. 5
( 13)


The important point here is that the potential ally may prefer not just
good indicator traits, but ones that will provide additional bene¬ts to
him or her. In many cases, group-bene¬cial signals will be more likely
to have this quality than other types of signals. Note that this explana-
tion, like the broadcast ef¬ciency one in the previous section, relies
solely on individual advantages (to signaler and to observer) of group-
bene¬cial signaling; it is based on mutualism rather than altruism and
thus provides an alternative to both reciprocal altruism and strong
126 Smith and Bliege Bird

4.5 Signaling and Collective Action

4.5.1 The Problem of Collective Goods
Generosity”such phenomena as sharing food outside the immediate
family, giving gifts, hosting public events, or helping neighbors in
need, all at some cost to one™s self”seems to be a ubiquitous cross-
cultural feature of human social life. But generosity is not universally
nor randomly extended. Instead, it appears to be strategic”the con-
texts in which such acts occur, as well as the characteristics of donors
and recipients, seem to be highly constrained and patterned. Some of
this variability may be adaptive and therefore explicable using theory
from evolutionary ecology. Such explanations have most frequently
been framed in terms of conditional reciprocity, involving such con-
cepts as reciprocal altruism, tit-for-tat, iterated Prisoner™s Dilemma,
and the like (Trivers 1971; Axelrod and Hamilton 1981; Cosmides and
Tooby 1989). Others have argued that sharing or other forms of puta-
tive generosity are due to coercion on the part of recipients and hence
are a form of ˜˜tolerated theft™™ (Blurton Jones 1984; Hawkes 1992). CST
provides a third explanation, involving mutualism rather than reci-
procity or coercion (Dugatkin 1997), although as we will argue, these
need not be mutually exclusive and indeed can work in concert.
One classic example of public generosity is the widespread practice
among hunter-gatherers of sharing individually harvested resources
with nonkin. This practice is commonly explained as a means of reduc-
ing the risk associated with acquiring productive but highly variable
resources such as big game (Smith 1988). Pooling individually har-
vested resources smoothes out consumption variance for all partici-
pants, a considerable bene¬t when harvests are so unpredictable that
individual hunters can expect many days or weeks to elapse between
successes (Winterhalder 1990). While risk-reduction effects are plausi-
ble and can be demonstrated to exist (Cashdan 1985; Kaplan, Hill, and
Hurtado 1990), these effects might be an outcome rather than a cause
of food-sharing. Sharing in order to reduce consumption variance in-
volves a Prisoner™s Dilemma payoff structure (Smith and Boyd 1990)
and thus creates incentives to slack off and free-ride on the efforts of
others (Blurton Jones 1986; Hawkes 1993). This insight has led to an
active debate about how to explain the undisputed fact of extensive
food-sharing by hunter-gatherers (Winterhalder 1996).
We believe that while conditional reciprocity may explain some cases
of hunter-gatherer food sharing, it cannot explain them all. When all
Costly Signaling and Cooperative Behavior 127

group members have rights to consume the resource regardless of their
past contribution, and the number partaking is dozens or more, the
conditions for conditional reciprocity are not met (Hawkes 1992). Such
lack of contingency (failure to direct shares so as to repay debts or
create indebtedness) occurs where there are strong social norms gov-
erning the distribution of shares and when resources are distributed
to individuals regardless of whether they ever repay the donor. This
situation applies most obviously in public ceremonial contexts such as
funerary rites (Smith and Bliege Bird 2000), big-man feasting (Wiess-
ner and Schiefenhovel 1995), Northwest Coast Indian potlatching
(Boone 1998), or charity galas in capitalist society (Veblen 1899). Here,
generosity takes place within a broad social arena, and resources are
distributed as a collective good simultaneously to large numbers of
recipients. The generous individual cannot ensure that the targets of
his generosity will return the favor. If sharing or generosity cannot be
made contingent upon reciprocation, then the fundamental condition
for evolutionarily stable reciprocity is absent.
Lack of attention to a recipient™s past history or future probability of
reciprocating has been described as part of food sharing patterns in
many hunter-gatherer societies, such as the Ache Indians of the Para-
guayan forest (Kaplan and Hill 1985a), and the Hadza of the East Afri-
can savanna (Hawkes 1993). In these cases (and others), at least some
types of harvested resources are shared unconditionally with most or
all members of the community, and some hunters consistently provide
more than others while sharing more or less equally in the catch. These
˜˜altruistic™™ providers in fact enjoy higher social status and reproduc-
tive success than their less productive peers, despite the absence of
any conditional exchange of ˜˜meat for mates™™ (Kaplan and Hill 1985b;
Marlowe 2000; Bliege Bird, Smith, and Bird 2001; Smith in press).
Enhanced reputations, social status, and its subsequent mating advan-
tages could be the ˜˜selective incentive™™ (Olson 1965) that motivates cer-
tain individuals to provide collective goods. But why should recipients
reward generous providers with high status? Is this not just another
form of reciprocity? If so, we would seem to have solved one collective
action problem by posing another (Smith 1993).

4.5.2 Is ˜˜Indirect Reciprocity™™ a Solution?
Early social theorists analyzing public generosity interpreted some of
its many forms as an outgrowth of social competition, a rivalry in
divesting oneself of goods where the most generous individual gains
128 Smith and Bliege Bird

the highest prestige, and the recipients of goods gain material bene¬t
at the expense of their reputations (Veblen 1899; Mauss 1926; Fried
1967). Recently, behavioral biologists have begun to modify models of
reciprocal altruism to account for reputation enhancement and status
bene¬ts associated with giving, what Alexander (1987, 85) termed ˜˜in-
direct reciprocity.™™ Indirect reciprocity, in Alexander™s scheme, results
when ˜˜the return explicitly comes from someone other than the recipi-
ent of the original bene¬cence.™™
Nowak and Sigmund (1998) proposed that the bene¬t gained from
advertising one™s cooperative tendencies through costly acts of altru-
ism is the increased chance of becoming the recipient of another™s al-
truistic act at a later date. They constructed computer simulations in
which one of a pair of players could choose whether or not to donate
help based on the potential recipient™s behavior in previous pairings
with others, as measured in their ˜˜image score.™™ While their simulation
showed that reputations did matter in choosing partners, others have
demonstrated that the particular form by which the ˜˜image score™™ is
instantiated in Nowak and Sigmund™s analysis is not evolutionarily
stable (Leimar and Hammerstein 2001). However, a related form of
indirect reciprocity, the ˜˜standing strategy™™ (Sugden 1986), can both
invade a noncooperative population and resist invasion by other strat-
egies, even when errors and incomplete information are allowed (Pan-
chanathan and Boyd 2003).
While this work illuminates theoretical possibilities, it will take em-
pirical research to determine if altruists are indeed preferred targets of
the altruism of third parties. For example, experiments conducted by
Wedekind and Milinski (2000) showed that those who were more gen-
erous in dyadic semi-anonymous interactions received more donations
in return, which the experimenters interpret as support for the ˜˜image
scoring™™ version of indirect reciprocity. However, the more generous
players preferred to give indiscriminately, rewarding the generous
and stingy alike. This is precisely the behavior one would predict if
generosity were a costly signal of the ability to donate resources. Milin-
ski, Semmann, and Krambeck (2002) report experimental results show-
ing that players who donated more to a charity (information which
was known to other players in the game) received both more aid from
fellow players in an indirect reciprocity game and more votes for elec-
tion to a student organization. As the authors note (p. 883), ˜˜Donating
to those who are in need might serve as an honest and ef¬cient (be-
cause it is done in public) signal for one™s reciprocity reliability.™™
Costly Signaling and Cooperative Behavior 129

Boyd and Richerson (1989) also modeled indirect reciprocity and
concluded that strategies based on the principle ˜˜be nice to people
who are nice to others™™ were relatively successful. However, they and
others have noted that the evolutionary stability of indirect reciprocity
is likely only in small groups, where there is some way of keeping
score of one™s giving and receiving and of targeting cooperation con-
ditionally at reciprocators”precisely the same limitations faced in di-
rect conditional reciprocity. Roberts (1998) proposed a solution to this
problem by arguing that reputation could be an indirect bene¬t of
altruistic behavior if interactions are modeled in two stages: (1) an
assessment stage, in which individuals establish reputations for gen-
erosity through public and non-reciprocal displays; and (2) a subse-
quent stage involving dyadic interactions, where individuals choose
a cooperating partner based on reputations previously established.
Like Milinski et al., Roberts is thus proposing essentially a costly sig-
naling argument, where the key function of establishing reputations
through costly public displays of altruism is to facilitate trust in dyadic
While ˜˜indirect reciprocity™™ models tend to focus on bene¬ts gained
through subsequent pair-wise cooperative interactions following dis-
play, CST helps us see that these may not be necessary in order for
both signalers and observers to bene¬t. As we have pointed out (Smith
and Bliege Bird 2000), observers might respond to signals by subse-
quently avoiding the signaler altogether, in much the same way as red
deer roaring provides a way for competitors to evaluate the probability
of winning a ¬ght without actually risking injury (Clutton-Brock and
Albon 1979). Indirect reciprocity explanations also overlook the possi-
bility that there may be bene¬ts gained simply as a function of the dis-
play, both by those who display and those who observe.

4.5.3 The Costly Signaling Solution: An Ethnographic Example
Meriam turtle hunting provides an interesting test of the ability of CST
to explain seemingly inef¬cient (and costly) foraging and food distri-
bution patterns (Smith and Bliege Bird 2000; Bliege Bird, Smith, and
Bird 2001). There are two primary types of marine turtle acquisition
on Mer (an island in Torres Strait, northern Australia)”hunting and
collecting. Turtle hunting occurs primarily in the context of public
feasting events; hunters choose to hunt in response to a request from
feast organizers to provide turtles for consumption at a previously
announced feast. Among Meriam turtle hunters, there are three distinct
130 Smith and Bliege Bird

roles: hunt leader, jumper, and driver. In addition, turtles are also col-
lected, primarily in the context of household provisioning, but also for
feasts, by men of all ages, women, and children. This occurs only when
they can be harvested on beaches during the nesting season (October
through April).
Hunting turtles is a competitive pursuit, with a very different com-
plement of participants than collecting. As the Meriam put it, ˜˜Anyone
can collect turtle in the nesting season, but only certain men have the
ability to succeed at turtle hunting.™™ Compared to collecting, hunting
is more costly (in terms of time, energy, and risk), provides meat less
ef¬ciently (due to higher travel, search, and pursuit costs), and is asso-
ciated with wider distributions of meat (¬gure 4.1). Hunters keep no
meat for themselves, except in the rare occasion when they are hunting
for household consumption, in which case they still keep less and
share more than turtle collectors. Hunters take on a variety of costs for
which they are not materially compensated, including time and energy

#HH consuming
After Sharing
Before Sharing

25000 50

20000 40
Mean Return Rate (kcal/hr)

15000 30
#HH Consuming

10000 20

5000 10

0 0

-5000 -10
Collect/House Hunt/House Collect/Feast
Hunt Type/Distribution Type

Figure 4.1
Meriam turtle hunting versus collecting returns before and after sharing, and number of
households consuming for each hunt/distribution type.
Costly Signaling and Cooperative Behavior 131

in hunting, money for fuel, and time organizing and preparing the
hunting team and its equipment prior to the hunt. The ability to bear
such costs appears to be linked to hunter quality: because a hunt leader
is an organizer and decision-maker, his abilities peak as he gains skill
and experience. The signals sent by hunting also are ef¬ciently broad-
cast: hunts are associated with larger numbers of consumers and thus
attract a broader audience than collections during the nesting season
and during household consumption events (¬gure 4.1). When quizzed,
most feast-goers (audience members) know the identity of hunters,
while the identity of jumpers seems to be common knowledge only
among their own peer group of young males.
The bene¬ts hunters receive from generously providing turtle for
public consumption do not appear to come in the form of increased
shares of collected turtle or other foods, as we might predict if risk re-
duction reciprocity were structuring the payoffs for hunting (Bliege
Bird et al. 2002). Those who acquire turtle (both hunted and collected)
more frequently and share more widely (¬gure 4.2) or acquire turtle in
greater quantity (¬gure 4.3) do not receive turtle more frequently as
compared to those who share less or not at all. In addition, generous
turtle sharing does not appear to be repaid through receiving shares of
¬sh or other foods (¬gure 4.4).
The CS explanation of collective goods provisioning as applied to
the Meriam turtle hunting case proposes that turtle hunters bene¬t
from unconditional sharing because their harvesting success sends
honest signals about their quality to the community in which they will

Turtle Shares Received

R2 = 0.0026
0 20 40 60 80 100 120 140
Total Shares Out

Figure 4.2
Do those who give turtle to more households receive turtle more frequently?
132 Smith and Bliege Bird

#Turtle Shares Received



-100 0 100 200 300 400 500 600
Sum Kg Shared
Y = 2.978 + .003 * X; R2 = .037

Regression Coefficients
#Turtle Shares Received vs. Sum Kg Shared

Coefficient Std. Error Std. Coeff. t-Value P-Value
Intercept 2.978 .541 2.978 5.506 <.0001
Sum Kg Sh .003 .002 .192 1.127 .2680

Figure 4.3
Do those who share more kilograms of turtle receive more frequently?

play out their lives as mates, allies, and competitors. Paying attention
to such signals can bene¬t observers because the costs and potential
for complete failure inherent in the signal guarantee that it is an honest
measure of the underlying qualities at issue. Only those endowed with
the necessary skills will succeed and be asked to serve regularly on
crews or as hunt leaders. The bene¬ts accruing to signalers (hunters)
will depend upon the speci¬c signal and audience. For hunt leaders,
they might consist of being deferred to by elders or obtaining bene¬ts
of a hard working wife™s labor (Smith, Bliege Bird, and Bird 2003). For
jumpers, they might include a means of establishing social dominance
among peers and hence preferential access to various social resources,
including enhanced mating opportunities.
Interestingly, the Meriam themselves interpret unconditional gener-
osity such as donating turtles to feasts or sharing collected turtles with
neighbors as different from conditional reciprocity (which also has a
Costly Signaling and Cooperative Behavior 133


Amount Received/day






-2000 0 2000 4000 6000 8000 10000 12000 14000
Amount Shared/day

Y = 2657.796 - .012 * X; R2 = 1.39 * 10“4

Regression Coefficients
Amt Rec/day vs. Amt Sh/day

Coefficient Std. Error Std. Coeff. t-Value P-Value
Intercept 2657.796 2139.059 2657.796 1.243 .2604
Amt Sh/day -.012 .399 -.012 -.029 .9779

Figure 4.4
Do those who share more food receive more food?

place in their social and economic lives). The Meriam conception of
generosity involves reference to a concept called ˜˜debe tonar,™™ which
means ˜˜the good way.™™ Debe tonar is a set of principles for everyday so-
cial interactions, and adhering to those principles is considered a signal
of Meriam identity. Following debe tonar is said to provide long-term
bene¬ts in the form of an enhanced social reputation as a ˜˜good per-
son.™™ Subsistence decisions intersect with that aspect of debe tonar that
incorporates generosity with food. There are strong social sanctions
against conditional or contingent sharing of ˜˜table food™™ (derapeili”to
share portions out) between households following the harvesting of
wild or cultivated foods. Similarly, there are social norms governing
the provisioning of food to feasts: such foods are explicitly designated
public goods (kies), open to unconditional consumption. Debe tonar
states that such unconditional sharing is its own reward; to share
with the expectation of a return in kind is to share sel¬shly, and to
134 Smith and Bliege Bird

return a portion given freely is to imply that the giver had sel¬sh
motives. A Meriam adage goes, ˜˜When you expect payment back,
you do not esak gem blo em™™ (make an lasting impression). Sharing
sel¬shly is equivalent to not sharing at all: both are referred to as
being ˜˜gobar,™™ greedy with food. While derapeili and kies sharing are
explicitly unconditional, there are conditional sharing contexts within
which reciprocal exchange is considered proper: markets (tama), labor
exchange/work party feasts (irapu), ceremonial exchange partners
(wauri tebud), and repayments (bodomelam) for use-rights to certain
individually or corporately owned items (land, foraging areas, boats,
or tools).

4.5.4 Why Does Crowding Out Occur?
As discussed in chapter 1 and elsewhere in this volume, it often hap-
pens that voluntary contributions to a public good will decrease when
such contribution is supplemented or replaced by direct material in-
centives (e.g., payments to blood donors). The costly signaling frame-
work can be used to generate the hypothesis that this ˜˜crowding out™™
may be due to a dilution of signaling value. If the social value of donat-
ing blood is cheapened by an increased supply from sellers or if blood
donation might be con¬‚ated with blood selling, the signaling value of
donating is likely to fall, perhaps suf¬ciently to lie below the donation
cost for many former donors.
We are not arguing that all examples of ˜˜crowding out™™ can be
explained by costly signaling processes, nor that we are certain that
some can be so explained. However, it appears to be a plausible hy-
pothesis, supported by a formal model and empirically testable under
the right circumstances. The alternative explanations in terms of strong
reciprocity and histories of group selection proposed elsewhere in this
volume are equally testable, but it appears in most cases that the data
to distinguish between the explanatory ef¬cacy of the two explanations
are lacking.

4.5.5 Signaling and Group-Bene¬cial Enforcement
Another type of collective good that may be a form of costly signaling
involves punishing those who free-ride on the group™s cooperative
activities or otherwise violate group-bene¬cial norms. It is well known
that although enforcing cooperation by punishing defectors will solve
collective action problems, such enforcement is costly to those who
carry out the enforcement and is itself a collective good, thus posing
Costly Signaling and Cooperative Behavior 135

a second-order collective action problem (Hardin 1982)”albeit a less
demanding one at the cooperative (low“enforcement-frequency) equi-
librium (Boyd et al. 2003). Boyd and Richerson (1992) demonstrated
that if enforcement takes the form of punishing both noncooperators
and nonpunishers, then cooperation (or anything else) can be evolutio-
narily stable, even in large groups and even if enforcement is only car-
ried out by a small fraction of the group™s members. Such enforcement
may be a potent element in stabilizing cooperation in many types of
social systems (Clutton-Brock and Parker 1995; Frank 1995; Richerson
and Boyd 1998; chapter 7 in this volume).
Note, however, that enforcement can serve as a costly signal in its
own right, as long as the costs of enforcement are quality-dependent.
If the qualities that make one a good (low-cost) enforcer also make one
attractive as a potential ally or someone it will pay to defer to in other
contexts, enforcement costs can be recouped by signaling bene¬ts. This
costly signaling dynamic can then provide a private bene¬t to the
enforcer and thus in principle solve the second-order collective ac-
tion problem. The GSB model provides one mechanism for the evolu-
tion of such a system. In this version, enforcement”punishment of
noncooperators”is itself the behavior that signals high quality. This
model readily allows such punishment or enforcement to serve as the
costly signal and hence to be maintained when the conditions for evo-
lutionary stability speci¬ed in the model are met.
Here is a brief account to show how the model captures this form of
signaling (adapted from Gintis, Smith, and Bowles 2001). Suppose that
a group of n members can cooperate to provide some collective good.
By cooperating, each member contributes a total bene¬t of g to others
at a ¬tness cost of d to himself. Thus, the gain from defecting is d, and
to prevent this, members must be punished at least d for defecting.
Now suppose that a high-quality individual can impose d on defectors
at a personal cost of c, whereas a low-quality individual must incur
cost c 0 > c to achieve the same effect. Following the model summarized
in section 4.3, under the range of parameter values noted, there will be
an equilibrium in which high-quality individuals will punish and low-
quality ones will not. In turn, observers will bene¬t by using such pun-
ishment behavior as a signal of underlying qualities that will provide
useful information for future social interactions. And of course, all
group members will bene¬t from the effect of punishment in enforcing
cooperation in collective action. To our knowledge, this argument has
not been directly applied to any empirical cases of group-bene¬cial
136 Smith and Bliege Bird

punishment. However, it does seem to be consistent with a variety
of ethnographic observations (e.g., Boehm 1999, chapter 5 of this
A common observation that can be experimentally replicated (see
chapter 5 of this volume), is that in many circumstances people enforce
a norm of fairness or equity. Of course, the speci¬c meaning of ˜˜fair-
ness™™ is culturally variable and often highly contested within any given
culture or society (compare the arguments of U.S. Democratic versus
Republican lawmakers on tax policy, for example). Nevertheless, even
if subject to con¬‚icting interpretations and strategies of deception and
manipulation, enforcement of fairness or redistributive equity in divi-
sion of the social product is a pervasive feature of human social life
(see various chapters of the present volume for evidence and discus-
sion). While models of strong reciprocity (chapter 6) and conditional
reciprocity (chapter 3) offer plausible hypotheses to explain this phe-
nomenon, in the spirit of theoretical pluralism we want to sketch how
CST might have something to contribute to this topic.
In brief, we suggest that fairness norms allow assessment of the abil-
ity (or willingness) of individuals or coalitions to ˜˜pay their share™™”
that is, to pay the cost that ensures signal reliability. This could apply
equally (but with different forms and signaling details) to both egali-
tarian systems”where fairness means contributing equally”and to
hierarchical systems”where some are allowed to possess more wealth
and power than others but are expected to contribute to the common
good (e.g., hosting ceremonial events) according to their greater abili-
ties. In either case, norms that require members of a collective to do-
nate their surplus to have-nots (Boone 1998) or to contribute equally to
production of a public good could be motivated by signaling concerns.
Given this multiplicity of social contexts and underlying qualities
being signaled, failure to pay one™s share could have one of several dis-
tinct meanings and consequences, including:
1) inability to do so (a signal of low quality, leading to reduced social
2) defection from the game (a decision to withdraw from a given arena
of status competition), and
3) ¬‚aunting the norm (a signal of social power or superior status, reli-
able to the extent that norm violation is more costly than simply ˜˜pay-
ing one™s share,™™ and hence not viable for low-status or subordinate
Costly Signaling and Cooperative Behavior 137

Thus, we suggest that the meaning of fairness-norm violations must be
interpreted in light of contextual information. In any case, the CS
framework suggests that attention must be paid to the information
value of fairness norms, and that enforcement of these norms may
be aimed at ensuring signal reliability and solving status-competition
games as much as (or more than) ensuring equity per se.

4.5.6 Signaling and Inter-Group Con¬‚ict
A related set of phenomena involve individually costly contributions
to violent con¬‚ict between social groups. Participating in group raid-
ing or defense is common among chimpanzees as well as human soci-
eties (Boehm 1992; Manson and Wrangham 1991), yet this kind of
activity poses some thorny challenges to evolutionary analysis. Such
behavior provides bene¬ts that are available to all group members and
cannot be hoarded or individually consumed, and therefore approach
the classic de¬nition of a pure public good. Yet the costs to contribu-
tors can be extraordinarily high (including of course death), which
means that the payoffs to free-riding should be considerable. Some
have argued that the widespread occurrence of organized intergroup
violence thus can only be explained as the result of a history of genetic
group selection (Alexander 1979, 1987; Hamilton 1975; Eibl-Eibesfeldt
1982). Others suggest that group selection acting on cultural variation
could be responsible for favoring self-sacri¬cial aspects of intergroup
con¬‚ict (Peoples 1982; Richerson and Boyd 1998; chapter 7 of this
volume). CST offers a possible alternative to both of these views, with
participation in group military defense and offense serving as a display
of underlying qualities useful for status competition within one™s
The evidence that military contribution and self-sacri¬cial bravery is
a primary avenue to male status enhancement in small-scale societies is
substantial (Chagnon 1990; Otterbein 1970; Patton 2000). It is easy to
see that status enhancement may ensure the spread of even dangerous
status-enhancing behavior if it has suf¬cient bene¬ts to material, repro-
ductive, or cultural success. The key question, as we saw with the issue
of unconditional generosity, is why others grant such status to war-
riors. The CS explanation is that success in warfare signals underlying
qualities that are valued by prospective allies and deferred to by pro-
spective competitors. In turn, females may mate preferentially with
successful warriors (Chagnon 1988) because they bene¬t from the
138 Smith and Bliege Bird

social dominance of such individuals, even if the particular qualities
signaled are not of direct bene¬t to a spouse.
The group-level bene¬ts of status-seeking by warriors is thus inci-
dental to the CS explanation, and indeed the CS dynamic could pro-
duce an oversupply of military adventuring as well as an undersupply
of dedicated warriors. We read the ethnographic and historical record
as providing ample evidence of both (e.g., Boone 1983; Keeley 1996;
Mesquida and Wiener 1996; Otterbein 1970). Of course, CS dynamics
could work in concert with multilevel selection of either genetic or cul-
tural variation. For example, depending on initial conditions, CS might
yield a variety of equilibria in the intensity of intergroup con¬‚ict, and
those equilibria that happen to optimize the supply of belligerence (as
measured by enhanced persistence and/or spread of the social group)
would then be favored. This would be an example of the process of
group selection among alternative local equilibria modeled by Boyd
and Richerson (1990) (see also chapter 7 in this volume).

4.6 Signaling and Commitment

4.6.1 Signaling and Common Goals
The presence of con¬‚icting interests among social organisms often sets
high barriers to cooperation. Gender can be a source of profound dis-
parities in reproductive interests such that ˜˜even when cooperating in
a joint task, male and female interests are rarely identical™™ (Trivers
1972, 174). Even the most closely related groups of cooperating indi-
viduals have imperfectly coincident genetic interests and can therefore
exhibit intense con¬‚ict and competition. Although siblings share a con-
siderable degree of common genetic interest, they also often compete
more with each other (for parental investment and other resources)
than with other individuals in the social group (Sulloway 1996). Con-
¬‚icting genetic interests between mother and fetus institute a form of
maternal-fetal warfare during gestation, even though by cooperating
each could maintain an outcome better for both (Haig 1993).
Individuals often come into con¬‚ict about working toward a shared
goal when each faces different tradeoffs and gains different bene¬ts
from working toward that goal. For example, given the fact that males
and females have different reproductive strategies and life histories,
they will often face con¬‚icts between working toward a common goal
of household provisioning (in effect, an investment in self-maintenance
and the rearing of children) and alternative productive and reproduc-
Costly Signaling and Cooperative Behavior 139

tive goals, such as gaining status in the wider community, producing
children with other mates, acquiring more mates, or accumulating
wealth. To have an incentive to cooperate, however, partners must
have goals in common to some degree, such that both bene¬t more
from the partnership than each would if they acted independently.
In choosing a partner for a dyadic cooperative endeavor”such
as a mate to cooperatively raise offspring, a research colleague, or a
coauthor”each partner must be convinced that he or she will gain
a net bene¬t from the interaction. This is particularly important if the
association will be a long-term one, with opportunities for cheating or
periods of one-sided costs for one partner (e.g., carrying a child to
term or writing the ¬rst draft of a manuscript) that would allow the
other partner to maximize short-term returns by defecting. How can
the potential partners discern each other™s intentions with any degree
of reliability?
One way they might do so is by sending honest signals of commit-
ment to common goals for the project or relationship in question. Pur-
suing our coauthorship example, con¬‚ict can occur when one partner
seeks individual status over the collective status achieved through
coauthorship or seeks to free-ride on the efforts of a harder-working
coauthor. In order to honestly signal his commitment to common
goals, he must show that when given the opportunity to gain individ-
ual status at the expense of his partner, he chooses to forgo this in
favor of promoting the interests of the partnership. For example, one
partner might be approached by a publisher interested in securing the
rights to a book deal based on jointly conducted research. If this part-
ner accepts the deal and promotes himself as sole author, he gains indi-
vidual bene¬ts but jeopardizes the continuation of the partnership. If
he declines the publisher™s offer (or brings the partner into the deal),
he pays an opportunity cost in terms of his own individual status but
signals his ongoing commitment to the partnership. Since the tempted
partner can only expect the short-term cost of forgoing sole authorship
to be repaid if he expects the relationship to last long enough to pro-
duce bene¬ts greater than these costs, his actions are an honest signal
of commitment. Thus, the cost paid by not defecting (if suf¬ciently
high) guarantees the honesty of the signal.

4.6.2 Coalition Commitment
Signaling of common goals applies not only to dyadic partnerships, but
might also help solve multi-agent cooperative dilemmas. Relationships
140 Smith and Bliege Bird

among members of relatively stable social groups often involve such
interactions; for example, some of the most important interactions are
cooperative behaviors involving group territorial defense. When chal-
lenges by neighboring groups come at unpredictable times, periodic
testing of the commitment of group members to engage in costly terri-
torial defense helps ensure that they can be relied upon when the time
comes to ¬ght. Zahavi and Zahavi (1997) suggest that among Arabian
Babblers (a group-living bird), many social behaviors such as hud-
dling, grooming, and group dances serve to test the social bond and,
by extension, demonstrate commitment to group defense. Dances and
ceremonies in many human groups that are performed prior to head-
ing into battle might serve much the same function. However, plausi-
bility arguments such as this one need to be stated in a more directly
falsi¬able manner and then subjected to careful empirical tests.
Cooperative dilemmas also arise in the context of within-group com-
petition, involving political coalitions and alliances. Most models of
political power assume that politicians gain power as part of a recipro-
cal exchange: a politician promises ˜˜pork™™ to his constituents in return
for the favor of their vote. Given the delayed return here (˜˜If I support
you now, you will return the favor by providing collective goods in the
future™™), defection is always a distinct possibility. Costly signaling may
not eliminate the risk of defection, but it could help in predicting which
individuals are less likely to do so. If a politician can reliably signal a
superior ability to obtain resources for redistribution during period 1,
he should be more likely to actually do so in period 2.
Here, costly signaling does not guarantee honesty of intent to deliver
collective goods, but it may guarantee honest advertisement of the
ability to do so. A variety of political systems, ranging from the semi-
egalitarian ˜˜big-man™™ systems of Melanesia to the strati¬ed chiefdoms
of the Northwest Coast Indians, appear to display various elements of
this costly signaling dynamic of garnering and maintaining political
support through magnanimity (Boone 1998). In these cases, and argu-
ably in many instances of electoral politics in modern industrialized
democracies, political candidates use distributions of goods to honestly
signal their ability to bene¬t supporters in the future. The big man,
chief, or congressional candidate encourages others to donate wealth
or labor in his support by displaying honest signals of his skill in accu-
mulating resources, thus ameliorating the most problematic aspect of
delayed reciprocity: the risk of default.
Costly Signaling and Cooperative Behavior 141

4.7 Conclusions

Costly signaling theory provides the basis for arguing that gen-
erosity”incurring the costs of providing collective goods (including
those shared with partners in dyadic relationships)”is one means
by which individuals and coalitions compete for status, and ultimately
for the material and ¬tness-enhancing correlates of status (such as po-
litical power, mates, and economic resources). The quality-dependent
cost of providing the collective good guarantees the honesty of the
signaler™s claim to such qualities as resource control, leadership abili-
ties, kin-group solidarity, economic productivity, or good health and
vigor”information that is useful to the signaler™s potential mates,
allies, and competitors (Boone 1998; Smith and Bliege Bird 2000). If
this explanation is correct, it means that those who engage in acts of
unconditional generosity by providing collective goods are not acting
in hope of reciprocation in kind, nor sacri¬cing for the good of the
group or their partner, but rather are competing for status and its per-
quisites. We extended these arguments to address the issue of commit-
ment in cases of both dyadic reciprocity and n-person coalitions and
argued that signaling might provide an alternative to (or at least
strengthen) more conventional analyses of these phenomena.
The n-player game-theoretical model summarized earlier in this
chapter (and developed more fully in Gintis, Smith, and Bowles 2001)
speci¬es conditions under which an honest-signaling equilibrium will
be stable. At this equilibrium, only high-quality individuals signal
while observers respond only to signalers (these being the player™s re-
spective best moves). This model shows that group-bene¬cial signals
(such as unconditionally providing a collective good) can meet the
conditions for a signaling equilibrium. However, these results apply
equally well to socially neutral or even harmful signals, and hence this
model alone cannot tell us why group-bene¬cial signals would be fa-
vored over other signals.
To address this last question, we brie¬‚y discussed three distinct (but
not mutually exclusive) hypotheses. One of these involves equilibrium
selection among alternative (some more group-bene¬cial, some less)
signaling equilibria, possibly through a process of cultural group selec-
tion. A second proposes that the value to the signaler of providing a
collective good over some more ˜˜wasteful™™ display may lie in the
broadcast ef¬ciency of the signal in competing for observer attention,
142 Smith and Bliege Bird

given that observers are more likely to be attracted to signals that pro-
vide an additional consumption bene¬t. Our third hypothesis is that
when members of a social group bene¬t directly from cooperative sig-
nals, these signals can be favored because they serve as reliable indica-
tors that the signaler™s allies have an increased probability of receiving
similar private bene¬ts from the signaler in the future.
While the costly-signaling approach to cooperation and collective
action opens up some exciting vistas, it raises many questions that will
require extensive theoretical and empirical work. Of these, we will
mention three major ones: ¬guring out (1) what underlying qualities
are being signaled, (2) who the intended signal recipients are, and (3)
under what conditions signaling to the ¬eld is a better strategy than
targeting speci¬c signal recipients. The ¬rst two questions are primar-
ily empirical, although potentially very dif¬cult ones to answer in par-
ticular cases. The third question is one that is side-stepped in the GSB
model”which simply presumes that attracting a large audience by
providing a collective good will increase the payoff to the signaler.
This is likely to be the case under at least two circumstances: 1) when
each signaler can bene¬t from attracting multiple partners of a single
type (e.g., multiple potential mates), and 2) when each signaler can
bene¬t from in¬‚uencing multiple types of observers (e.g., attracting
allies and mates, intimidating competitors).
Another issue that presents intriguing possibilities for theoretical
development is to extend the one-shot analysis of the GSB model to
situations of repeated signaling and extended interaction. Repeated
signaling is likely to occur when reputations need to be built or main-
tained. We suspect that this is likely to be important under two main
(but not mutually exclusive) conditions: (1) where there is lots of
˜˜noise™™ (i.e., variation in signaling that is not due to underlying qual-
ities), and (2) where the qualities being signaled are likely to change
quickly over time (e.g., due to ecological and economic variations).
The ¬rst situation is exempli¬ed by the hunting of large game. In this
scenario, elements not controlled by the hunter can have a large
but unpredictable effect on success, yet underlying qualities (such as
ethological knowledge, visual acuity, stamina, and so forth) can lead
to consistent differences between individuals in long-term average
returns. The second situation”where underlying qualities are subject
to rapid decay”is characteristic of a variety of economic situations,
ranging from subsistence regimes subject to severe ecological ¬‚uctua-
tion (e.g., pastoralists in arid lands) to a speculative capitalist economy,
Costly Signaling and Cooperative Behavior 143

where entrepreneurs and members of the underclass are subject to
boom-and-bust conditions. We can expect the ¬rst situation to feature
repeated signaling early in an individual™s career, in order to establish
a reputation that will then be relatively stable and require little or no
future reinforcement (e.g., the turtle-hunting careers of Meriam that
typically take a decade or so to establish and may then persist for only
a few years more, yet result in long-term social and reproductive
gains). The second situation is likely to be much more dynamic, with
signalers™ fortunes (and signaling intensities) rising and falling rapidly
with changing circumstances; this scenario will presumably present a
more challenging arena for formal modeling.

4.8 Acknowledgments

For many fruitful discussions that have strongly in¬‚uenced the ideas
developed in this paper, we thank Doug Bird, Carl Bergstrom, Sam
Bowles, Herb Gintis, Kristen Hawkes, Hillard Kaplan, and Joan Silk.
The n-player signaling model outlined in this chapter was developed
in collaboration with Gintis and Bowles and a more detailed version
is published elsewhere (Gintis, Smith, and Bowles 2001). Fieldwork
among the Meriam was carried out in close collaboration with Doug
Bird, Craig Hadley, and many Meriam friends, especially members of
the Passi family. This ¬eldwork was supported by grants from the
National Science Foundation, the Leakey Foundation, the Wenner-
Gren Foundation, and the Australian Institute of Aboriginal and Torres
Strait Islander Studies.


1. By ˜˜cooperative behavior™™ we mean any actions by Ego that enhance the well-being of
members of Ego™s social group beyond his or her immediate kin. Such actions need not
impose a net cost on Ego in the short term, and the social bene¬ts need not be evenly dis-
tributed among Ego™s social group, though these special cases are of particular interest in
this chapter and elsewhere in the present volume.
2. Lotem, Fishman and Stone (2002) subsequently published a model of the evolution of
cooperation via costly signaling that shares some aspects of the GSB model. However,
Lotem et al. model only dyadic interactions, and start with a population containing high
frequencies of conditional reciprocators, whereas Gintis et al. (2001) develop a multi-
player game with no initial (pre-signaling) levels of cooperation.
3. GSB show that the results are unchanged if (following Johnstone 1997 and Getty 1998)
signaling bene¬ts rather than signaling costs are made quality-dependent. They did
not, however, analyze the effect of allowing Responders to vary in ways that affect the
144 Smith and Bliege Bird

interaction payoff to Signalers. This two-sided matching problem is dif¬cult to model,
and has not been attempted in the costly signaling framework, though it has been devel-
oped for marriage markets and mate choice (e.g., Bergstrom and Real 2000).


Alexander, Richard D. (1979) Darwinism and Human Affairs. Seattle: University of Wash-
ington Press.

Alexander, Richard D. (1987) The Biology of Moral Systems. Hawthorne, NY: Aldine de
Arak, A., and M. Enquist. (1995) Con¬‚ict, receiver bias, and the evolution of signal form.
Phil Trans R Soc Lond B 1355:337“344.
Axelrod, Robert, and William D. Hamilton. (1981) The evolution of cooperation. Science
Bergstrom, Carl, and Leslie A. Real. (2000) Towards a theory of mutual mate choice: Les-
sons from two-sided matching. Evolutionary Ecology Research 2:493“508.
Bliege Bird, Rebecca L., Eric A. Smith, and Douglas W. Bird. (2001) The hunting handi-
cap: Costly signaling in human foraging strategies. Behavioral Ecology and Sociobiology
Bliege Bird, Rebecca L., Douglas W. Bird, Eric A. Smith, and Geoffrey Kushnick. (2002)
Risk and reciprocity in Meriam food sharing. Evolution and Human Behavior 23(4):297“
Blurton Jones, Nicholas G. (1984) A sel¬sh origin for human food sharing: Tolerated
theft. Ethology and Sociobiology 5:1“3.
Blurton Jones, Nicholas G. (1986) Fitness returns from resources and the outcome of
contests: Some implications for primatology and anthropology. In Primate Ontogeny, Cog-
nition and Social Behaviour, vol. 3. Ed. J. G. Else and P. Lee. Proceedings of the 10th Con-
gress of the International Primatology Society. Cambridge: Cambridge University Press,
Boehm, Christopher. (1992) Segmentary warfare and management of con¬‚ict: A compar-
ison of East African chimpanzees and patrilineal-patrilocal humans. In Coalitions and Alli-
ances in Humans and Other Animals. Ed. A. Harcourt and F. B. M. de Waal. Oxford:
Oxford University Press, 137“173.
Boehm, Christopher. (1999) Hierarchy in the Forest: The Evolution of Egalitarian Behavior.
Cambridge, MA: Harvard University Press.

Boone, James L. (1983) Noble family structure and expansionist warfare in the late Mid-
dle Ages: A socioecological approach. In Rethinking Human Adaptation. Ed. R. Dyson-
Hudson and M. A. Little. Boulder, CO: Westview Press, 79“96.
Boone, James L. (1998) The evolution of magnanimity: When is it better to give than to
receive? Human Nature 9:1“21.
Boyd, Robert, and Peter J. Richerson. (1988) The evolution of reciprocity in sizable
groups. Journal of Theoretical Biology 132:337“356.
Costly Signaling and Cooperative Behavior 145

Boyd, Robert, and Peter J. Richerson. (1989) The evolution of indirect reciprocity. Social
Networks 11:213“236.
Boyd, Robert, and Peter J. Richerson. (1990) Group selection among alternative evolution-
arily stable strategies. J. of Theoretical Biology 145:331“342.
Boyd, Robert, and Peter J. Richerson. (1992) Punishment allows the evolution of coopera-
tion (or anything else) in sizable groups. Ethology and Sociobiology 13:171“195.
Boyd, Robert, Herbert Gintis, Samuel L. Bowles, and Peter J. Richerson. (2003) The
evolution of altruistic punishment. Proceedings of the National Academy of Sciences, USA
Caro, Tim M. (1994) Ungulate antipredator behaviour: Preliminary and comparative evi-
dence from African bovids. Behaviour 128:189“228.
Cashdan, Elizabeth A. (1985) Coping with risk: Reciprocity among the Basarwa of north-
ern Botswana. Man 20:454“474.

Chagnon, Napoleon A. (1988) Life histories, blood revenge, and warfare in a tribal popu-
lation. Science 239:985“992.

Chagnon, Napoleon A. (1990) Reproductive and somatic con¬‚icts of interest in the gene-
sis of violence and warfare among tribesmen. In The Anthropology of War. Ed. J. Hass.
Cambridge: Cambridge University Press, 77“104.
Clutton-Brock, T. H., and S. D. Albon. (1979) The roaring of red deer and the evolution of
honest advertisement. Behaviour 69:145“170.
Clutton-Brock, T. H., and G. A. Parker. (1995) Punishment in animal societies. Nature
Cosmides, Leda, and John Tooby. (1989) Evolutionary psychology and the generation of
culture, part II: Case study: A computational theory of social exchange. Ethology and Soci-
obiology 10:51“97.

Dugatkin, Lee Alan. (1997) Cooperation among Animals: An Evolutionary Perspective. New
York: Oxford University Press.

Eibl-Eibesfeldt, Iraneus. (1982) Warfare, man™s indoctrinability, and group selection.
Zeitschrift fur Tierpsychologie 67:177“198.
Frank, Steven A. (1995) Mutual policing and repression of competition in the evolution of
cooperative groups. Nature 377:520“522.
Fried, Morton H. (1967) The Evolution of Political Society. New York: Random House.
Getty, Thomas. (1998) Handicap signalling: When fecundity and viability do not add up.
Animal Behaviour 56:127“130.
Gintis, Herbert, Eric A. Smith, and Samuel Bowles. (2001) Cooperation and costly signal-
ing. Journal of Theoretical Biology 213:103“119.
Grafen, Alan. (1990) Biological signals as handicaps. J. of Theoretical Biology 144:517“
Guilford, Tim, and Marian Stamp Dawkins. (1993) Receiver psychology and the design
of animal signals. Trends in the Neurosciences 16:430“436.
146 Smith and Bliege Bird

Haig, David. (1993) Genetic con¬‚icts in human pregnancy. Quarterly Review of Biology
Hamilton, William D. (1975) Innate social aptitudes of man: An approach from evolu-
tionary genetics. In Biosocial Anthropology. Ed. R. Fox. London: Malaby, 133“155.
Hardin, Russell. (1982) Collective Action. Baltimore: John Hopkins University Press.

Hawkes, Kristen. (1992) Sharing and collective action. In Evolutionary Ecology and Human
Behavior. Ed. E. A. Smith and B. Winterhalder. Hawthorne, New York: Aldine de
Gruyter, 269“300.
Hawkes, Kristen. (1993) Why hunter-gatherers work. Current Anthropology 34(4):341“362.
Johnstone, Rufus A. (1995) Sexual selection, honest advertisement and the handicap prin-
ciple: Reviewing the evidence. Biological Reviews 70:1“65.
Johnstone, Rufus A. (1997) The evolution of animal signals. In Behavioural Ecology: An
Evolutionary Approach. Ed. John R. Krebs and Nicholas B. Davies. Oxford: Blackwell,
Kaplan, Hillard, and Kim Hill. (1985a) Food sharing among Ache foragers: Tests of ex-
planatory hypotheses. Current Anthropology 26(2):223“246.
Kaplan, Hillard, and Kim Hill. (1985b) Hunting ability and reproductive success among
male Ache foragers: preliminary results. Current Anthropology 26(1):131“133.

Kaplan, Hillard, Kim Hill, and A. Magdalena Hurtado. (1990) Fitness, foraging and food
sharing among the Ache. In Risk and Uncertainty in Tribal and Peasant Economies. Ed. E.
Cashdan. Boulder, CO: Westview Press.
Keeley, Lawrence H. (1996) War before Civilization: The Myth of the Peaceful Savage. New
York: Oxford University Press.
Leimar, Olof, and Peter Hammerstein. (2001) Evolution of cooperation through indirect
reciprocity. Proceedings of the Royal Society of London, Series B 268(1468):745“753.
Lotem, Arnon, Michael A. Fishman, and Lewi Stone. (2002) From reciprocity to uncondi-
tional altruism through signalling bene¬ts. Proceedings of the Royal Society of London B
Manson, Joseph H., and Richard W. Wrangham. (1991) Intergroup aggression in chim-
panzees and humans. Current Anthropology 32:360“390.

Marlowe, Frank. (2000) The patriarch hypothesis: An alternative explanation of meno-
pause. Human Nature 11:27“42.
Mauss, Marcel. (1967[1926]) The Gift: Forms and Functions of Exchange in Archaic Societies.
New York: Norton.
Mesquida, Christian G., and Neil I. Weiner. (1996) Human collective aggression: A be-
havioral ecology perspective. Ethology and Sociobiology 17:247“263.
Milinski, Manfred, Dirk Semmann, and Hans-Jurgen Krambeck. (2002) Donors to charity
gain in both indirect reciprocity and political reputation. Proceedings of the Royal Society of
London, Series B 269:881“883.
Neiman, Fraser D. (1997) Conspicuous consumption as wasteful advertising: A Darwin-
ian perspective on spatial patterns in Classic Maya terminal monument dates. In Redis-
Costly Signaling and Cooperative Behavior 147

covering Darwin: Evolutionary Theory and Archeological Explanation. Ed. C. Michael Barton
and Geoffrey A. Clark. Washington, D.C.: Archeological papers of the American Anthro-
pological Association, No. 7, 267“290.
Nowak, Martin A., and Karl Sigmund. (1998) Evolution of indirect reciprocity by image
scoring. Nature 393:573“577.
Olson, Mancur. (1965) The Logic of Collective Action: Public Goods and the Theory of Groups.
Cambridge: Harvard University Press.
Otterbein, Keith. (1970) The Evolution of War: A Cross-Cultural Study. New Haven: HRAF.
Panchanathan, Karthik, and Robert Boyd. (2003) A tale of two defectors: The importance
of standing for evolution of indirect reciprocity. Journal of Theoretical Biology 224:115“126.
Patton, John Q. (2000) Reciprocal altruism and warfare: A case from the Ecuadorian Am-
azon. In Adaptation and Human Behavior: An Anthropological Perspective. Ed. Lee Cronk,
Napoleon Chagnon, and William Irons. Hawthorne, New York: Aldine de Gruyter, 417“
Peoples, James G. (1982) Individual or group advantage? A reinterpretation of the
Maring ritual cycle. Current Anthropology 23:291“310.
Petrie, M. (1994) Improved growth and survival of offspring of peacocks with more elab-
orate trains. Nature 371:598“599.
Richerson, Peter J., and Robert Boyd. (1998) The evolution of human ultra-sociality. In
Ideology, Warfare, and Indoctrinability. Ed. I. Eibl-Eibesfeldt and F. Salter. Oxford: Ber-
ghahn, 71“95.
Roberts, Gilbert. (1998) Competitive altruism: From reciprocity to the handicap principle.
Proceedings of the Royal Society of London B 265:427“431.
Smith, Eric Alden. (1988) Risk and uncertainty in the ˜˜original af¬‚uent society™™: Evolu-
tionary ecology of resource sharing and land tenure. In Hunters and Gatherers: History,
Evolution, and Social Change. Ed. T. Ingold, D. Riches, and J. Woodburn. Oxford: Berg,
Smith, Eric Alden. (1993) Comment on Hawkes. Current Anthropology 34:356.
Smith, Eric Alden. In press. Why do good hunters have higher reproductive success?
Human Nature.
Smith, Eric Alden, and Robert Boyd. (1990) Risk and reciprocity: Hunter-gatherer socio-
ecology and the problem of collective action. In Risk and Uncertainty in Tribal and Peasant
Economies. Ed. E. Cashdan. Boulder, CO: Westview Press, 167“191.
Smith, Eric Alden, and Rebecca L. Bliege Bird. (2000) Turtle hunting and tombstone
opening: Public generosity as costly signaling. Evolution and Human Behavior 21(4):245“
Smith, Eric Alden, Rebecca L. Bliege Bird, and Douglas W. Bird. (2003) The bene¬ts of
costly signaling: Meriam turtle hunters. Behavioral Ecology 14(1):116“126.
Spence, Michael. (1973) Job market signaling. Quarterly J. of Economics 87:355“374.
Sugden, R. (1986) The Economics of Rights, Co-operation and Welfare. Oxford: Basil Black-
148 Smith and Bliege Bird

Sulloway, F. (1996) Born to Rebel: Birth Order, Family Dynamics, and Creative Lives.
New York: Vintage.
Trivers, Robert L. (1971) The evolution of reciprocal altruism. Quarterly Review of Biology
Trivers, Robert L. (1972) Parental investment and sexual selection. In Sexual Selection and
the Descent of Man, 1871“1971. Ed. B. G. Campbell. Chicago: Aldine, 136“179.
Veblen, Thorstein. (1994 [1899]) The Theory of the Leisure Class. New York: Dover.
Wedekind, Claus, and Manfred Milinski. (2000) Cooperation through image scoring in
humans. Science 288:850“852.
Wells, K. D. (1988) The effect of social interactions on anuran vocal behavior. In The Evo-
lution of the Amphibian Auditory System. Ed. B. Fritzch. New York: Wiley, 433“454.
¨vel, eds. (1995) Food and the Status Quest: An Inter-
Wiessner, Polly, and Wulf Schiefenho
disciplinary Perspective. Oxford: Berghahn Books.
Winterhalder, Bruce. (1990) Open ¬eld, common pot: Harvest variability and risk avoid-
ance in agricultural and foraging societies. In Risk and Uncertainty in Tribal and Peasant
Economies. Ed. E. Cashdan. Boulder, CO: Westview Press, 67“87.
Winterhalder, Bruce. (1996) Social foraging and the behavioral ecology of intragroup re-
source transfers. Evolutionary Anthropology 5(2):46“57.

Wright, Jonathan. (1999) Altruism as signal”Zahavi™s alternative to kin selection and
reciprocity. J. of Avian Biology 30:108“115.

Zahavi, Amotz. (1975) Mate selection”A selection for handicap. J. of Theoretical Biology
Zahavi, Amotz. (1977) Reliability in communication systems and the evolution of altru-
ism. In Evolutionary Ecology. Ed. B. Stonehouse and C. M. Perrins. London: Macmillan
Press, 253“259.
Zahavi, Amotz. (1995) Altruism as handicap”the limitations of kin selection and reci-
procity. J. of Avian Biology 26:1“3.
Zahavi, Amotz, and Avishag Zahavi. (1997) The Handicap Principle: A Missing Piece of
Darwin™s Puzzle. New York: Oxford University Press.
III Modeling and Testing
Strong Reciprocity
5 The Economics of Strong

Ernst Fehr and Urs

5.1 Introduction

The purpose of this chapter is to show that economists and other so-
cial scientists fail to understand core questions in economics and social
theory if they insist on the self-interest hypothesis and rule out hetero-
geneity in the realm of social preferences. Two developments support
our argument. First, during the previous decades, experimental psy-
chologists and economists have gathered overwhelming evidence that
systematically refutes the self-interest hypothesis and suggests that a
substantial fraction of the people demonstrate social motives in their
preferences”in particular, preferences for strong reciprocity. Second,
there is also strong evidence indicating that deviations from self-
interest have had a fundamental impact on core issues in economics
and social theory.
Social preferences are other-regarding preferences in the sense that
individuals who exhibit them behave as if they value the payoff of rel-
evant reference agents positively or negatively. Depending on the situ-
ation, the relevant reference agents may be a person™s colleagues,
relatives, trading partners, or neighbors. It is important to keep in
mind that a person may have different reference agents indifferent
domains. Strong reciprocity means that individuals behave as if their
positive or negative valuation of the reference agent™s payoff depends
on the actions of the reference agent. If the actions of the agent are per-
ceived as kind, a strong reciprocator values the payoff of the reference
agent positively. If the actions are perceived as hostile, the payoff of
the reference agent is valued negatively. As we will see, strong reci-
procity is a particularly important form of social preference.1
152 Fehr and Fischbacher

One core question is to understand the workings of competition and
the interplay of competition and cooperation in markets, organizations,
and politics. Other core questions pertain to understanding the condi-
tions for successful collective actions, the prevailing structure of con-
tracts and property rights, and, above all, the workings of economic
incentives, because the workings of incentives constitute the essence of
economics. We claim that a satisfactory understanding of these ques-
tions is impeded by the self-interest hypothesis. In particular, we pro-
vide evidence suggesting that preferences for strong reciprocity shape
the functioning of competition, govern the laws of cooperation and
collective action, and have a decisive impact on how economic incen-
tives are constituted and how they function. The evidence we present
also indicates that, by changing the incentives for sel¬sh types, strong
reciprocity affects the prevailing interaction patterns and the con-
straints on individual behavior, that is, the prevailing contracts and
The structure of this chapter is as follows. In ˜˜The Nature of Social
Preferences,™™ we shortly describe the most important types of social
preferences. We then illustrate the preference for strong reciprocity by
means of two simple one-shot experiments and discuss whether recip-
rocal behavior in these experiments can be interpreted as a cognitive
mistake”that is a kind of habit that is learned in the repeated interac-
tions outside the laboratory and inappropriately applied to one-shot
situations”or whether reciprocal behavior is better interpreted as ra-
tional behavior driven by a preference for strong reciprocity. ˜˜Compe-
tition™™ then shows that if one neglects strong reciprocity, one cannot
understand the important effects of competition on market prices.
˜˜Cooperation™™ deals with cooperation and shows that decisive deter-
minants of cooperation cannot be understood on the basis of the self-
interest hypothesis. ˜˜Economic Incentives and Property Rights™™ deals
with economic incentives, contracts, and property rights. We present
evidence indicating that neither the effects nor the determinants of eco-
nomic incentives can be adequately understood if one neglects strong
reciprocity and that the interaction between economic incentives and
strong reciprocity is likely to have important effects on the optimality
of different types of contracts and property rights. ˜˜Proximate Models
of Strong Reciprocity™™ discusses some problems in the modeling of
strong reciprocity and fairness preferences, and to what extent it is pos-
sible to mimic preferences for strong reciprocity by simpler and more
tractable models of inequity aversion.
The Economics of Strong Reciprocity 153

5.2 The Nature of Social Preferences

The last ¬fteen years have seen a large number of studies indicating
that”in addition to economic self-interest”social preferences shape
the decisions of a substantial fraction of people. A person exhibits so-
cial preferences if the person does not only care about the economic
resources allocated to her but also cares about the economic resources
allocated to relevant reference agents. In this chapter, we do not at-
tempt to summarize the empirical evidence on social preferences (for
surveys, see Fehr and Schmidt 2003, Sobel 2001, and chapter 6).
Instead, we are interested in the social and economic implications of
people™s social preferences. Before we proceed, it is nevertheless useful
to mention the quantitatively most important types of social prefer-
ences that have been established.
A particularly important type of social preference is the preference
for strong reciprocity. A strongly reciprocal individual responds kindly
toward actions that are perceived to be kind and hostily toward
actions that are perceived to be hostile. Whether an action is perceived
to be kind or hostile depends on the fairness or unfairness of the inten-
tion underlying the action. The fairness of the intention, in turn, is de-
termined by the equitability of the payoff distribution (relative to the
set of feasible payoff distributions) caused by the action. It is important
to emphasize that strong reciprocity is not driven by the expectation of
future economic bene¬t. It is, therefore, fundamentally different from
˜˜cooperative™™ or ˜˜retaliatory™™ behavior in repeated interactions. These
behaviors arise because actors expect future economic bene¬ts from
their actions. In the case of reciprocity, the actor is responding to
friendly or hostile actions even if no economic gains can be expected.
Models of strong reciprocity have been developed by Rabin (1993),
Levine (1998), Falk and Fischbacher (chapter 6, this volume), Dufwen-
berg and Kirchsteiger (2004), and Segal and Sobel (1999).2
A second type of social preference is inequity aversion as modeled in
Fehr and Schmidt (1999) and Bolton and Ockenfels (2000). Fehr and
Schmidt (1999) assume, for example, that inequity-averse persons want
to achieve an equitable distribution of economic resources. This means
that they are altruistic towards other persons”that is, they want to in-
crease other persons™ economic payoff if the other persons™ economic
payoffs are below an equitable benchmark. However, inequity-averse
persons also feel envy”that is, they want to decrease the other per-
sons™ payoffs when these payoffs exceed the equitable level. In many
154 Fehr and Fischbacher

situations, strongly reciprocal persons and inequity-averse persons be-
have in similar ways. For example, both strong reciprocity and ineq-
uity aversion imply the desire to reduce the payoff of another person if
that person made a decision such that the payoff of the strongly recip-
rocal or inequity-averse person is much lower than the payoff of the
other person. Recent evidence (Falk, Fehr, and Fischbacher [henceforth
FFF] 2000 and 2001; Offerman 2002) suggests, however, that strong
reciprocity is the quantitatively more important motive.
The similarity in the behavior of reciprocal and inequity-averse per-
sons is due to the fact that both concepts depend in important ways on
the notion of a fair or equitable payoff. Since models of inequity aver-
sion are much simpler and more tractable than models of strong reci-
procity, it is often convenient to ˜˜mimic™™ or to ˜˜black box™™ reciprocal
behavior by inequity aversion (see the section on proximate models of
strong reciprocity later in the chapter). Some authors (e.g., Charness
and Rabin 2002) have also found evidence suggesting that subjects
tend to help the least well off. Such behavior is, however, often not dis-
tinguishable from inequity aversion, in particular, nonlinear inequity
aversion. Recently, Neilson (2000) provided an axiomatic characteriza-
tion of a nonlinear version of Fehr-Schmidt type inequality aversion.
Strong reciprocity and inequity aversion are very different from un-
conditional altruism, which constitutes a third type of social preference.
Unconditional altruists do not condition their behavior on the actions
of others”that is, altruism given does not emerge as a response to
altruism received (Andreoni 1989; Andreoni and Miller 2002; Cox,
Sadiraj, and Sadiraj 2001). In technical terms, unconditional altruism
means that a person values the economic resources allocated to a rele-
vant reference agent positively. An unconditional altruist, therefore,
never takes an action that decreases the payoff of a reference agent.
Yet, as we will see later in the chapter, an important stylized fact con-
cerns people™s willingness to punish other people for unfair or hostile
actions. Unconditional altruism also cannot explain the phenomenon
of conditional cooperation, that is, the fact that many people are will-
ing to increase their voluntary cooperation in response to the coopera-
tion of the others.
Finally, research has also shown that a fraction of the people exhibits
spiteful or envious preferences (FFF 2001). A spiteful or envious person
always values the economic payoff of relevant reference agents nega-
tively. The person is, therefore, willing to decrease the economic payoff
of a reference agent at a personal cost to himself (Kirchsteiger 1994;
The Economics of Strong Reciprocity 155

Mui 1995), irrespective of the payoff distribution and irrespective of
the reference agent™s fair or unfair behavior. Spiteful choices seem to
be quantitatively less important than reciprocal choices. Moreover,
spitefulness (as well as unconditional altruism) cannot explain why
the same people often are willing to help others at a personal cost in
one situation while they harm other people in other situations (FFF
Although previous research clearly indicates that many people ex-
hibit social preferences it is important to keep in mind that not every-
body exhibits social preferences. In fact, most studies indicate that
there is also a substantial number of people who behave in a purely
sel¬sh manner. A key question, therefore, is how the heterogeneity of
motives at the individual level can be captured by parsimonious mod-
els and how the different individual motivations interact. In this chap-
ter, we concentrate on the existence of strongly reciprocal and sel¬sh
types. The reason for this is three-fold. First, empirical evidence clearly
suggests that in the domain of payoff-decreasing or punishing behav-
ior, strong negative reciprocity is the dominant motive (FFF 2000,
2001; Kagel and Wolfe 2001; Offerman 2002). Second, in the domain of
helping or rewarding behavior, strong positive reciprocity also plays
an important role”although other motives like unconditional altru-
ism, inequity aversion, and ef¬ciency-seeking behavior play a role
as well (Cox 2000; Charness and Rabin 2002; FFF 2000; Offerman
2002). For reasons of parsimony we will, however, neglect these other
motives in this chapter.
Theory as well as empirical evidence suggest that the interaction be-
tween strongly reciprocal and sel¬sh types is of ¬rst-order importance
for many economic questions. The reason for this is that the presence
of reciprocal types often changes the economic incentives for the sel¬sh
types, which induces the sel¬sh types to make ˜˜nonsel¬sh™™ choices.
For example, a sel¬sh person is deterred from behaving opportunisti-
cally if the person expects to be punished by the reciprocators. Like-
wise, a sel¬sh person may be induced to behave in a cooperative and
helpful manner because she expects the reciprocators to return the fa-
vor. Since the presence of strongly reciprocal types changes the pecuni-
ary incentives for the sel¬sh types, the strongly reciprocal types often
have a signi¬cant impact on the aggregate outcome in markets and
We focus our presentation on laboratory experiments because it is
impossible to unambiguously isolate the impact of strong reciprocity
156 Fehr and Fischbacher

in most real-life situations. A skeptic may always discount ¬eld evi-
dence with the argument that, in the ¬eld, the notion of fairness is
only used for rhetorical purposes that disguise purely self-interested
behavior in an equilibrium of a repeated game.3 The experimental evi-
dence on ultimatum games (Guth, Schmittberger, and Schwarze 1982)
and the gift exchange games (Fehr, Kirchsteiger, and Riedl 1993) pro-
vides transparent illustrations of strong positive and negative reciproc-
ity.4 Since both games have been described in the introductory chapter
to this volume, we do not describe the experiments in detail here.
It suf¬ces to mention that the rejection of positive offers by the res-
ponders in the ultimatum game can be interpreted as strong negative
reciprocity, whereas the generous effort choices by the employees in re-
sponse to generous wage offers in the gift exchange game can be inter-
preted as strong positive reciprocity. These results have been replicated
in numerous studies”in many countries and across a variety of differ-
ing conditions.5 It is also worth mentioning that strong reciprocity has
been observed in the ultimatum and the gift exchange game even
when the stake size in the experiment has been raised to the level of
three months income (Cameron 1999; Fehr, Tougareva, and Fisch-
bacher 2002).

5.2.1 One-Shot and Repeated Interactions
Sometimes it is argued that reciprocal behavior in anonymous one-shot
experiments is due to subjects™ inability to adjust properly to one-shot
interactions. One idea (see Binmore 1998) is that outside the laboratory,
subjects are typically involved in a network of repeated interactions. It
is well-known from repeated game theory that in repeated interactions,
rewarding and punishing may be in the long-term self-interest of an in-
dividual. According to this argument, subjects who routinely interact
in the repeated game of life import routines and habits that are appro-
priate for repeated interactions into the laboratory™s one-shot situation
because they do not understand the strategic differences between one-
shot and repeated interactions. Therefore, the observation of reciprocal
behavior in one-shot interactions should not be interpreted as a devia-
tion from self-interest but merely as a form of rule-of-thumb behavior,
that is, as a cognitive failure to properly distinguish between one-shot
and repeated interactions.
Our response to this argument is twofold. First, and most impor-
tantly, the argument is refuted by evidence indicating that the vast
The Economics of Strong Reciprocity 157

majority of the subjects understand the strategic differences between
one-shot and repeated interactions quite well. Later in this chapter we
discuss the experiment reported by Fehr and Fischbacher (2003) that
explicitly tested for this phenomenon. Second, even if the argument
were correct, social scientists would have strong reasons to take the
habits and routines that shape people™s behavior into account. The im-
portance of reciprocal behavior for the social sciences does not depend
on whether it is interpreted as a deviation from self-interest or as a
form of bounded rationality. Reciprocal behavior is important because
it affects the functioning of markets, organizations, incentives, and col-
lective actions in fundamental ways.
In principle, it is testable whether people have the ability to distin-
guish temporary one-shot play from repeated play. Fehr and Fisch-
bacher (2003) investigated this problem in the context of the ultimatum
¨chter and Falk (2002) provided evidence for the gift ex-
game, and Ga
change game. Fehr and Fischbacher conducted a series of ten ultima-
tum games in two different conditions. In both conditions, subjects
played against a different opponent in each of the ten iterations of the
game. In each iteration of the baseline condition, the proposers knew
nothing about the past behavior of their current responders. Thus, the
responders could not build up a reputation for being ˜˜tough™™ in this
condition. In contrast, in each iteration of the reputation condition, the
proposers knew the full history of the behavior of their current res-
ponders, that is, the responders could build up a reputation for being
˜˜tough.™™ In the reputation condition, a reputation for rejecting low
offers is, of course, valuable because it increases the likelihood to re-
ceive higher offers from the proposers in future periods.
If the responders understand that there is a pecuniary payoff from
rejecting low offers in the reputation condition, one should in general
observe higher acceptance thresholds in this condition. This is the pre-
diction of the social preferences approach that assumes that subjects
derive utility from both their own pecuniary payoff and a fair payoff
distribution. If, in contrast, subjects do not understand the logic of rep-
utation formation and apply the same habits or cognitive heuristics to
both conditions, one should observe no systematic differences in res-
ponder behavior across conditions. Since the subjects participated in
both conditions, it was possible to observe behavioral changes at
the individual level. It turns out that the vast majority (slightly more
than 80 percent, N ¼ 72) of the responders increase their acceptance
158 Fehr and Fischbacher

thresholds in the reputation condition relative to the baseline condi-
tion.6 This contradicts the hypothesis that subjects do not understand
the difference between one-shot and repeated play.

5.3 Competition

In this section, we illustrate our claim that it is not possible to un-
derstand the effects of competition if concerns for fairness and strong
reciprocity are neglected. We will show, in particular, that the self-
interest hypothesis hinders social scientists from understanding impor-
tant comparative static effects of competition. In addition, we present
results indicating that competition may sometimes have much less im-
pact than predicted by models based on the self-interest hypothesis.

5.3.1 The Effects of Competition under Exogenous Contract
Consider the following slightly modi¬ed ultimatum game. Instead of
one responder, there are now two competing responders. When the
proposer has made his offer, the two responders simultaneously accept
or reject the offer. If both accept, a random mechanism determines
with probability 0.5 which one of the responders will get the offered
amount. If only one responder accepts, he will receive the offered
amount of money. If both responders reject, the proposer and both res-
ponders receive nothing.
This version of the ultimatum game”with responder competition”
can be interpreted as a market transaction between a seller (proposer)
and two competing buyers (responders) who derive the same eco-
nomic payoff from an indivisible good. Moreover, if the parties™ pe-


. 5
( 13)