third target by $6, and four to reduce a fourth target by $8, using 10 red coins and costing a total of $5, leaving none for Ego. We also described the different ways that a set number of coins could be used: Were they to buy seven red coins, costing $3.50 and leaving themselves $1.50, they could reduce seven targets by $2 each, or two targets by $4 and three targets by $2, or one target by $8 and one by $6, or one target by $14—any combination, their decision (mo lewa ga).

Postgame Interviews The specific questions we asked after each game were the following:

Allocation Game: 1) How did you decide to distribute the money as you did? 2) Why did you distribute the money to some men? 3) Why did you not want to give money to some other men? 4) Do you think others will make the same kinds of decisions as you? 5) How should [e dodonu me] someone distribute money in this task? 6) Did this task remind you of any of your life experiences? (or Fijian life? or village events?) Taking Game: 1) How did you decide to distribute the money as you did? 2) Why did you take money from some of the men? 3) Why did you not take money from some of the men? 4) Do you think others will make the same choices as you? 5) What should [e dodonu me] someone do in this game? 6) Did this task remind you of anything you have experienced in your life? (or aspects of Fijian life? Or village events?) Costly Reduction Game: 1) How did you decide to distribute (or not distribute) the red coins as you did? 2) Why did you place red coins on the men you did? 3) Why did you place different numbers of red coins on different men? 4) Do you think others will make the same choices you did? 5) What should [e dodonu me] someone do in this game?

6) Does this game remind you of anything you have experienced in life? (or aspects of Fijian life? Or village events?)

Results Experiential Analogues of the Games for Participants Allocation Game Twenty-six (51%) said the AcG reminded them of helping (vukei) others, including distributing (wasea) fish or money to those in need (leqa), looking after others (veinanumi), or having love (lomani) or compassion (loloma) for others. Five (9.8%) mentioned times of money trouble (leqa vakailavo), while four (7.8%) mentioned decision making with money. Three (5.9%) mentioned paying fees for schooling or travel, while two (3.9%) mentioned tithing. Five (9.8%) mentioned competition (veisisivi) or gambling. Taking Game Twenty-two participants (44%) said the TkG reminded them of helping others, showing love, joint production (solesolevaki), or taking care of others (veikauwaitaki). Two (4%) mentioned caring for elders, while four (8%) mentioned being responsible (vakayalomatua), or not being greedy (kocova). Eight (16%) mentioned facing financial difficulties, while one mentioned seeking help from others (kere veivuke). One mentioned cross-cousins (tavale) taking money from each other, while four (8%) mentioned duties to relatives (vakaveiwekani). Costly Reduction Game The CRG elicited a wider diversity of answers than the other two games. Eight (16%) mentioned sharing in some sense, but half of these (8% overall) mentioned difficult or heavy (vakabibi) decisions, not sharing, or unequal distributions. Only four (8%) mentioned helping those in need,

while three (6%) mentioned looking out for relatives. Five (10%) mentioned solving household problems (ni bula ni vuvale). Only one mentioned working together. Six (12%) mentioned experiencing financial hardship, while five (10%) mentioned being asked for money (kere ilavo) or losing money in business. One mentioned theft (“taking without knowledge”), one competition (veiqati), and one a test or trial (veivakatovolei). One compared the game to the “many ways” Fijians can be made equal (vakatautauvata).

Coding Decision Rationales Deciders gave a range of reasons for their allocation decisions. I compiled the open-ended responses to the first three questions for each game (relating to reasons for decisions) and used them to inductively generate coding categories. Recurring distinctions included aspects of self or target (e.g., need or character), relationship (e.g., quality, kinship), attitudes (e.g., love, respect), and generalized considerations (e.g., equality, fairness). Final categories, coded separately for both action and inaction (e.g., taking and not taking), included Ego need, Target need, Target character, Relationship quality, Kinship, Interpersonal feelings, Abstract values/rules, and two other categories: “My decision/desire” if that was the only reason mentioned, and “Other,” for a handful of idiosyncratic, uninterpretable, or uninformative answers (e.g., “Because I took money from them.”). Coding took the form of binary presence or absence of each code for each respondent, and allowed for the possibility of multiple codes per question per participant. I coded responses with reference to both the Fijian responses and English translations made by a bilingual Fijian RA. I trained an American anthropology undergraduate as a second coder through iterated coding of exemplars, and she coded all English translations. We then compared results and built consensus, using the agreed on coding as the final data. I calculated

interrater reliability for each game as the total number of “present” codes on which both raters agreed divided by the total number of “present” codes by either rater alone. Initial coding agreement between myself and the RA across the three games were 88.1% (AcG), 76.7% (TkG), and 72.2% (CRG).

Game Decisions Allocation Game Decisions On average, deciders allocated money to 19% of the possible targets. For comparison, 20 coins maximally distributed across 53 targets could have gone to 38% of them, while 17.5 coins (allowing for the mean keep of $2.49, or 2.5 coins kept) could have gone to 33% of possible targets. Decisions across Games The total amount Deciders kept for themselves in the AcG was significantly correlated both with the total amount they took from others in the TkG (r = .45, p . 40). Spearman’s rank correlation produced qualitatively similar results. Across the three games, decider earnings varied widely. The mean aggregate earnings from keeping in the AcG, taking in the TkG, and not spending in the CRG was FJD $12.31 ± $10.56. The maximum earned was $40.70. The minimum earned was $0 (N = 4), in which case each decider kept nothing (AcG), took nothing (TkG), and spent everything (CRG).

Figure S1:

Histograms of the FJ$ Deciders kept in the AcG, took in the TkG, and spent on reductions in the CRG.

Figure S2:

Histograms of the total FJ$ Targets received in the AcG, had taken from them in the TkG, and were reduced in the CRG.

Target Outcomes Allocation Game The number of allocations a Target received, and the maximum size of a received allocation, were highly correlated (r = 0.72, p < .0001). Taking Game A Target’s number of takers, and the number of large takes they suffered (i.e., of >4 coins, or >50%), were highly correlated (r = .69, p