[Paleopsych] Larry Samuelson: Foundations of Human Sociality: A Review Essay

[Premise Checker]

Larry Samuelson: Foundations of Human Sociality: A Review Essay
Journal of Economic Literature
Vol. XLIII (June 2005), pp. 488–497

1. Introduction

The scientists and engineers at the University of Wisconsin have periodic open 
houses. Thousands of visitors "ooh" and "aah" at the gleaming 
equipment--nuclear reactors, particle accelerators, electron microscopes, wind 
tunnels--and sophisticated experiments--cloning, fusion reactions, artificial 
hearts, and so on. It is no wonder that people walk away anxious to give money. 
Economists could surely benefit from putting on a similar display, but what 
would we show?

Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence 
from Fifteen Small-Scale Societies provides one answer. 1 It reports on a 
research program involving coordinated experiments in fifteen societies 
scattered around the world. This is a project as sophisticated as anything the 
scientists can produce. The fifteen societies are described in the title as 
"small-scale," but this is an

* Samuelson: University of Wisconsin. I thank the National Science Foundation 
(SES-0241506) and Russell Sage Foundation (82-02-04) for financial support.

1Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence 
from Fifteen Small-Scale Societies. Edited by Joseph Henrich, Robert Boyd, 
Samuel Bowles, Colin Camerer, Ernst Fehr, and Herbert Gintis. Oxford University 
Press, 2004.

understatement. They are remote, they speak a variety of uncommon languages, 
and they live in circumstances that make even basic data collection difficult. 
They fit the stereotype of societies that would be studied by anthropologists, 
and the editors are joined in the project by a team of (primarily) 
anthropologists.

Three questions lie behind this book. First, experimental economists have 
amassed a wealth of data in recent years, featuring considerable consensus in 
some respects. The bulk of the subjects in these data are university students, 
often American university students. Could the relatively consistent data be an 
artifact of a homogeneous subject pool, with a world of unexplored variety 
lurking beyond? Second, a growing number of economists have explored models in 
which people are concerned with more than simply their own material well-being. 
How might we model such "social" preferences, and how might we use experimental 
data to bring some discipline to what otherwise looks like a game with no 
rules? Finally, having designed such a research program, what do the data have 
to say?

Reflecting these questions, the book splits into two parts. One of these is 
chapter 3, written by Colin Camerer and Ernst Fehr. This chapter lays out a 
program for the experimental investigation of social preferences, centered 
around a series of games that allows one to evaluate and refine hypotheses 
about preferences and their induced behavior. The second part, comprising the 
bulk of the book, includes chapter 2, written by the six coeditors and Richard 
McElreath, and chapters 4–14. The latter report the field work done with the 
fifteen small-scale societies, emphasizing the economic experiments. Chapter 2 
ties the field work together with a description of the common elements as well 
as the inevitable differences.

2. Social Preferences: A Research Program

It will help organize the subsequent discussion to begin with the research 
program for studying social preferences outlined in chapter 3. Some terms will 
be helpful. Let us say that an agent’s preferences are personal if they depend 
only on the amount of money the agent receives.2 If not, then they are social. 
An agent’s social preferences can in turn be substantive, in which case they 
depend only upon the final allocation of (perhaps everyone’s) monetary payoffs, 
or can be procedural, in which case they also depend upon aspects of the 
process by which these payoffs were determined.3

The book offers a useful example of substantive social preferences, taken from 
Fehr and Klaus M. Schmidt (1999) (see Gary E.

2 More generally, an agent’s preferences are personal if they depend only upon 
the agent’s own consumption, but it suffices here to consider preferences over 
amounts of money.

3 The authors refer to procedural preferences as "reciprocal" preferences, 
carrying the connotation that agent i’s attitude toward payoffs may depend upon 
choices j has made, so that i’s behavior may reciprocate j's. This term fits 
well, but I prefer an alternative because reciprocal is used in the literature 
with a variety of different meanings. For example, the chapter in question 
refers to players "reciprocating cooperation" and "reciprocating defection" in 
the prisoners' dilemma, while explaining how the substantive preferences given 
by (1) (below) can explain such behavior, thus using the word in its common 
form in a setting where its technical meaning is inappropriate.

Bolton and Axel Ockenfels 2000 for a similar model). The utility ui(pi,pj) 
derived by player i from i's monetary payoff pi and j's monetary payoff pj is 
given by

.p ap ( -p if p=p (1) i (i, j )..i - ij i ) ij upp = .p ßp p if p= -p .ii (i - 
j ) ij

i =1. The special case
  =0 gives personal preferences.
  are positive, player i still likes high monetary payoffs, but now is also 
averse to inequality, being especially unhappy with inequality in which she 
receives the short end of the stick.

Chapter 3 notes that either substantive or procedural preferences are 
consistent with behavior observed in the experiments discussed in the chapter. 
However, the authors suggest that procedural preferences provide better 
descriptions of behavior in other experiments.4 In light of these remarks, it 
would have been helpful for the chapter to present an example of procedural 
preferences analogous to (1), and to have the experimental results discussed in 
this chapter explained in terms of this model.5

The research described in this chapter is centered around a series of five 
games that have played prominent roles in the experimental literature:6

The prisoners' dilemma. The point of departure is the prisoners' dilemma.7

4 For example, they refer (p. 83) to an "increasing number of experiments that 
compare predictions of competing theories," providing "clear evidence for 
reciprocity beyond inequality aversion . . . "

5 The chapter refers the reader to Matthew Rabin (1993) as well as other 
models.

6 The authors include two additional games, a public goods game that gives rise 
to incentives analogous to those of the prisoners' dilemma, and a gift exchange 
game that gives rise to incentives analogous to those of the trust game.

7 An example of a prisoners' dilemma is:

2 CD C 1 D 2, 2 3, –1 –1, 3 0, 0

The key characteristics are that defection (D) is a dominant strategy, while 
mutual cooperation (C) yields a superior outcome.

Personal preferences yield an unambiguous prediction, namely that players 
should invariably defect. Instead, laboratory experiments routinely find that 
some agents defect but others cooperate.8 There are manipulations of the 
experimental environment that make cooperation less frequent, but there is no 
experimental treatment that reliably produces universal defection.

If we are to account for this behavior while retaining the organizing principle 
of economics, namely that rational players make optimal choices guided by well- 
defined and stable preferences, then we must admit the possibility of social 
preferences: some people prefer to cooperate in the prisoners' dilemma. In 
terms of (1), we thus have evidence that for at least some
  > 0.9 For cooperation to be optimal, however, an individual must not only , 
but must also think the opponent is sufficiently likely , or are they simply 
pessimistic about the prospects that their opponents will cooperate, while 
standing ready to cooperate against sufficiently cooperative opponents?

The ultimatum game.

In response, attention turns to the ultimatum game.10 Personal preferences, 
coupled with sub- game perfection, call for player 1 to offer player 2 nothing 
(or at most the smallest possible positive amount) and for player 2 to accept. 
Conditional on player 2 always accepting, player 1's choices should provide . 
Larger values of correspond to larger offers to player 2.

8 See John O. Ledyard 1995 for a survey.

  > 0 is necessary for C to be preferred to D. The interpretation is that such 
an agent prefers the more equal (but personally less lucrative) outcome of 
mutual cooperation to the more asymmetric payoffs produced by defecting against 
a cooperator.

10 Player 1 proposes a division of a sum of money to player 2, who either 
accepts the division, in which case it is implemented, or rejects it, in which 
case both receive nothing.

In experiments, player 1 often offers a substantial portion of the money to 
player 2, while player 2 often rejects smaller offers.11 The latter behavior 
provides evidence that > 0 for some players, so that an equal outcome of no 
money for both is preferred to receiving a relatively small amount of money. 
However, this player-2 behavior obscures the lessons concerning player 1's . 
Does player 1 offer money to player 2 because 1 prefers not to have too 
unequally large a share, or because player 2 rejects small offers?

The dictator game.

To answer this question, the authors direct attention to the dictator game.12 
Personal preferences call for player 1 to retain all of the money. However, 
player 1 typically offers some money to player 2, but less than in the 
ultimatum game (Roth 1995). The positive offers again provide evidence for 
positive , but the comparison suggests that 1's behavior in the ultimatum game 
is a mixture of concerns for player 2's payoff and concerns about being 
rejected.

The trust game.

Attention next turns to the trust game.13 Personal preferences, again coupled 
with subgame perfection, call for player1 to retain everything, sacrificing the 
efficiency gains of making a contribution to player 2. Experimental findings, 
with significant variation, are that player1 typically contributes about 
halfof1'sendowmentto2, who returns a sum leaving player 1 slightly worse off 
than if1had contributed nothing.

Given that player 1's behavior reflects a mixture of preferences and 
expectations about 2's behavior, as in the ultimatum game, the clearest 
implications for preferences

11 See Werner Güth, Rolf Schmittberger, and Bernd Schwarze (1982) for the 
initial study and Alvin E. Roth (1995) for a survey.

12 Player 1 divides a sum of money between player 1 and 2.

13 Player 1 and 2 each receive an endowment of S. Player 1 first decides how 
much of S to give to player 2, retaining the rest. The money given to player 2 
is tripled, at which point player 2 decides how much money to return to player 
1.

14 The preferences given by (1) are linear in payoffs, as is the feasible set 
facing player 2. Taken literally, (1) then implies that player 2 should 
(generically) either return none of the money or equalize payoffs, but 
generalizations give interior solutions. come from player 2's behavior. The 
fact that player 2 returns anything at all provides > 0. However, the data pose 
puzzles for both substantive and procedural preferences. Player 2 fills the 
position of a dictator, with the amount to be divided depending upon player 1's 
first move. In this dictator portion of the trust game, the final payoff 
typically allocated to player 1 does not increase as the amount to be divided 
increases. One might have expected preferences such as those captured by (1) to 
call for the player 1's payoff to increase.14 Perhaps the answer is that 
procedural considerations are important here. The second move in a trust game 
is not a de novo dictator game, but a game that follows a choice on the part of 
player 1 that affects player 2's preferences. The difficulty here is that as 
player 1 contributes more, and hence acts so as to enhance the welfare of 
player 2, player 2 "reciprocates" with actions that decrease player 1's 
relative payoff, leaving player 1 no better off and appearing to put increasing 
emphasis on player 2's payoffs. This is an unusual brand of reciprocity.

The trust game illustrates how subtle the match between experimental results 
and simple economic models can be. In light of this, a more detailed discussion 
would have been helpful, including a demonstration of how the substantive model 
given by (1) and an analogous procedural model might match behavior.

Punishment games.

People may have social preferences, but are the departures from personal 
preferences large enough to have real effects on behavior? The point behind the 
punishment games described briefly in this chapter is that seemingly small 
effects in preferences can be leveraged into large effects in behavior. Suppose 
that there is an accepted standard of behavior, and that people have the 
opportunity to sanction those who stray from this standard. These sanctions may 
have a large cumulative effect on their target while imposing relatively small 
costs on those doing the sanctioning. Experiments show that people will impose 
such sanctions, even against those whose actions have no direct payoff 
consequences for the sanctioner, and that the sanctions can have significant 
effects on behavior (Fehr and Urs Fischbacher 2004, Fehr and Simon Gächter 
2000, 2000, 2002).

3. The Experiments

Chapter 3 offers a template for the study of social preferences. The second and 
larger part of the book pursues this program, reporting on experiments with 
fifteen groups of people. The following table identifies the groups with whom 
the experiments were done, their location, and the researchers who did the 
experiments and wrote the relevant chapters. Every study involved the ultimatum 
game. In some cases, experiments with one or more of the public-goods game, the 
trust game, or the dictator game were also performed. The most striking aspect 
of the experimental work is the chance to compare ultimatum-game behavior 
across a wide range of settings, and I'll concentrate on the ultimatum game.

The authors organize their results around five themes.

Variability. The behavior observed in the fifteen societies exhibits more 
variation than found in the familiar experimental literature based on 
university students. Mean offers in ultimatum-game experiments with university 
students tend to lie between 40 and 50 percent. Among the fifteen societies 
studied here, the means range from 25 to 57 percent. Most existing experiments 
have produced modes at 50 percent, while modal offers here range from 15 to 50. 
Rejection behavior is similarly varied. Four groups (Kazakhs, Ache, Quichua, 
and

Chapter Group Country Researcher

4 Achuar Ecuador John Q. Patton

4 Quichua Ecuador John Q. Patton

5 Machiguenga Peru Joseph Henrich and Natalie Smith

5 Mapuche/Huinca Chile Joseph Henrich and Natalie Smith

6 Hadza Tanzania Frank Marlowe

7 Tsimane' Bolivia Michael Gurven

8 Au Papua New Guinea David P. Tracer

8 Gnau Papua New Guinea David P. Tracer

9 Kazakhs Mongolia Franciso J. Gil-White

9 Torguuds Mongolia Franciso J. Gil-White

10 Shona Zimbabwe Abigail Barr

11 Sangu Tanzania Richard McElreath

12 Orma Kenya Jean Ensminger

13 Ache Paraguay Kim Hill and Michael Gurven

14 Lamalera Indonesia Michael S. Alvard


Tsimane') rejected no offers (with sample sizes from 10 to 70), even though in 
two of these cases about half of the offers were for less than 30 percent of 
the surplus. Another group (Machiguenga) rejected only one offer (out of 21), 
though 75 percent of the offers were for less than 30 percent of the surplus. 
Two of the groups (Au and Gnau) often rejected offers of more than fifty 
percent, and appeared to be just as likely to reject high as low offers. In 
other groups, rejections were relatively frequent, especially of low offers.

Preferences. The prediction of subgame perfection with personal preferences, 
that the proposer offers virtually nothing and the responder accepts, does not 
provide a good match for the data. As one might expect, given the variability 
in behavior, there are aspects of the data that appear to be both closer to and 
further from this benchmark than the bulk of the existing literature.

Intergroup differences. The authors construct measures of two characteristics 
for each group. A group is deemed to have higher potential gains from 
cooperation if productive activities in the group are more likely to require 
interaction with nonrelatives. A group is deemed to have higher aggregate 
market integration the more often its people engage in market exchange, the 
larger its settlements, and the more complex its political structure. A 
regression suggests that mean ultimatum-game offers are higher for those groups 
exhibiting a higher potential benefit from cooperation and higher aggregate 
market interaction.15

These results nicely exploit the strength of the research design and data, 
namely the variability in groups and behavior. At this point, it would be 
helpful to have more attention devoted to how one might interpret the link 
between aggregate market integration and ultimatum-game behavior. If asked for 
the direction of this link before seeing this book, I'm not sure which I would 
have predicted. I can imagine market

15 A standard deviation increase in either variable gives about half a standard 
deviation increase in mean offer.

integration making people more likely to treat others "fairly" or more likely 
to induce them to drive and sometimes accept hard bargains.16 Some additional 
scrutiny and perhaps modeling would be helpful in interpreting the correlation.

Intragroup differences. A variety of individual characteristics, such as sex, 
age, wealth, education, market participation, and others, showed little 
relationship to behavior in the experiments.

Everyday life. The authors note that one can find parallels between the 
behavior of various groups in the experiments and their everyday life. For 
example, foragers who routinely share the meat they catch (the Ache), to the 
point that a successful hunter often takes none of the catch, sometimes 
declining even the credit for making the catch by leaving it outside the 
village to be found anonymously, made quite generous offers in the ultimatum 
game and generated no rejections. Foragers who appear to share only under 
duress (the Hadza), often attempting to sneak their catch into camp unseen, 
made smaller offers that often generated rejections.

The links between everyday and experimental behavior are especially intriguing. 
First, they suggest a view of preferences as being shaped by one's culture and 
way of life, a suggestion that runs throughout the book.17 Among the clearer 
examples, John Patton (pp. 121–22) suggests that the Achuar tend to be involved 
in more stable interactions in their daily lives than do the Quichua, which may 
be reflected in different perceptions of what behavior is fair or acceptable, 
and hence different behavior in the ultimatum game. Richard McElreath (pp. 
350–51) offers a similar explanation for

16 Jean Ensminger's contribution (pp. 356–357) provides a brief discussion of 
ideas suggesting that market participation may either enhance or attenuate 
tendencies to be fair.

17 Henrich (2004) presents a model in which cultural evolutional shapes social 
preferences.

behavioral differences between Sangu herders and farmers. Michael Gurven (p. 
227) closes his discussion with a call for further investigation of such 
cultural foundations of social preferences.

At the same time, this link raises issues in interpreting the experimental 
results. For example, the Orma are said (pp. 38–41) to have readily recognized 
the public goods game as a "harambee" game, referring to the contributions they 
make when constructing public goods such as roads or schools. However, the 
harambee itself appears to be best modeled as a repeated relationship, unlike 
the one-shot nature of the experiment. If subjects responded to the experiment 
with behavior appropriate for the repeated environment of the harambee, then 
their behavior may tell us little about their preferences. We return to this 
issue in the following section.

The experiments provide a wealth of material for those interested in studying 
experimental methods. How does one attempt standardized experiments when 
dealing with twelve researchers in fifteen societies spread across the globe in 
challenging settings? As the authors note, a variety of compromises had to be 
made. Some actual ultimatum-game offers were supplemented with sham offers to 
generate more variation. Some sample sizes were small. The game was typically 
presented in the abstract, but sometimes with the help of analogies to concrete 
situations. Subjects were typically paid in money,but not always. How does one 
control the variation inevitable in oral presentations? How does one ensure 
that the various settings have not introduced framing effects that swamp other 
considerations? How does one work, not only without the sophisticated computer 
interfaces typical of experimental laboratories, but with subjects who may not 
be able to do even simple arithmetic?

The authors are aware of these difficulties. I think it a sensible response to 
note that if one waited for perfect procedures and perfect answers to all of 
these questions, no study of this type would ever get off the ground. It is 
clear from the recurring discussions of method in the book that the authors 
worked hard in quest of the best experimental practice they could achieve, and 
worked hard to control and standardize their procedures. This is as much as one 
can ask.

4. Discussion

The book provides clear answers, sketched in the previous two sections, to the 
first two motivating questions raised in section 1. There are ways to 
systematically use experiments in investigating social preferences. There is 
more variety in behavior than previous experiments might lead us to believe. 
The third question is more challenging. What do the results tell us about the 
nature of people's social preferences? Using (1) to organize the discussion, 
what do the ? In particular, do the experimental observations imply that we 
should rethink our economic models, in which personal preferences currently 
play a prominent role, to allow more room for social preferences?

The answer appears to be yes, if the games played by the subjects are a good 
match for the ultimatum game that appears in our theoretical models. This 
qualification, however, is important. A key feature of the ultimatum game is 
that it involves anonymous opponents with no future interaction. If the 
participants are concerned with the implications of current play for future 
behavior, perhaps in the form of an effect of current play on their reputation, 
then the simple link between behavior in the ultimatum game and preferences is 
broken.

How might such a concern with the future or a feeling of nonanonymity arise? 
The subjects in the experiments reported here are typically drawn from small 
societies, with whom they know they will have subsequent interactions, often 
while living with virtually no privacy. Despite the best attempts of the 
experimenter to make experimental play

18 Notice that this possibility differs from a maladaption account, in which 
evolution is said to have neglected to equip us even with an understanding of 
one-shot interactions (because we purportedly evolved in an environment in 
which the norm was repeated interactions with small groups of primarily 
relatives). Fehr and Henrich (2003) and Henrich (2004) argue that such 
maladaption does not provide an adequate model of social preferences. Instead, 
we are considering here the possibility that subjects may understand the 
implications of one-shot interactions, but not view the experiments as such 
interactions.

anonymous, the subjects may have perceived, perhaps correctly, that their 
current play would have future repercussions.18 This possibility is raised 
several times in the book. Gurven (p. 221) suggests that rejections in the 
ultimatum game may have been relatively rare because subjects viewed rejection 
as giving rise to costs or punishments in subsequent interactions. Ensminger 
(p. 358) notes that anonymity may be impossible to achieve in small-scale 
societies. Hill and Gurven suggest (p. 403) that their subjects live in a 
sufficiently small community that aggressive behavior in the game could be 
deterred by the attendant negative impact on subsequent community relations, 
and note (pp. 406-07) that subjects appeared to treat the public goods games as 
part of an iterated sequence of social interactions.

These difficulties are exacerbated by the challenges of getting the 
experimental subjects to understand the abstract experimental environment. 
Patton (p. 105) indicates that the experiment was described to the potential 
subjects as a minga, or cooperative labor exchange. This may have helped in 
many respects, but a minga entails future obligations (p. 101), potentially 
introducing the ideas of repetition or lack of anonymity into subjects' views 
of the experiment. As we have noted, Ensminger (p. 376) reports that her 
subjects immediately recognized the public goods game as a harambee, a 
village-level pubic goods contribution process that again potentially 
introduces elements of repetition or the lack of anonymity. It does not appear 
as if one can explain all of the data by simply muttering "they acted

Samuelson: Foundations of Human Sociality: A Review Essay 495 19 Ensminger 
follows her comment with the observation that something more appears to be at 
work. Hill and Gurven note that the shadow of the future was most pronounced in 
a version of the public goods game in which contributions were made publicly, 
an indication that subjects understand that the future is sometimes more 
important than others. as if the game was repeated."19 However, as various of 
the authors' comments indicate, the possibility that such considerations played 
a role in shaping play is heightened by the nature of the subject populations, 
and makes it difficult to identify the games and the forces that shaped the 
subjects' behavior. This in turn makes it difficult to draw inferences from the 
observed behavior. Why is this a problem? Everyone recognizes that people 
sometimes do things-- make anonymous contributions, render aid to strangers, 
chastise transgressors--that are most readily modeled with social preferences. 
Differences of opinion arise in assessing the importance of social preferences 
in explaining economic behavior, and it is here that experiments can be 
especially valuable. But if the behavior in the experiments depends importantly 
on features beyond the experimental design, then conclusions concerning the 
nature and importance of social preferences are difficult to draw. This 
difficulty becomes more pronounced if we consider procedural preferences rather 
than the substantive preferences of (1). Not only are outcomes now important, 
but so is the process by which these outcomes are achieved. In addition, we 
cannot expect the experiment to control all of the details of this process. For 
example, Gurven (p. 226) notes that preferences for how resources are to be 
allocated may depend importantly on how the resources are made 
available--whether as the result of a windfall, for example, or as a result of 
having been earned. How do the experimental subjects perceive the surplus, and 
what effect does it have on their behavior? Preferences now depend upon the 
game and the context in which the game is played. The experiment can control 
the former, but we can expect much of the latter to be supplied by the 
subjects. A recurring theme throughout this book is that behavior in the 
experiments reflects the subjects' behavior in their everyday lives, as one 
would hope if the experiments are to tell us about the preferences that guide 
the subjects' lives. However, running along with this is the theme that the 
experimental behavior may be borrowed from a variety of real-life situations, 
not all of them an appropriate match for the setting one typically hopes to 
capture with the ultimatum game. Henrich and Smith (p. 164) suggest that 
behavior in the ultimatum game depends upon which of a diverse set of 
behavioral rules is triggered by the experimental implementation. McElreath (p. 
344) suggests that different players may map the ultimatum game into different 
social experiences and hence effectively play different games. Patton (p. 98) 
suggests that ultimatum-game behavior reflects a logic of reciprocal fairness 
that is not well suited for the ultimatum game, but also not easily overruled 
by a conscious understanding of the game. Tracer (p. 255) suggests that his 
players "inescapably brought the understandings, beliefs, expectations, and 
values that they apply to daily life into the experiment," including a belief 
that current actions would incur future obligations. Hill and Gurven (p. 403) 
offer a similar interpretation. Gurven (p. 226) notes that people can be more 
or less likely to cooperate depending upon the type of cooperation required--a 
group may readily cooperate in hunting or fishing, but not in 
conservation--making it difficult to make the link from experimental behavior 
to any particular real-life behavior. We thus have two related obstacles 
standing between the observed behavior and conclusions about social 
preferences. First, the subjects may have viewed their behavior as having 
future implications. If so, then subjects with personal preferences could 
exhibit behavior that could only be rationalized by social preferences in a 
literal implementation of the ultimatum game. Second, social preferences, 
especially procedural social preferences, may call into play a variety of 
factors beyond those captured by the ultimatum game model and the surrounding 
experimental design. Distinguishing between personal and social preferences is 
thus not straightforward.

These difficulties are not new. The existing experimental literature is full of 
experiments that are interpreted by some as being consistent with social 
preferences and others as reflecting personal preferences, with the differences 
frequently revolving around the connection between the experimenter's model of 
the strategic interaction and the subjects' model of the interaction. Indeed, 
this diversity of interpretations and conclusions is reflected in miniature in 
the studies in this volume.20 One can conclude that the experimental findings 
in this volume are consistent witha modelin which social elements play an 
important role in people's preferences and behavior, but also that they are 
consistent with a model in which behavior is guided primarily by personal 
preferences.

Further work is required to distinguish these alternatives. Work that moves us 
outside the traditional experimental economics laboratory, like the experiments 
reported here, is likely to be especially useful. In this respect, perhaps the 
most valuable part of this book is its analysis linking experimental play to 
group characteristics such as the importance of cooperation and degree of 
market integration. Once again, the results lend themselves to multiple 
interpretations. It could be that cultural evolution has led different groups 
to different configurations of social preferences that reflect differences in 
their ways of life. It could also be that subjects from different groups 
associate the experiment with different experiences from

20 As the editors note in their introduction to this volume (p. 7), they made 
no attempt to force a party line on the studies.

their contrasting ways of life, prompting different behavior from personal 
preferences.

Comparative static studies of this kind are likely to hold the key to 
separating the various hypotheses and getting a better idea of preferences.21 A 
single study cannot be expected to resolve every question, even a study as 
large as this one. But it has enriched the discussion and opened new directions 
for research, directions that hold great promise and that one hopes become 
standard.

5. Conclusion

This is a fascinating book to read on many levels. It provides a captivating 
window into the world of anthropologists.22 It is impressive both in the goals 
the authors are willing to set for an experimental study and in the lengths 
they are willing to go in pursuit of these goals.

We have much to learn from these experiments, most notably that humans exhibit 
a richer variety of behavior than many of us have imagined. The authors 
indicate they are working on additional experiments that will allow them to 
make use of what they have learned in this study and to fill in some more 
pieces of the puzzle. Readers of this book will surely look forward to the 
next.

REFERENCES

Bewley,TrumanF. 1999. Why Wages Don't Fall During a Recession. Cambridge: 
Harvard University Press. Bolton, Gary E., and Axel Ockenfels. 2000. "ERC: A 
Theory of Equity, Reciprocity, and Competition."

American Economic Review, 90(1): 166-93. Fehr, Ernst, and Urs Fischbacher. 
2004. "Third Party

21Truman F. Bewley (1999) offers another promising direction for such work, 
seeking evidence that considerations such as fairness and morale play a role in 
workplace behavior, as does the growing body of work with field experiments 
(cf. Glenn W. Harrison and John A. List 2004). It may be useful to pursue 
experiments analogous to those of this book with larger societies in which 
anonymity may be more readily obtained.

22 The introduction briefly mentions a dinner in which the anthropologists on 
this project entertained the economists with stories of what the former had 
eaten in the field. The book is less graphic, but no less interesting in 
describing how work in economics and anthropology can be combined.

Punishment and Social Norms." Evolution and Human Behavior, 25(2): 63-87.

Fehr, Ernst, and Simon Gächter. 2000. "Cooperation and Punishment in Public 
Goods Experiments." American Economic Review, 90(4): 980-94. Fehr, Ernst, and 
Simon Gächter. 2000. "Fairness and Retaliation: The Economics of Reciprocity." 
Journal of Economic Perspectives, 14(3): 159-81.

Fehr, Ernst, and Simon Gächter. 2002. "Altruistic Punishment in Humans." 
Nature, 415(6868): 137-40.

Fehr, Ernst, and Joseph Henrich. 2003. "Is Strong Reciprocity a Maladaptation? 
On the Evolutionary Foundations of Human Altruism," in The Genetic and Cultural 
Evolution of Cooperation. Peter Hammerstein, ed. Cambridge: MIT Press, 55-82.

Fehr, Ernst, and Klaus M. Schmidt. 1999. "A Theory of Fairness, Competition, 
and Cooperation." Quarterly Journal of Economics, 114(3): 817-68.

Güth, Werner, Rolf Schmittberger, and Bernd Schwarze. 1982. "An Experimental 
Analysis of

Ultimatum Bargaining." Journal of Economic Behavior and Organization, 3(4): 
367-88.

Harrison, Glenn W., and John A. List. 2004. "Field Experiments," Mimeo. 
University of Central Florida and University of Maryland.

Henrich, Joseph. 2004. "Cultural Group Selection, Coevolutionary Processes and 
Large-Scale Cooperation." Journal of Economic Behavior and Organization, 53(1): 
3-35.

Ledyard, John O. "Public Goods: A Survey of Experimental Research," in Handbook 
of Experimental Economics. John Kagel and Alvin E. Roth, eds. Princeton: 
Princeton University Press, 111-94.

Rabin, Matthew. 1993. "Incorporating Fairness into Game Theory and Economics." 
American Economic Review, 83(5): 1281-1302.

Roth, Alvin E. 1995. "Bargaining Experiments," in Handbook of Experimental 
Economics. John H. Kagel and Alvin E. Roth, eds. Princeton: Princeton 
University Press, 253-348.