Cognitive Adaptations for n-person Exchange: The ... · Cognitive Adaptations for n-person...

MANAGERIAL AND DECISION ECONOMICS

Manage. Decis. Econ. 27: 103–129 (2006)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/mde.1287

Cognitive Adaptations for n-personExchange: The Evolutionary Roots

of Organizational Behavior

John Toobya,*, Leda Cosmidesa and Michael E. Priceb,c

aCenter for Evolutionary Psychology, Department of Anthropology, University of California,Santa Barbara, CA 93106-3210, USA

b Indiana University Workshop in Political Theory and Policy Analysis and The Santa Fe Institute,Indiana University, 513 N. Park, Bloomington, IN 47408-3895, USA

cOlin School of Business, Washington University in St. Louis, Campus Box 1133, 1 Brookings Drive,St. Louis, MO, 63130-4899, USA

Organizations are composed of stable, predominantly cooperative interactions or n-personexchanges. Humans have been engaging in n-person exchanges for a great enough period of

evolutionary time that we appear to have evolved a distinct constellation of species-typical

mechanisms specialized to solve the adaptive problems posed by this form of social interaction.

These mechanisms appear to have been evolutionarily elaborated out of the cognitiveinfrastructure that initially evolved for dyadic exchange. Key adaptive problems that these

mechanisms are designed to solve include coordination among individuals, and defense against

exploitation by free riders. Multi-individual cooperation could not have been maintained overevolutionary time if free riders reliably benefited more than contributors to collective

enterprises, and so outcompeted them. As a result, humans evolved mechanisms that

implement an aversion to exploitation by free riding, and a strategy of conditional

cooperation, supplemented by punitive sentiment towards free riders. Because of the designof these mechanisms, how free riding is treated is a central determinant of the survival and

health of cooperative organizations. The mapping of the evolved psychology of n-partyexchange cooperation may contribute to the construction of a principled theoretical

foundation for the understanding of human behavior in organizations. Copyright # 2006John Wiley & Sons, Ltd.

EVOLUTIONARY FORMS OF

RATIONALITY: THE EVOLUTIONARY AND

COGNITIVE BACKGROUND TO EXCHANGE

AND MULTI-INDIVIDUAL COOPERATION

Voluntary exchange for mutual benefit improvesthe net welfare of participants. Consequently,traditional economic and social science theoriesoften treat exchange simply as the straightforwardproduct of human rationality plus self-interest,

transparently based on general-purpose abilities toreason, pursue goals and select the highest payoffs.Although collective action has a similar ability toimprove the net welfare of participants, econo-mists and organizational theorists have longrealized that there was a problem in attemptingto explain collective action as a parallel expressionof rationality plus self-interest (Olson, 1965;Ostrom, 1998; Price et al., 2002). In n-partyinteractions that produce public and commongoods (that is, produce outcomes where individualeffort is initially unlinked to the ability to consumethe benefits of a joint effort), free riding provideshigher payoffs than contributing. As a result,

Copyright # 2006 John Wiley & Sons, Ltd.

*Correspondence to: Center for Evolutionary Psychology,Department of Anthropology, University of California, SantaBarbara, CA 93106-3210, USA. E-mail: [email protected]

rational agents}who according to standard eco-nomic theory will be essentially everyone}willchoose not to contribute. Because of this free riderproblem, it is puzzling why any rational, self-interested individual would contribute at all to acollective action. Yet, across all known cultures,throughout history, humans have commonly en-gaged in collective actions (Ostrom, 1990; Priceet al., 2002; Tooby and Cosmides, forthcoming).Where interactions are small in scale, joint effortsare routine. How can this be?

Locating Rationality, Choice, and Organizational

Behavior in an Evolutionary Psychological

Framework

Although contributing to collective actions ap-pears to violate standard rationality and the theoryof choice behavior that emerges from it, rationalityitself appears less straightforward when theattempt is made to locate formal principles ofrationality and economic choice in the causalmatrix of the physical world. First, rationality andchoice are not, of course, mechanism-free, but arereal solely to the extent they are embodied in theinformation-processing architecture of neural pro-grams in the human brain. Economic and organi-zational behaviors do not flow from disembodiedprinciples, but are the computed outputs ofstructured cognitive mechanisms incarnated inbrain organization. So, cognitive science is onefoundation needed to understand human econom-ic and organizational decision-making. Second,species-typical mechanisms of reasoning andchoice, in whatever form they exist, came intobeing because they (or their developmental bases)were produced by the evolutionary process. So,evolutionary biology provides a second founda-tion for understanding human decision-making.

Evolutionary psychology unites the two projectsof evolutionary functionalism and cognitivescience into a single integrated research program(Tooby and Cosmides, 1992a). It explicitly usesknowledge of what natural selection would havefavored during human evolution to guide theempirical mapping of the computational architec-ture of the human mind/brain. Thus, it asks twosets of interrelated, complementary, and mutuallyilluminating questions. First, what are the actualcomputational procedures that the human mind/brain embodies in its reasoning, emotion, andmotivational systems? Second, which aspects of

these procedures constitute functional designfeatures that led natural selection to incorporatethem into the human species-typical architecture?That is, what adaptive problems are these proce-dures designed to solve, and in what way do theirpatterned outputs correspond to solutions? Evolu-tionary psychology has been effective as a researchstrategy because understanding the adaptive pro-blems faced by our ancestors provides detailedpredictions about the designs}the functionalarchitecture}of our evolved psychological me-chanisms, guiding new empirical initiatives.

Consequently, taking an evolutionary psycho-logical approach offers a new framework for thestudy of rationality, choice, and organizationalbehavior that differs in certain respects from moretraditional approaches. Because reliably develop-ing species-typical mechanisms are the product ofevolution, the functional logic they embody will bea species of evolutionary rationality}what wehave called ecological rationality (Tooby andCosmides, 1992b). Ecological rationality differsfrom ordinary concepts of rationality and func-tionality in several ways:

First, the functional product that evolvedmechanisms are designed to produce is not utilitymaximization, social welfare or the general abilityto realize goals (although it may overlap with theseunder many circumstances). Instead, the mechan-isms were designed (in interaction with the othermechanisms in the architecture) to produce out-puts that typically promoted genic fitness underancestral conditions (that is, that increased thefrequency of the mechanisms’ genetic basis insubsequent generations relative to pre-existing ormutational alternatives). This definition of func-tionality often departs from both formalized nor-mative theories of rationality and common senseintuitions about functionality. For example, me-chanisms involved in dyadic or n-person exchangemay, by design, cause choices that do not maximizeabsolute payoffs, violating standard rational prin-ciples (e.g. Hoffman et al., 1998; Price et al., 2002).Similarly, mechanisms underlying jealousy violatecommon sense notions of functionality. Jealousyoften causes desperate unhappiness for the in-dividual in which it is activated, as well as for itstargets}a byproduct of the fact that its function isto spread the genetic basis of jealousy at theexpense of the alternative design (indifference tothe sexual behavior of one’s mate), regardless of itsimpact on the jealous individual’s happiness.

J. TOOBY ET AL.104

Copyright # 2006 John Wiley & Sons, Ltd. Manage. Decis. Econ. 27: 103–129 (2006)

Second, these mechanisms were designed to beevolutionarily functional within the ecological(causal) structure of the ancestral world humansevolved in, but not necessarily in environmentswhose structure departs from ancestral conditions.On this view, modern ‘irrationality’ will commonlybe the expression of ancestral functionality: Ataste for salt evolved under conditions wheresalt}a necessary nutrient}was so chronicallyscarce that modern humans ‘irrationally’ over-consume it now that it is abundant.

Third, our cognitive architecture system has notbeen designed by evolution to reach with equalefficiency any goal that arbitrary preferences mightnominate. Rather, our species-typical neurocom-putational architecture evolved to solve particularfamilies of adaptive problem that recurred fre-quently enough to select for mechanisms to solvethem. Hence, the design of specific problem-solving systems is matched to the recurrentstructure of ancestrally significant adaptive pro-blems just as specific keys are machined to fitparticular locks.

How does this apply to understanding multi-individual cooperation and collective action? Wethink that human evolutionary history hasequipped the human mind with specialized psy-chological adaptations designed to realize gains intrade that occur both in 2-party exchanges and inn-party exchanges, including collective actions. Webelieve that the specific characteristics of thesemechanisms (e.g. cheater detection circuits) reflectthe ancestrally recurrent structure of these adap-tive problems (the existence of payoffs to cheat-ing), just as parts of the key shape complementdetailed parts of the lock. We think the under-standing of collective action and organizationalbehavior can be improved by exploring theproperties of this evolved psychology. We expectthat emerging evolutionarily psychological the-ories of n-person exchange and collective actionwill replace theories of standard economic ration-ality by explaining their puzzles. In particular, wethink that violations of standard economic ration-ality that are often found in situations studied byresearchers in organizational behavior, behavioraleconomics, and anthropology are expressions ofwell-engineered, ecologically rational mechanismsproducing outputs that would have been advanta-geous in ancestrally typical conditions. Indeed, wethink that many of the ‘irrational’ behavioralexpressions of these mechanisms (such as voting

behavior or donating blood) will come to berecognized as engineering byproducts of thesefunctional designs when they are activated outsideof the ancestral envelope of conditions for whichthey were designed.

A Surprise: Exchange is not Produced by General

Rationality but by Evolved Neurocomputational

Programs Functionally Specialized for

Exchange Interactions

One of the first applications of the evolutionarypsychological research program was to the phe-nomenon of dyadic exchange (i.e. where twoparties deliver benefits to each other, each deliverybeing made conditional on the other). The primaryquestions addressed were: What do the actualcomputational procedures underlying exchange inhumans look like, and how do their design featuresreflect solutions to the adaptive problems imposedby the recurrent structure of ancestral socialexchanges? Perhaps the largest surprise to emergefrom this research was the finding that exchange,despite being precisely the kind of thing thatgeneral rationality would guide rational agents toengage in, is not after all produced by a general-purpose set of reasoning abilities. That is, thetraditional view that exchange is the expression ofgeneral economic rationality turns out to be false.Instead, exchange is produced by evolved reason-ing specializations}social exchange procedur-es}tailored by natural selection to solve thecomputational problems specific to social ex-change (Cosmides, 1989; Cosmides and Tooby,1989, 1992, 2005; Hoffman et al., 1998). Whatleads to this conclusion? Exchange depends onconditionally delivered behavior (e.g. I will do X ifyou do Y), and so necessarily requires conditionalreasoning for its regulation. In theory, of course,this could be accomplished by general reasoningabilities}the traditional assumption. Yet whenmethods were developed to study conditionalreasoning performance in humans, subjects turnedout to be very poor at detecting potentialviolations of conditional rules across a very broadrange of familiar and unfamiliar contents (seereview in Cosmides, 1989). One could not attributethe human ability to engage in exchange to ageneral ability to reason about conditionals.

Based on an adaptationist analysis of exchange,it was hypothesized that humans have an evolvedcognitive specialization for reasoning about social

COGNITIVE ADAPTATIONS FOR N-PERSON EXCHANGE 105


exchange, including a subroutine for detectingcheaters (Cosmides, 1989; Cosmides and Tooby,1989). This led to a number of novel and specificpredictions. For example, despite being generallypoor at detecting potential violations of condi-tional rules, subjects should nevertheless detectthem readily when the rule involves social ex-change and looking for violations corresponds tolooking for cheaters. Second, subjects’ reasoningchoices should correspond to the adaptivelycorrect choices laid out in a specialized logic ofsocial exchange, even when these choices conflictedwith logically correct choices. These and relatedpredictions were subsequently confirmed by nu-merous experiments conducted on subject popula-tions drawn from many societies, from Harvardundergraduates to Amazonian hunter–horticultur-alists (for review, see Cosmides and Tooby, 2005).Indeed, evidence from cognitive neuroscienceshows it is possible through brain damage to havethe mechanisms underlying social exchange rea-soning selectively impaired, while other reasoningabilities remain intact}something that wouldnot be possible if reasoning about differentcontents were accomplished using the same setof general-purpose rational mechanisms (Stoneet al., 2002.)

Identifying the fitness advantages of exchange,of course, was not a deep evolutionary puzzle: Ashas been recognized at least since Adam Smith (ifnot for hundreds of thousands of generations),exchange increases the welfare of both parties.What is critical from an evolutionary perspective iswhether the mechanisms that cause us to engage inexchange have been shaped in any specific ways byselection pressures that are particular to exchangeinteractions (Cosmides, 1989; Cosmides and Too-by, 1989). If exchange had turned out to be simplyone out of an indefinitely large set of activitiesmade possible by general learning abilities orgeneral rationality, then an evolutionary approachwould have had little to contribute. Of course, theevolutionary biology community has not widelyaddressed the superordinate category of exchangeper se (Tooby and Cosmides, 1996), but primarilya narrower subset of exchange interactions: alter-nating, deferred (usually implicit) exchange. (Tri-vers, 1971) confusingly labeled this behaviorreciprocal altruism, although it did not meet thebiological or common sense definition of altruism.We will use reciprocation to refer to alternatingdeferred exchange and exchange or reciprocity to

refer to the superordinate category of intercontin-gent interactions involving gains in trade).

George Williams was, characteristically, the firstto introduce the topic of reciprocation intoevolutionary biology, perceptively identifying thekey strategy as one of the conditionality of one actof benefit delivery on the other (Williams, 1966).Trivers (1971) elaborated Williams’ insights into afar richer treatment, in the light of the alreadyexisting experimental game theory literature oniterated prisoners’ dilemmas and the social psy-chology literature on reciprocity. This was fol-lowed by more formal and systematic analysesusing evolutionary game theory, such as Axelrodand Hamilton’s exploration of tit-for-tat andMaynard Smith’s ESS analyses of reciprocation(Axelrod and Hamilton, 1981; Maynard Smith,1982). One result that robustly emerges from mostevolutionary game theory formalizations of reci-procation is that it cannot easily evolve unlessreciprocation strategies avoid cheaters. This im-plicitly assumes that reciprocation strategies havethe capacity to detect instances of cheating(noncompliance, defection). Thus, although thefitness advantages of exchange are no puzzle, andtheir elucidation is a ratification rather than acontribution to economics, the theoretical analysisof reciprocation highlighted an evolutionary vul-nerability to the strategy that our cognitive andmotivational machinery would necessarily havebeen selected to address: A reciprocation strategy,to be successful, must incorporate defenses againstbeing outcompeted by cheaters. This result appliesnot only to reciprocation, but to the moreencompassing category of exchange (Cosmidesand Tooby, 1989, 1992). This conclusion led tothe prediction and discovery that humans have acognitive specialization that allows them to detectviolations of conditional rules when the condi-tional relationship involves exchange and a viola-tion would constitute an act of cheating (Cosmidesand Tooby, 2005).

If 2-party exchange is not caused by generalrationality, then contributing to collective action isnot}after all}an anomaly because it departsfrom general rationality. Because general ration-ality (as a set of real, physical mechanisms)seemingly does not exist, departures from it donot require explanation.1 Instead, both types ofbehavior are explained by the actual designs of theevolved neurocomputational programs that causethem. But what exactly are these designs, and how

J. TOOBY ET AL.106


could they have emerged over evolutionary time inthe face of potentially fitter competitive challengesby alternative designs? On this view, both 2-personand n-person exchange behavior appear (at firstexamination) to be anomalous from an evolu-tionary perspective because both suffer fromparallel vulnerabilities to evolved counterstrategiesthat should have outcompeted them: Cheaters andfree riders both take benefits without makingcontributions, and so would have outcompetedfrom the beginning simple first-order designs for 2-party or n-party unconditional ‘exchange’ (benefitdelivery). Consequently, what an evolutionaryperspective predicts instead is that the mechanismsthat evolved must necessarily embody morecomplex, conditional exchange strategies whosecomputational properties defend them againstoutcompetition by cheaters (in 2-party exchange)and free-riders (in n-person exchanges such ascollective actions). These strategies do not just pickthe highest absolute payoffs}they often sacrificesuch payoffs in favor of enduring practices thatgenerate higher relative payoffs against exploitivestrategies, when averaged across all interactions.Antiexploitation computational elements are thuscentral and indispensable design features of acooperative strategy}essential if the strategy is tosuccessfully emerge and stably persist. Hence, therelevant question becomes: What exactly do theseexploitation-resistant strategies look like, whendescribed as psychological (information-proces-sing) designs?

The Functional Logic that Regulates Motivation is

an Integral Part of Evolved Social Rationality

Another cardinal difference between standardapproaches to rationality and evolutionary ap-proaches to rationality involve the role thatmotivation plays in decision-making. Instead ofpreferences being arbitrary, wholly acquired, orexogenous to rationality, evolved motivationalspecializations are (we believe) intrinsic compo-nents of more encompassing, ecologically rationalproblem-solving adaptations, such as adaptationsfor 2-party exchange and collective action. That is,motivational mechanisms are endogenous andindissoluble constituents to systems of rationalityand cognition, with regulatory architectures thatemploy evolved, proprietary forms of representa-tion in order to direct motivation so thatindividuals correctly implement solutions to adap-

tive problems (Tooby et al., 2005). Thus, a fear ofsnakes is not the product of a general rationalsystem for avoiding negative payoffs, but rather anevolved system with evolved proprietary represen-tations (snake) and motivational dispositions(fear/avoidance). This system is ecologically ra-tional in that its specialized motivational circuitsmotivate behavior that was ancestrally functio-nal}snake avoidance}even though it may some-times lead to unnecessary (‘irrational’) avoidanceof some harmless snakes (i.e. it is not perfect). Thekey claim is that for motivation to direct behavioradaptively, it requires guidance systems equippedwith evolved representational elements capable ofcorrectly specifying the targets and intensities ofthe motivation. Examples of such target-specifying‘innate ideas’ that reliably develop within then-person exchange motivational system includecontributor, free rider, benefit to the group,contribution required for entitlement, individualreceipt of joint benefit, entitlement to benefit,undercontribution, exploitation, and so on. Theseevolved conceptual elements are necessary todirect the specific motivations that can lead tothe realization of potential gains in trade whilesimultaneously defending against exploitation.

Properly targeted and calibrated motivationalcircuits are central to defending against free riders(and cheaters), and hence essential to makingmulti-individual cooperation (and dyadic ex-change) evolutionarily stable against exploitation.This regulation is delivered through a set ofmotivational specializations that, for example:

(1) calibrate willingness to devote effort tocollective projects,

(2) direct punitive sentiment toward free riders,and

(3) mobilize reward sentiments towards contri-butors.

It is important to emphasize that the controlsystem that regulates the deployment of thesesentiments is cognitive or computational in nature,and must apply procedures to representations andinputs to determine, for example, under what con-ditions to feel punitive sentiment, how much to feel,and toward whom to feel it. Despite the frequentmystification of feeling, feelings are not blind,ineffable, randomly emerging forces, but areneurally computed outputs of evolved programswhose structure can be mapped and whose func-tions can be identified (Tooby and Cosmides, 2005).



THE EVOLVED FUNCTIONAL DESIGN OF

THE PSYCHOLOGICAL MACHINERY

UNDERLYING N-PERSON EXCHANGE

Using the foregoing framework, we would like toexplore how multi-individual cooperation (n-per-son exchanges, including collective action) as apervasive human activity can be accounted for. Wewould like to sketch out some of the cognitive andmotivational design features that we proposeunderlie the human ability to engage in collectiveenterprises. We would like to link these designfeatures in a general way to selection pressures andancestral situations that we believe led to theirincorporation into the human psychological archi-tecture.

The evolution of the ability to engage in 2-partyexchange would almost certainly have precededthe evolution of the ability to navigate the gametheoretic complexities of n-party exchange andcollective action. Consequently, we would like toconsider what additional components would havehad to be added to cognitive mechanisms fordyadic social exchange in order to extend thesystem to n-person exchange, including the specialcase of collective action. Finally, we will considerhow such a model fares against criticism thatmulti-individual cooperation, including collectiveaction, could not have evolved via individualselection for such exchange strategies.

Dyadic Exchange Builds Many of the Components

Necessary for n-Person Exchange

The first claim is that the evolution of mechanismsfor 2-party exchange built a large part of thecomputational infrastructure necessary for enga-ging in n-party exchange (see, e.g. Cosmides andTooby, 1989 for a detailed description of mechan-isms; for a review of evidence, see Cosmides andTooby, 2005). The evidence indicates that thesocial exchange algorithms that evolved forexchange use conceptual primitives such as self;agent (or party); the welfare or interest of a party;exchange; entitlement to benefit, benefit to begained; cost or requirement to be met to gainentitlement to the rationed benefit; cheater (a partythat has taken the benefit while having intention-ally not met the requirement); intended outcome vsaccidental outcome; consent to a social contract (anintercontingent plan of action in which each partyagrees to undertake a course of action conditional

on the other party’s execution of a correspondingcourse of action); and so on. The exchange systemalso involves specialized procedures that interre-late these concepts, map actual situations intothese representations, evaluate magnitudes, andcompute necessary regulatory outputs. Two ob-vious examples are a system that scans the socialand instrumental world for opportunities toincrease welfare through exchange, and the look-for-cheaters algorithm.

Many of these components are not proprietaryto the exchange system, but evolved because theyare more broadly useful across a wide array offunctions. For example, to plan an agent needsmechanisms of valuation}that is, procedures thatcan turn representations of a given situation(actual or imagined) into a magnitude representingthe welfare that the situation holds for the agent.Similarly, agents must be equipped with proce-dures that represent potential or actual changes insituations as positive or negative magnitudes inwelfare or interest, which generates the perspec-tive-specific concepts of benefits, and costs. Hu-mans can do this for themselves, and attribute thisability to other agents (i.e. have a capacity tointerpret behavior in terms of an intuitive theory ofinterests that is, we think, one subcomponent ofthe evolved theory of mind system (Baron-Cohenet al., 1985)). Interests are representations of thesets of changes to situations that positively ornegatively modify an agent’s welfare. Acts cansimultaneously modify the welfare of more thanone party. When this is foreseeable such behavioris commonly interpreted as expressing a stableinternal variable that regulates how much thewelfare of the self is traded off against the otherparty: a welfare trade-off ratio or WTR (Toobyand Cosmides, 2002, 2005, forthcoming). Thus, ifone party reliably incurs high costs in order todeliver benefits to another, that behavior expressesa high welfare trade-off ratio toward the recipient.

When single exchanges morph into extendedseries or enduring exchange relationships, theymay not be structured in the long run by tit-for-tat-like rules (which require alternating acts ofreciprocation), but rather by a rule that matchesone exchange partner’s welfare trade-off ratio tothe other’s, scaled by the relative symmetry andfrequency of each party’s opportunities to help theother (e.g. Tooby and Cosmides, 1996, 2002). Thisexchange psychology represents what are, in effect,accounts}how much each party has done for the

J. TOOBY ET AL.108


other. Accounts and their interrelationships (suchas a ‘balance of trade’ index) generate conceptualelements such as owe, debt, obligation, exploitation,cheating, and so on. Computations on theseaccounts and their interrelationships also regulatemotivation: Others’ kindness to us moves us (i.e.recalibrates our welfare trade-off ratio towardthem so that it is higher than it was before). Theirindifference, unwillingness to help, or their cheat-ing harden our hearts (i.e. downregulates ourwelfare trade-off ratio with respect to them). Ifexploitive or deceptive enough, these acts maymotivate anger and punitive sentiment towards theoffender. Obviously, this neurocomputational ma-chinery is beyond conscious awareness, althoughwe are aware of some of its outputs.

Recursion as One Modification to 2-party Exchange

Psychology that can Help to Adapt it to Solving the

Problems Arising from n-party Exchange

Understanding the relationships that 2-party ex-change embodies seems effortless and intuitive. Inreality, these interactions involve a surprisinglylarge number of interrelationships, variables,steps, conditional branch points, unexpressedcontingencies, and so on (see Cosmides andTooby, 1989, for a list of the conditional relation-ships required to encapsulate a simple dyadicexchange). What makes this maze transparent,intuitive, and navigable is that humans areequipped with evolved machinery designed toautomatically represent and track this intricateset of implicated relationships. What would needto be modified in the architecture of 2-partyexchange to allow it to be able to coordinate thebehavior of three, four, or more individuals torealize n-person gains in trade?

Theoretically, much of what is needed could beachieved through a relatively modest change to thearchitecture: adding recursion to the exchangestructure. The pre-existing computational templateused for navigating 2-party exchange could beleveraged to navigate n-party exchange throughadding the ability to reduplicate an additionalagent-centered unit for representing exchangerelations for each added person. That is, if the 2-party form can be abbreviated as ‘I will do x if youdo y’, then the generalization of this intercontin-gent regulatory structure to n-person social ex-change could be abbreviated as ‘For individuals 1though n, I (individual1) will do x1 if individual2

does x2 and individual3 does x3 . . .’ (or ‘I will ifyou1, you2, you3. . .will.’). It is important torecognize that the gain that is produced need notbe a public good, but can be a series of privategoods (for benefit1, for benefit2, for benefit3. . .).

Discoordination and Complexity as Factors that

Sharply Limit the Scope of n-Party Exchange

One important limiting factor for n-person ex-change is the problem of reaching and sustainingcoordination among the interactants. Complexityin the set of possible alternative arrangements andbehavioral interdependencies mounts explosivelyas the numbers in the exchange grow. The firstproblem is that of establishing a mutual under-standing among many different minds. In dyadicexchange, this is relatively easy: If the proposedexchange is not worthwhile for one participant,then the other participant increases the value ofthe offer (or refuses to engage in the exchange). Topersuade a holdout in a multisided negotiation,however, it is in every other party’s interest to haveothers contribute the needed additional induce-ment. This leads to an n�1 sided game of chicken.Similar negotiative problems and instabilitiesemerge, for example, in deciding who is to dowhat, whenever complementary tasks with differ-ent costs need to be performed.

Another important limiting factor for n-personexchange is the rapidly increasing cognitivedemand posed by the rapid growth of combina-torial complexity as the numbers in an exchangegrow. For example, if a participant has to divideher attention among an increasing number ofexchange partners, how are instances of under-contributing or other forms of exploitation to bedetected? With increasing numbers of participants,the problems of negotiation, communication,coordination, and agreement become far morecomplex. Complexity is injected by the number ofdifferent roles individuals could take, by thenumber of different goals the group could pursue,by the possibility of different costs being incurredby different participants doing their respectiveparts, by the possibility of different individualsderiving different magnitudes of benefit fromthe n-sided exchange, and so on. Limitations onthis type of exchange emerge from logisticalconstraints in negotiating and implementingn-sided exchanges, cognitive bottlenecks inattending and representing, and corresponding



limitations of the ability to defend oneself againstexploitation within them. These problems arerelatively small if the set of interactants is small,such as three, four, five, or six. But with increasingnumbers of interactants, communication becomesnoisy, mutual surveillance becomes difficult, cheat-ing becomes easier, and transaction costs becomeprohibitive.

For these reasons, it is inevitable that in theregular course of human life immense numbers ofmutually beneficial potential gains in trade from n-person exchange are never realized. Nevertheless,the human ethnographic record indicates that overour evolutionary history humans have been able tosuccessfully carve out and master some subsetsfrom the space of n-person exchanges (Ostrom,1990; Price et al., 2002). The persistence of theseopportunities would have favored selection forcognitive and motivational mechanisms that couldcost-effectively implement and increase the scopeof the various families of solution.

Different Modes of n-Party Exchange Exploit

Different Strategies for Solving Coordination

Problems

We think that there exist a number of different butinterrelated pathways to solving coordinationproblems, and that each has left its imprint onour evolved social psychology. For example, thesecoordinative problems selected for specializedcognitive and motivational adaptations that createthe uniquely human interlocking roles of leader-ship and followership. Humans find these conceptsand behaviors intuitive, and they appear sponta-neously in all human cultures (within delimitedsocial contexts). That is, we think a set ofneurocomputational adaptations subserving lea-dership and followership evolved as solution to thecoordination problems that emerge in complexexchanges and intercontingent social structures(Tooby and Cosmides, 2002, forthcoming). Lea-dership allows efficient coordination by routingcoordinative decisions through a single individualmind so their mutual implications can be compu-tationally integrated rapidly and dynamically.Leadership (when competent) simplifies the dy-namics of determining new projects and allocatingefforts to achieve them}something that becomesincreasingly important when the numbers in apotentially cooperating group get larger than ahandful. Cheating in leadership involves directing

the group in excessively self-interested, group-injurious ways, while in followership it encom-passes not only general undercontribution, butalso the more specific concepts of unreliability,disloyalty and disobedience (discoordination). Byeffective use of joint attention, persuasion, directeddisapproval and approval, threat, punishment,and reward, leaders can collapse discoordinatingpossibilities entertained by the other n�1 membersof the group into a single shared representation ofa plan of action and benefit allocation. The factthat such a plan is shared itself should excitegreater willingness to engage in an n-personexchange, precisely because a lack of a sharedrepresentation of coordinated action is one of theprimary obstacles to achieving n-person gains intrade.

Indeed, certain events that invite a common andmutually manifest response can sometimes sub-stitute for leadership in solving coordinationproblems. If an event draws joint attention, andinherently invites the same kind of response frommost observers, then the necessary coordination totrigger a collective effort can be triggered. Theright kind of event can spontaneously lead toacephelous defense during attack, hunts precipi-tated by the sudden appearance of game animals,emergency-provoked rescues, and outrage-pro-voked mobbing, riots, and so on. The key featureof event-organized exchanges is existence of asingle supersalient event that not only coordinatesthe minds of the co-participants, but makes it clearfor each participant that this coordination ispresent in the minds of the other participants.Implicitly present is a shared appreciation of thepotential for gains offered by coordination, whichactivates motivation once the near achievement orreality of coordination is perceived.

A third pathway that also collapses many ofthese problems into manageable units is theactivation of a mutually enforced system of strictegalitarianism of external conditions and mar-kers}what Alan Page Fiske calls equality match-ing (Fiske, 1991). This is not an equality of costs,benefits, and inner accounts, but of externallydemarcated units of shares and burdens (e.g. oneman, one vote), greatly simplifying monitoringagainst exploitation. We think this intuitivemethod of social organization is anchored in ourevolved exchange circuitry as one natural mode ofn-person exchange. In general, n-person exchangesare far more cognitively tractable to the extent that

J. TOOBY ET AL.110


contributions from the participants consist ofparallel behaviors: If everyone is expected tocontribute in the same way, then it is vastly easierto communicate and to monitor such an exchange.

Another natural mode involves collective actionto produce public goods, policed by punitivesentiment against free riders (Price et al., 2002;Price, 2003, in press A, B). The entire destabilizingnegotiation over the allocation of a jointlyproduced resource or desirable state of affairs iseliminated when what is produced is a publicgood-like diffused benefit. Collective actions forpublic goods simplify the issue of defendingagainst exploitation by focusing surveillance onthe issue of undercontribution. If the good to beproduced is not excludable, then producers cannotbe cheated by being denied consumption of theproduct of the joint effort. As we will explain,many groups that may initially emerge to pursuenonpublic goods eventually also produce publicgoods. This makes the evolved psychology ofcollective action relevant to most and perhaps allcooperating groups.

The Resuscitation or Perpetuation of a Pre-existing

n-Person Exchange System is Another Pathway to

Solving the Problems of Coordination

Because successfully coordinated n-person ex-change is so difficult to achieve, its existence onceattained becomes a kind of resource to be valuedand effortfully perpetuated by its participants.That is, another condition that greatly simplifiessubsequent coordination and the fruits that itmakes available is the prior existence of somepreviously existing set of understandings thatarose from a previous coordinated interaction.This means that it is in the interests of participantsto give the precedents set by a previous n-personcoordination a weight and an inertia that theywould not necessarily give those expressed inprevious 2-person exchanges, which can be farmore easily generated, modified, and discarded asneeded. The coordination problems inherent inestablishing n-person exchanges make it far toocostly simply to create them de novo as 1-shotarrangements in the face of every new opportunityfor gains in trade. Because the great difficulty ofarriving at coordination usually prevents mutuallybeneficial n-person exchanges from coming intoexistence where n is large, the default willingnessto invest effort in their creation for a specific

short-term end should be low. Correspondingly,however, the existence of recent and mutuallyknown precedents gives a coordinative startingpoint that allows potential interactants to circum-vent considering, negotiating, and sorting throughthe huge number of logically possible exchangearrangements that arise in situations involvingmore than a few people. Hence, the prior existenceof n-person exchange structures should facilitatethe motivation to engage in such exchanges byoffering the opportunity for lower cost coordina-tion. The magnitude of this effect should be afunction of the number of individuals involved.The larger the set of minds to be coordinated, thegreater the cost of arriving at a new set of mutuallyconsistent representations. This means that thegreater the number of participants is, the greaterthe comparative advantage of conservativelyperpetuating pre-existing arrangements will be,however beneficial or flawed those pre-existingarrangements were.2

The key point is that our evolved coalitionalpsychology is designed to appreciate that it iseasier to resuscitate or perpetuate successfulcoordinative arrangements than to create them.3

In reality, groups do not objectively exist}theyonly exist to the extent that they are represented inmutually consistent ways in the minds of assortedindividuals. So, groups only persist if individualsrepresent them as having continuity over time. Ourcognitive circuits for representing groups areadvantageously designed to endow them withcontinuity.

The coordinative arrangements of an existing n-person exchange system themselves constitute auseful resource or instrumentality. This is becausethey make it possible for the individuals in theexchange system (coalition) to consistently takeadvantage of the flow of opportunities to generatejoint benefits that uncoordinated individuals couldnot. That means that collective efforts do notsimply emerge rationally at the time individualscome to anticipate the joint benefits that could beattained from reaching a specific goal, and ceasewhen the goal has been reached. Instead, ourminds evolved to inherently value the existence ofthe groups for their own sake, so that we have asuitable coordinative structure ready and waitingwhen the need strikes. Just as humans value anyother intrinsic good, we are designed to prize theexistence of and continuation of the groups we aremembers of. We feel pleasure upon becoming a



valued member of a group, satisfaction in itscreation and successes, and sadness at its dissipa-tion. Our evolved psychology includes a motiva-tional variable that tracks the social sufficiency ofthe group or groups the individual is a member of.This variable regulates the appetite for finding ordeveloping groups if the individual lacks asufficient number of sources of support. We seekto acquire, create, and perpetuate group identities.

Groups Tend to Turn into Coalitions, Imposing

Exchange Obligations on their Members

For this reason, there is a tendency for groups tooutlive whatever rationales, circumstances, orevents gave rise to their creation,4 and to graduallymorph into what we call amplification coalitions(Tooby and Cosmides, 2002, forthcoming). Thatis, the aboriginal n-party exchange system is onewhose purpose is the amplification of the ability ofeach of its members to realize her interests in dailyevents by cost-effectively combining welfare trade-offs and joint efforts from the other members.Alexandre Dumas memorably encapsulated thenature of the amplification coalition as all for oneand one for all (Dumas, 1844). Dyadic alliances orfriendships are minimal 2-party instantiations ofsuch a coalition, but they can be easily extended toinclude more individuals. Gangs, cliques, factions,friendships, alliances}these coalitions exist when-ever individuals feel motivated to provide generalsupport to someone else by virtue of implicit orexplicit co-membership. What earns you the rightto the support of others is the contribution youmake to supporting others when they could cost-effectively benefit from assistance. Reciprocally,when others observe a pattern of undercontribu-tion on your part, this erodes the willingness ofothers to support you.

For example, disputes are endemic in small-scalesocial interaction, and are often decided in waysthat reflect the relative power of the disputants.Other things being equal, two individuals havemore power than one, three more than two, and soon, so support for an individual in a dispute byothers is often a considerable help in resolving thematter in a way that favors the supportedindividual. Everything can be taken from apowerless individual. So in addition to the otherjoint resources that were produced by collectiveeffort ancestrally, ordinary social competition inhunter–gatherer interactions would have selected

for a coalitional psychology. Indeed, experimentalevidence indicates that our minds contain anevolved alliance detector that scans for themembership in coalitions bound by mutual alli-ance. This evolved cognitive specialization, as aroutine part of person representation, takes thepattern of cues of mutual alliance present inindividual behavior in the social world as input,and deduces coalitional categories in the mind asoutput (Kurzban et al., 2001b). That is, our mindsare designed to interpret the social world in termsof coalitions, spontaneously and automatically.

Trade-offs arise from the opposing costs andbenefits to the individual of belonging to largerversus smaller coalitions: The larger the coalition,the stronger it is, and the more resources andsupport it can supply to an individual. However,the larger the coalition, the less joint attention andvalue the ‘group mind’ can give to an individual,and the less indispensable she is to it. Even moreimportant, the larger the coalition, the morefrequently disputes involving welfare trade-offswill occur among ingroup members rather thanbetween members and nonmembers. Wheneverdisputes are internal to the coalition, the individualcannot draw on it for support. At the logicalextreme, if everyone belongs to a single encom-passing coalition, then it is not a coalition thatusefully amplifies individual power in the disputesthat emerge in daily life. Help from the coalitioncannot be dependably asymmetric when bothdisputants happen to be in the same coalitionand so qualify for the same support. Reciprocally,the smaller the coalition, the more differentiallyvalued and attended to a member will be. More-over, the smaller the coalition, the more indivi-duals lie outside of it, and so more disputes withothers will occur between ingroup members andoutgroup members. As a result, the smaller thecoalition, the more often the coalition can be usedto amplify one individual’s power against anotherin a dispute. Consequently, within each larger scalecoalition, individuals benefit by calving off asmaller scale coalition to side with them whenthere are internal disputes}support that is earnedby siding with others in other internal disputes.(Indeed, more marginal individuals may covertlywelcome or encourage disputes among others thatcan give them opportunities to assist others andhence cement their status as someone entitled todefense and support.5) As a result, individuals areusually members of many concentric amplification

J. TOOBY ET AL.112


coalitions from dyads up to the largest scale ofintergroup conflict. Coalitions existing at differentscales are fractally organized by links of mutualalliance. This chronic set of problems suggests thatthe psychological system that carves out a coali-tional identity includes regulatory circuitry toassess the local fractal structure of groups, andto allocate effort to forging alliances across a rangeof scales. For similar reasons, each individual in anamplification coalition should spontaneously dis-approve of the formation of strong individualloyalties toward outgroup members by otherindividuals in the ingroup: Such loyalties willcontract the set of circumstances in which thecompromised individual can be expected toamplify the power of other ingroup members.6

These considerations extend beyond help indisputes. If, for example, responsibility for materialassistance is diffused among too many individuals,no one helps (leading to the well-known bystandereffect; Darley and Latane, 1968). To be cheated inthis context is to fail to receive assistance when onehas given it to other ingroup members. Cheatingamong large undifferentiated groups is more likelybecause the blame for the absence of assistance isdistributed among too many others. For assuranceof help from others to be strong, there must bebonds of mutual dependence among a small enoughnumber of individuals that withdrawal by a cheatedindividual strongly hurts a few specific others.Individuals accordingly feel a sense of precarious-ness among homogeneity, and a desire to beindividually valued by a small number of localothers (see also, Tooby and Cosmides, 1996).Indeed, Marilyn Brewer proposed that individualidentification with groups is shaped by a motivationfor group optimal distinctiveness, which may alsoserve this function (Brewer, 1991). In short, it is notenough just to be nominally a member of acoalition}a sufficient number of local individualsin the coalition must know you, differentially valueyou, monitor your welfare, and support you whenyou need it.

Both for reasons of social competition andmaterial insurance, when homogeneous coalitionsget too large, our evolved coalitional psychologiesprefer to give them internal factional structure. AsMadison pointed out in the Federalist Papers, the‘latent causes of faction are thus sown in thenature of man’ (Madison, 1788/1982). Morebroadly, modern humans feel naked without theprotection of a sufficient set of group identities

because the well-designed hunter–gatherer wasunprotected against threats from the social andnatural world without the support of others(Tooby and Cosmides, 1996). Individuals havean evolved appetite for creating coalitional iden-tities independent of and prior to a real need forthem. If individuals were designed to wait untilneed struck to build a coalition, the organizationaldifficulties and latencies would make it too late.So, to serve these appetites, modern humans inflecttheir lives with notional and recreational groupsthat often seem detached from utilitarian need,such as churches, clubs, sports teams, politicalgroups, and the like (see, e.g. Tiger, 2004).Moreover, recurrently spending time with othersin group settings is itself a primary generator of theexpectation that one is in an amplification coali-tion. This means that whatever the formal or legaltheory present, life in the workplace invitesimplicit assumptions of mutual amplification andsupport that transcends work-specific activities.Conformity to these expectations is consideredelementary decency, and their repudiation ordiscouragement is experienced as an alienatingfeature of many corporate and nonprofit work-place cultures.

The Status of Coalitions is a Public Good, Making

Coalitions Collective Actions

Amplification coalitions are intuitively under-stood, spontaneously self-organizing, and mayemerge into awareness when one or more indivi-duals are challenged by outsiders. One cruciblethat led to the evolution of the psychologyunderlying amplification coalitions was their an-cestral operation in zero-sum competitions overaccess to resources by contending sets of indivi-duals. Individuals take or relinquish resources tothe extent that they feel empowered or over-matched. Public signals of support (or its absence)lead individuals and sets of individuals to revisewhat they attempt to possess, consume, or do. Inconsequence, our species-typical psychologyevolved to represent coalitions as having status(Brase, 1998; Sidanius and Pratto, 1999). Indivi-duals (where cost-effective) expend effort todefend, advertise or enhance the status of thecoalitions they belong to, and to deflate the statusof rival groups. This is one set of individualbehaviors that humans are designed to interpret asexpressions of the group as an entity. For a given



action by individuals, this interpretive systemlicenses the transformation of He did this to me(one individual taking action with respect toanother) into They did this to us (one group takingaction with respect to another).

A central feature of interactions over groupstatus is that their product is a public good. Thatis, anyone who is a member of an amplificationcoalition automatically partakes of its status, andgets the benefits in daily transactions of the greaterweight nonmembers will place on members’interests out of respect for how members coordi-nate to support each other in conflicts. Acts ofalliance cause observers to categorize individualstogether as members of the same coalition(Kurzban et al., 2001b). The representations inthe minds of observers of the status of a coalitionare a common resource that has the propertiesdiagnostic of a public good. That makes the workof maintaining group status intrinsically a collec-tive action (Tooby and Cosmides, 2002, forth-coming). Because amplification coalitions ariseeasily and spontaneously from the mere existenceof any cooperative precedent, and individuals tendto assume their existence and continuity, it followsthat all groups will gravitate toward beingexperienced, to some extent, as collective actionsby their members. This will be especially truearound issues that affect the status of the group inthe mind of outsiders.

Amplification coalitions by their nature areassumed to continue indefinitely as ongoingoperations, so they are formally different fromcollective actions that exist to achieve a realizable,finite goal, with a specific termination point.Extended cooperative interactions invite the im-plicit assumption of the existence of an amplifica-tion coalition, binding its members into anexchange relationship. The individual is neverdone with the obligation to contribute to thegroup so long as the group persists. One can cheatby not contributing one’s share of support toothers in the coalition, including not contributingone’s share to maintaining the representation ofgroup status in the minds of nonmembers.Individuals can also cheat by assuming member-ship and signaling it to outsiders without havingearned it (Brase, 1998). This makes policing theboundaries of such a coalition a sensitive issue.Natural questions that express this intuitivecoalitional psychology include: Is individual i reallya member of the group? Is i a contributing member

of the group (fully paid up; not a free rider)? If thegroup has been in existence for some time, newmembers at the threshold of joining will intrinsi-cally strike existing members as resembling chea-ters in that they are attempting to gain the benefitsof membership, without having yet paid any costs(Tooby and Cosmides, 2002). Social practices willemerge to express or disarm the punitive sentimentas well as the distrust recruits spontaneouslyattract from veterans. A strong feature of coali-tional psychology will be organized around issuesof membership, group identity, the price of entry,initiation, genuineness of membership (loyalty,commitment), exclusivity, and the pressure onnew members to make contributions to bring theminto the implicit exchange relationship in theminds of veteran members. This exchange viewpredicts that organizations in all human cultureswill spontaneously tend to pass costlier, lessattractive tasks to younger, newer recruits, andrecruits will be motivated to sacrifice more in orderto make more dramatic and observable contribu-tions than established members will. The punitivesentiment veterans feel towards new entrants,combined with the advantages for veterans ofcreating strong representations of the dominanceof the coalition over new members, together oftenlead to the coordinated infliction of costs byveterans on new entrants. Correspondingly, whenrecent or potential entrants assume increasingcosts attendant upon membership, the motiva-tional systems of established members will increas-ingly recategorize them as potentially worthwhilerecipients of group benefits in an n-personexchange. The motivation to police the boundarywill be calibrated by the magnitude of the benefitsthat go with membership, the labor recruitmentneeds of the coalition, the levels of trust andinterdependence needed to function as a coalition,and the frequency with which the welfare of thecoalition (or its participants) depends upon sub-stantial sacrifices by members.

The Addition of the Cognitive Ability to Represent

a Group as an Individual Allows an Additional

Approach to Applying Dyadic Exchange Machinery

to n-person Exchange

Although the addition of recursion is one im-portant computational modification to 2-partyexchange psychology, a second important mod-ification to this exchange psychology involves

J. TOOBY ET AL.114


widening the exchange system’s definition of whatkinds of entities can be computationally treated asa party or agent. During the initial evolution ofexchange circuitry, party would have meantindividual human. The next step would have beenwidening party or agent so that open arguments inprocedures that formerly would have referredstrictly to human individuals become able to referalso to coalitions, alliances, communities, andother entities as well. This allows humans torepresent, understand, engage in, enforce, andcheat on dyadic exchanges where one party is anindividual, and the other party is a group, as wellas where both parties are groups.

This cognitive ability to interpret a coalition asan agent is not just speculation: Two decades ofresearch on social exchange reasoning demon-strates that subjects can understand and reasonabout 2-party exchanges just as readily when theexchange rule is a social rule entered into with agroup as they can about exchanges between twoindividuals (Cosmides, 1989; Cosmides and Too-by, 1992, 2005). Such a broadening of the meaningof agent creates alternative ways for individuals torepresent and enter into n-person exchanges. Aninteraction could be represented as a single multi-lateral n-person exchange with every individual asa separate node; or, it could be represented foreach individual as a dyadic exchange between theindividual and the group, with two nodes for thetwo agents}self and group. Mathematically, an n-person multilateral exchange involves nðn� 1Þdirectional links, while an n-person exchangeconceptualized as a series of individual to groupexchanges involves 2n links (or n exchanges).Seeing an n-person exchange in terms of a seriesof dyadic exchanges between each group memberand the group as a single entity allows for greatcognitive simplification and transparency (when nis greater than 3): Simply using pre-existing dyadicmachinery, each individual can clearly understandand reason about what she herself owes the group,and what the group owes her. She does not need tosimultaneously represent links to every individualin the group. She can also monitor cheating on thepart of other members applying dyadic reasoningparallel to what the individual applies to the self: Isa given individual taking the benefit of the jointeffort? Is the individual holding up their part? Aperson can comprehend and cheater detect an n-person exchange as a series of dyadic exchanges bygoing individual by individual through the group,

first taking the individual’s perspective, and thenthe group’s (represented as a single party). Thisallows dyadic exchange psychology to generateand cognitively treat many kinds of social rules associal contracts (Cosmides and Tooby, 1989).

The widening of the agent slot to include groupsas well as individuals is a cognitive expansion thatapplies beyond exchange relationships, althoughthat may have been its selective crucible. In humanpsychology, groups are not treated simply ascategories or aggregations of individuals. Groupsare conceptualized and experienced as entities towhich we can properly attribute mental states, andwhich can have propositional attitudes (Cosmidesand Tooby, 2000). That is, humans have notrouble interpreting claims about the mental statesof groups as if a group constituted an individualagent with a single mind. (In reality, obviously,they are not.) For example: Our university ordepartment or company can want us to do things;our community can be angry, or disgusted orpleased with us; our nation can be intimidated orvindictive; our village can mistakenly believe wekilled someone; we can be afraid to disappoint ourgroup. Nelson signaled at Trafalgar, ‘Englandexpects that every man will do his duty’ withoutwondering whether his fleet would be philosophi-cally puzzled about how a nation can entertain anexpectation.

Moreover, humans also unproblematically treatgroups according to an intuitive theory of intereststhat we believe originally evolved to interpretindividual action. Humans not only representindividuals as having interests, but typicallyrepresent groups has having welfares or interestsas well. Accordingly, groups (like individuals) areexperienced as having a unitary exchange accountthat we can trade off our welfare against (and viceversa). In so doing, we can owe the group, supplyhelp to the group so that it becomes indebted to us,feel entitled to group assistance, or punish thegroup for not taking our welfare sufficiently intoaccount; reciprocally, the group can owe us,supply help to us so that we become indebted toit; it can feel entitled to our assistance, or punish usfor not taking its welfare sufficiently into account.Groups can have status, rank, stigma, anddominance relations, not to mention friendshipsand enmities. We can see them as exploitive(favoring their own welfare above ours) orgenerous (injuring their own welfare to benefitus). A group can be dominant over us, whether we



are a member or not. (Indeed, leading members ofmale fraternal organizations often want newmembers to appreciate the power the coalitionhas over them; Tiger, 2004). Many cognitivescientists are persuaded by the developmental,cognitive, and neural evidence that humans comeequipped with what has been called a theory ofmind, a specialized set of mechanisms that inter-prets individual behavior in terms of mental states(Baron-Cohen et al., 1985). We propose that a setof augmentations and modifications to the theoryof mind system constitute a theory of group mindsystem (Tooby and Cosmides, 2002, forthcoming).This system generates representations about whata group thinks and feels, triggers the individual’sfeelings with respect to imagined mental states ofgroups, and treats groups as entities that can bethe object of emotions and motivations thatinitially evolved for individuals.

In reality, of course, groups (as sets of indivi-duals) do not have a single mind, body, or voice.In order to treat groups as agents, the mind mustcontain procedures that can interpret the obser-vable acts of individuals as the behavior of groupsor the expression of group mental states andintentions. If groups are mentally represented asamplification coalitions, then this predicts that onemajor set of acts of individuals that invitereinterpretation as acts of the group are behaviorsby one or more individuals in one coalition thatnegatively impact the welfare of one or moreindividuals in another coalition. When members ofthe Los Angeles African–American communitysaw footage of several Los Angeles police officersclubbing Rodney King, they did not treat it as anaffair just between the interactants}it was seen asone community (coalitional entity) doing some-thing to another community (coalitional entity). Inaccepting or attacking the acts of individuals fromother coalitions, coalitions act as if they arenegotiating general precedents specifying howmuch members of each coalition are obliged totrade off their welfare on behalf of members of theother coalition (an equilibrium welfare trade-offratio expressing relative power or dominance).Some individual acts}outrages}serve as coordi-native signals for a group to act together in acollective action to obtain what is usually a publicgood: to act together to change the representationsin the minds of the other group as to the relativepower of the two groups, by threatening, attack-ing, or dispossessing them.

How humans represent the organizations theywork within will have a major impact on how theybehave in these organizations. Consequently, it ishard to overestimate the magnitude of the impactthat the ability to represent groups as individualshas had on human sociality. This key expansion inthe human ability to engage in n-person exchangepopulates our mental world and our social worldwith groups, using much of the same machinerythat governs how we think and feel aboutindividuals. Yet, we maintain in parallel the abilityto interpret groups as sets of individuals engagedin multilateral exchange. These alternative ways ofrepresenting n-person exchanges are different, andmobilize different ways of thinking and feeling: Inone case, I am in a multilateral exchange witheveryone else in the group. In the other, I am in adyadic exchange relationship between myself andthe group. It is possible for the same individual toactivate both of these representational modalitiestoward the same object, at least at different times.A corporation might for some purposes berepresented by the employee as a single agent(the company) in a dyadic exchange relationshipwith the employee, and on other occasions as ateam or amplification coalition (a multilateralexchange) that the employee is a member of.These alternative representational methods willactivate different motivational responses.

It is perhaps not an overstatement to say thatcorporations (and many other formal organiza-tions) as human social institutions are founded onthe evolved representational ability to represent agroup (i.e. a corporation) as a unified agent orperson, while collectives, associations, and unionsmore often invite representation as multilateralexchanges among (relatively) equal members. If Iinterpret my employment with a corporation asdyadic relationship, then my attention is focusedon allocational conflict over benefits and costs (e.g.salary, effort, risk) between the two parties, myselfand the corporation. Ordinarily, in dyadic ex-change, the marginal cost of moving a unit ofbenefit from one individual to the other iscompared, with some modulated expectation thatunits will go to the individual to whom it does themost good. If the corporation is viewed as anindividual, however, it strikes the employee as awealthy, dominant, and potentially exploitiveindividual}something not significantly harmedby transferring marginal units of benefit thatwould make a great deal of difference to the

J. TOOBY ET AL.116


employee. The fact that the corporation does notrespond according to this logic is a continuoussource of alienation for employees. This modalityof representation provides the foundation for theMarxian and labor union worldviews that frameworkers and management as inherently locked inan oppositional, zero sum relationship with eachother. Alternatively, if the individuals networkedtogether into a corporation represent themselvesas members of a team, then the invited motiva-tional consequences are a convergence of interests.In either case, workplaces usually bring togetherbounded sets of individuals in continuous associa-tion with each other, often facing an open-endedset of situations. Ancestrally, spending a majorfraction of each day together indicated mutualmembership in an amplification coalition, impli-citly obliging members to provide a form ofcooperative social insurance to each other. Theexpectations generated by different forms oforganization are not just provided by explicit rulesor by local cultural practices. Our evolvedmechanisms, triggered by experience, assign theirown system of evolved meanings to social interac-tions even in defiance of explicit rules or attemptsat contrary cultural indoctrination.

Anti-exploitation Motivational Circuitry in

n-person Exchange

Neither dyadic nor n-person exchange can evolveor be evolutionarily stable unless cheating isdetected and then effectively responded to. Nu-merous experiments have established the firstelement: that the human brain does indeed containevolved circuitry specialized for detecting cheatingin both dyadic and n-person exchange relation-ships (Cosmides, 1989; Cosmides and Tooby,1989, 1992). To function properly, however,cheater detection circuitry needs to be coupled toa component that motivates effective responses tocheating. An effective response is one that helps tomake participation in exchange fitter than eithercheating or nonparticipation (as well as fitter thanother, more complex uncooperative strategies).The key property that these motivational circuitsmust have is the effect of making the averagelifetime fitness of those equipped with cooperativecircuits, on balance, greater than the averagelifetime fitness of those with a stronger dispositionto free ride. Given this design criterion, what

should the outputs of this motivational componentlook like?

The problem of sustaining cooperation in theface of potential exploitation is severe andubiquitous: After all, the ordinary psychology ofeffort minimization should make everyone atactical free rider. In all conditions, includingexchange, motivational adaptations in humansshould have been selected to identify and curtailunnecessary investment (i.e. units of effort whoseinvestment does not trigger or sustain returns).Hence, our evolved cognitive equipment shouldconnect the motivation to expend effort in anexchange to an assessment of the sensitivity ofothers’ responses to one’s own contributions, bothupwards and downwards. That is, do they increasetheir contributions when I increase mine? Do theydownregulate their contributions when I down-regulate mine? Are they appreciative if I sacrificefor the joint effort? Do they even notice if I do not?

As a result, the motivational component shouldhave been designed by evolution to modulate one’sown level of compliance to one’s obligations in anexchange in response to others’ levels of compli-ance (or at least to others’ monitoring andeventual responsiveness). In dyadic exchange, Iam no longer motivated to live up to my part ofthe exchange when my partner is not living up tohers. In n-person exchange, motivational levelsshould be a function of the distribution ofcontributions by others. In general, such circuitrymotivates greater effort when others are investingsufficiently (and contingently), and curtailed effortin the presence of free riding (or indiscriminateinvesting). This set of conditional rules regulatingthe motivation to participate is an importantfirst line of defense against being exploited byothers (although it permits the conditional exploi-tation of others).

Accordingly, our evolved motivational systemshould be designed to recognize situations whereothers systematically benefit from one’s ownsacrifices without contributing proportionately.Such situations should be cognitively representedas distinct from other situations, and marked asintrinsically motivationally significant. The detec-tion of such a situation (or its possibility) shouldthen trigger motivational outputs designed toeffectively redress the potentially adverse fitnessordering manifested in an exploitive situation (i.e.some property of the cooperative strategy mustinsure that on average, noncooperative exploitive



strategies have lower fitness, at least when theybecome common). Hence, procedures for socialcomparison of effort and benefit should be veryimportant motivationally. Of course, the under-lying logic of exchange involves delivering benefitsonly when the cost of providing them is less thanthe benefit produced}no one expects you to cutoff your arm to make dinner for a friend. Giventhe wide range of cooperative endeavors humansengage in, this means that cooperators will usuallyneed to represent not just behavior but also theunderlying costs and benefits that are associatedwith behaviors. If your child is starving, I do notexpect you to give me food, even if I previously fedyou. We operate in a world where the inferredcosts and benefits of actions are relevant to ourunderstanding of our exchange relations.

Accordingly, to engage in n-party exchangebehavior, the system needs to represent theaccounts own welfare and group welfare in orderto make decisions governing one’s own allocationsbetween these two accounts. To do this, the mindneeds to compute a regulatory variable, the welfaretrade-off ratio (WTR), that governs what trade-offs (sacrifices) the individual will be willing toengage in that have a negative impact on ownwelfare and a positive effect on group welfare (andvice versa). To protect itself against exploitation,the system also needs to be able to monitor andrepresent events and actions in terms of theirsignificance with respect to others’ welfare ac-counts (e.g. the contributions others make, thecosts they incur in making contributions, thebenefits they derive from the exchange, the impactsI have on them, etc.). These variables need to betracked both for individuals, and (for somepurposes) on a pooled basis. This enables themotivational system to be able to compare one’sown welfare trade-off ratio to the welfare trade-offratio that others are using, and therefore todetermine whether the actor is being exploited byothers (or is exploiting others). Our motivationalsystems switch into different modes of activationdepending on the answer to questions like: Areothers contributing less (more) than I am? (e.g. Priceet al., 2002; Price, in press B).

The evolved aversiveness of low payoffs ininteractions should be a function of comparativepayoff, not just absolute payoff. In rational choicetheory, any payoff is better than no payoff, but inthe evolutionary competition of alternative de-signs, a positive absolute payoff that is a low

relative payoff should (under many conditions) beselected to be perceived as worse than a zeroabsolute payoff that is equally bad for bothparticipants. Widespread evidence from experi-mental economics suggests that this is the case,such as the high level of rejection of offers in theultimatum game (Hoffman et al., 1998; Henrichand Smith, 2004). Indeed, the sucker’s payoff (inprisoner’s dilemma situations) should not just beaversive because it is low paying in an absolutesense, but because it is relatively low payingcompared to the gains of the other interactant.So, algorithms in the brain characterize situationsin terms of whether others are benefiting from theindividual’s own sacrifices in a way that isdisproportionate to their contributions. For con-venience, we will call this an evolved characteriza-tion of the degree of personal fairness or personalexploitation exhibited by a situation. As historyand evolutionary theory both attest, evolvedmechanisms in the mind do not have an automaticappetite for global fairness of all to all (includingsacrificing personal gains to be fair to others). Ourevolved mechanisms are not designed to intrinsi-cally object to one’s own unfairness to third parties(although in the right social environment suchattitudes may emerge). Our evolved exchangepsychology is designed instead to manifest adistaste for exploitation or unfairness by othersto oneself, as well as by others to family members,to a friend or ally, to an ingroup member, or(through putting a group into the agent slot) topersonally valued groups (see, Fehr and Fischba-cher, 2003, for a different view).

There are several distinct motivational outputsfrom this system, when it recognizes a situation ofpersonal exploitation, in addition to contributiondownregulation. The balance of costs and benefitsarising from the situation the individual is indetermine which motivational output and beha-vioral response will dominate. Most fundamen-tally, the exploitive aspect of a situation should beexperienced and represented as intrinsically un-pleasant or aversive, independent of the otherpayoffs of the situation, so that the prospect ofremoving exploitation presents itself as a goal (ifachievable) with a positive payoff. Not only shouldexploitation be aversive, but individuals whoconsistently exploit more than others shouldthemselves come to be seen as aversive. Othermotivated responses that defend against adverseselection favoring free riders include: punishing

J. TOOBY ET AL.118


free riders (see, e.g. Price et al., 2002); avoidingindividuals who differentially undercontribute;leaving situations in which one is being exploited;refusing to initiate exchanges in which exploitationis likely to occur; refusing to continue in exchangesin which exploitation is occurring; aggressivelyforcing those disposed to free riding to contribute;and driving off those disposed to free riding. It isunparsimonious to assume that all of theseresponses were independently selected for. Rather,it seems plausible that many or all these responsesare generated by a single underlying motivationalelement: finding personal exploitation intrinsicallyaversive.

Opening Cooperative Moves and Subsequent

Monitoring Forms in n-person Exchange

Because mutual benefit through coordinatedn-party exchange is hard to achieve and easy toundermine, joint contribution levels will often befar lower than optimum. The recurrent opportu-nity to capture these underrealized benefitsselected for design elements that promote upwardmovement. Perhaps the most important of theseincluded an initial cooperative orientation parallelto tit-for-tat’s opening cooperative move (Toobyand Cosmides, 2000). This orientation involves areadiness to make the first move (at least where thenumber of parties is not too large). This poten-tially initiates new upward movements at thebeginning of any new event boundary thatplausibly invites new collective projects, coupledto a decision-rule to modulate downwards if theeffort goes unmatched. Additionally, we recognizeand respond to others’ supernormative initiatives(we can be ‘inspired’ by others, a complement toour being cooled by others’ hypocrisy or defec-tion). We are proud of our own supernormativeefforts, especially if they successfully invite othersto new levels of commitment, and shamed if ourefforts are discovered to be lower than others, andto have inhibited positive interactions. As ageneral rule, the motivational system should bemore willing to make an investment if it is public;investments should preferentially be made in acontinuous flow of consecutive increments (wherethis is not inconsistent with public delivery), sothat their magnitude can be modulated contingenton others’ degree of matching}a pattern thatreduces the opportunity for free riding. Oncea higher contribution level is made by one

participant, others must match it, or their failureto do so will establish the future collective ceilingbelow the optimum.

So, our evolved exchange psychology containsprocedures that motivate participation in jointefforts when they can be detected and assessed asbeneficial. Defense against exploitation requiresmonitoring of others’ actions and their contingen-cies. We believe that different structures ofmonitoring have selected for different modes ofn-party exchange. These include joint monitoring(where everyone in a collective enterprise monitorseveryone else), leadership monitoring (where singleindividuals}‘leaders’}monitor and direct enfor-cement), asymmetric monitoring (where multipleindividuals with disproportionate stakes in thejoint effort or lower costs of enforcement regulatethe effort), role monitoring (where specialistscultivate a socially valued identity based onmonitoring), or some mixture of them. Jointmonitoring was perhaps the first to evolve, butdepends on the practical opportunity for small setsof individuals to directly observe each other’slevels of participation. All of these organizationalalternatives are anchored in an evolved psychologyof social comparison that evolved in the context ofn-party exchange.

Indeed, a standard finding of experimentalcollective actions (public good games), is that themajority of subjects are conditional cooperators,i.e. they cooperate more when they perceive thatco-players are more willing to cooperate (Orbelland Dawes, 1991, 1993; Ledyard, 1995; Fischba-cher et al., 2001; Lubell and Scholz, 2001), and lesswhen they believe that co-players are free riding(Fehr and Gachter, 2000; Kurzban et al., 2001a).Moreover, a standard finding of anonymouspublic good games, in which players have noinformation about co-player cooperativeness atthe start of the game, is that cooperative behaviorfollows the pattern predicted from an evolutionaryperspective. Contributing is highest at the outset ofthe game (players open with invitations tocooperate), and then decays gradually over time,as players receive information that some co-players are free riding (Ledyard, 1995; Mascletet al., 2003). This gradual decay apparently occursbecause higher contributors in initial roundsratchet down their contributions as the gameprogresses, in order to avoid exploitation bymatching the average expected co-player contribu-tion. The average constantly dwindles due to



persistent free riding (Fischbacher et al., 2001;Kurzban et al., 2001a,b).

Further, as expected if cooperators are siftingfor exploitation, interactants appear to monitorone another not just constantly, but accurately.Among villagers in a hunter–horticultural society,perceptions of co-villagers’ engagement in generalpro-village altruism correlated positively withmeasures of co-villagers’ actual engagement inspecific altruistic activities (Price, 2003). In asugarcane cultivating workgroup in this samesociety, perceptions of co-worker cooperativeness(attendance record and physical effort in worksessions) correlated positively with more objectivemeasures of this cooperativeness (Price, in pressA). Moreover, participants accurately distin-guished ‘intentional’ low contributors (those whocould have contributed highly but chose not to)from ‘unintentional’ low contributors (those un-able to contribute highly).

A psychology of conditional punishment co-evolvedwith a psychology of n-person exchange. Theevolved motivational mechanisms regulating n-person exchange are designed to make eachindividual’s actions contingent on the actions ofthe others in the cooperative interaction. Conse-quently, these mechanisms endow joint efforts withdynamical properties. That is, an n-person ex-change is a sensitive network of intercontingentfeedback loops. Depending on variables such asthe opportunities for free riding and the frequencyof free riders, an intercontingent cooperativestructure can dynamically drive itself towardhigher levels of cooperation or downwards towardlower levels of cooperation or dissolution (Ehrhartand Keser, 1999; Fehr and Gachter, 2000; Mascletet al., 2003). For example, if many participants arecommitted and enthusiastic, then the probabilitythat others will join and contribute is increased. Incontrast, the presence of undercontributors (chea-ters, free riders) in the network inhibits contribu-tions by others. Individuals are more reluctant tojoin collective efforts in which they can anticipatethat others will be benefiting without contributing.Indeed, the presence of individuals with a trackrecord of greater free riding should prevent manyn-person exchanges from coming into existence inthe first place. At a minimum, the presence of freeriders degrades the performance of the network atrealizing potential joint gains. More seriously,sufficient penetration by free riding can trigger the

termination of an otherwise mutually beneficialongoing intercontingent effort. Consequently, un-less maintained by corrective social feedback,cheating or undercontribution will spread conta-giously, and contributions to joint efforts willratchet downwards (Kurzban et al., 2001a). Whenopportunities for monitoring and feedback areinadequate, effort minimization tends to operateunopposed, and the cooperative network erodesaway.

The key point is that free riding is not just aninefficiency, an injustice, or a marginal strategy.How free riding is treated is the central determi-nant of the survival and health of cooperativeorganizations. Given the existence of motivationalcircuitry that evolved to defend against exploita-tion, social penetration by free riders is a brake oreven an off switch, turning cooperation down oroff among individuals whose motivation to co-operate is dynamically interlinked. The social costof free riding is not just the marginal effort thatfree riders do not contribute nor is it primarily thebenefits they consume but do not earn: Thegreatest cost that free riders inflict is the loss ofall the potential gains from n-party exchanges thatotherwise would have been achieved if free ridinghad not triggered antiexploitation motivationaldefenses among cooperators. That is, by forestallingthe emergence or continuation of a range ofmutually beneficial n-person exchanges that wouldotherwise exist, free riders can and do inflict hugecosts on cooperators out of all proportion to theparasitic benefits they derive.

One can approach the same conclusion byconsidering the likely sequence of steps in theevolution of the psychology of cooperation: Ingeneral, any strategy for exchange that reliablyallows cheaters to outcompete cooperators cannotevolutionarily persist. But (under reasonable con-ditions) cheaters cannot outcompete those coop-erators designed to follow a strict conditionalexchange strategy involving the detection ofcheaters and the refusal to enter into exchangeswith them. Plausibly, the establishment of thisfirst-order exchange strategy as part of ourancestral species-typical design was thus an earlyphase in the evolution of our psychology ofcooperation (Cosmides and Tooby, 1989). More-over, this conditional strategy is reasonablyefficient for dyadic exchange. Individuals whoabandon or refuse dyadic exchanges with cheatersare only forgoing interactions in which they are

J. TOOBY ET AL.120


unlikely to get benefits anyway. However, asn increases for n-party exchanges, abandonmentof the exchange in the presence of a cheaterbecomes increasingly costly and inefficient in termsof forgone benefits. Under strictly conditionalcooperation, no matter how many cooperatorsare present, a set of interacting cooperators can beprevented from beneficial interactions by thepresence of a single cheater. Naturally, the greaterthe number of individuals involved, the more likelyone (or more) will be a cheater, preventing theexchange (assuming an early ancestral socialecology in which individuals are equipped onlywith a psychology of strict exchange). Such astrategy insures that only exchanges with verysmall numbers of individuals will ever take place,leaving the rest as a large, unachieved residualclass. Of course, despite its inability to capturegains from larger scale exchanges, a strategy ofstrictly conditional exchange would neverthelesshave been strongly favored by selection because itdid allow capturing gains from smaller scaleexchanges. Indeed, for our hunter–gatherer ances-tors, a substantial number of exchanges were verysmall-scale}dyads, triads, tetrads, and so on.Even now, in modern mass society, the greatmajority of daily cooperative interactions involveonly a handful of individuals.

In short, a strictly conditional exchange strategywould have evolutionarily emerged and been pro-pelled upward toward species-typicality among ourancestors because (1) it cannot be outcompeted byan exploitive strategy, (2) it is able to take advantageof any n-person exchange not involving a free rider,and (3) opportunities for such exchanges wereancestrally ubiquitous. This makes this strategy fitterthan strategies that do not engage in exchanges, orthat free ride (cheat). We think that these selectionpressures established something resembling a strictlyconditional psychology of exchange among ourancestors as a platform out of which a morecomplex psychology of cooperation subsequentlyevolved. Accordingly, we should not expect to see afirst-order cooperative psychology whose motivatedcontributions are unaffected by the presence of freeriders, because such mutant designs would have beenexploited and outcompeted. What we predict onselectionist grounds is what we observe empirically:The presence of free riders downregulates coopera-tive behavior.

Because the detection of free riders by coopera-tors inhibits contribution to joint efforts, free

riders stand in the way of reaching valuable gainsthrough cooperation. The first line of defense forconditional cooperators}abandoning the ex-change in the presence of free riders}is very costlyto conditional cooperators. As a result, the presenceof free riders in cooperative contexts was ancestrally(as it is now) a serious adaptive problem facingconditional cooperators. If they exist, selectionwould have strongly favored the subsequent evolu-tionary emergence of psychological mechanisms that(1) can realize gains from n-person exchanges even inthe presence of cheaters, while simultaneously (2)preventing cheaters from outcompeting cooperators.Are there any such designs?

An evolutionarily tailored motivational systemdirecting punitive sentiment can help to solve theadaptive problem posed by free riders (Price et al.,2002). If the presence of a small number of freeriders (or one) is inhibiting optimal contributionlevels to an n-person exchange, then punishmentdirected at the individual will drive her off, orinduce her to contribute. In either case, thesituation is no longer psychologically marked forcooperators by personal exploitation, and so theyare no longer inhibited from contributing effec-tively to the exchange. Equally, in selectionistterms, free riders are no longer outcompetingcontributors. That is, it is the dynamic sensitivityof n-person exchanges to the presence of exploita-tion that selects for punishment. Rather thaninefficiently removing the exploitation by depriv-ing a number of potential cooperators of thebenefits of n-person exchange, punishment can beefficiently targeted exactly at the source of theproblem, the exploitive individuals themselves.Hence, in an ancestral social environment ofpotential n-person exchanges, the evolution ofmechanisms directing punitive sentiments towardfree riders was favored by natural selection.Because punitive acts and the evolved dispositionsthat cause them unlocked access to a wide range ofpreviously unattained cooperative gains, the cir-cuitry underlying this punitive psychology wasadvantageous. These fitness advantages increasedits frequency until it became (we believe) asignificant part of our species-typical social psy-chology of cooperation. Widespread experimentalevidence indicates both that humans do havepunitive sentiments toward exploiters in n-personexchanges, and that n-person exchanges can bedriven to higher levels of cooperation whensubjects have the ability to punish (Fehr and



Gachter, 2000). Although the strategy of avoidingcheaters can work as the sole defense againstexploitation when exchange interactions are small(dyads, triads, tetrads, etc.), the larger the size ofthe interaction, the more efficient punitive co-operation strategies become at sustaining coopera-tive organization.

The design features and evolutionary dynamics of apsychology of punitively defended conditional co-operation. For the reasons outlined above, wethink that the human mind contains an evolved,functionally specialized motivational mechanismthat, when exposed to a situation of personalexploitation, generates a punitive sentiment to-ward the agent that is deriving an unfair advantagein an exchange. We think evidence shows that thismechanism becomes strongly activated in collec-tive actions, and evolved as an anti-free riderdevice (Price et al., 2002, in preparation; Price, inpress B; Tooby and Cosmides, 2002, forthcoming).In n-person exchanges, this sentiment motivatesindividuals to inflict a cost on free riders (ex-ploiters) sufficient to reorder the net benefits thatthe interactants derive from the exchange (whenthis can be done cost-effectively). That is, in theabsence of corrective action, free riders are betteroff than cooperators. After corrective (punitive)action, free riders are made worse off thancontributors. The functional product of punish-ment is an outcome in which the most exploitiveindividuals either end up with a lower welfare thancontributors, or end up less well off than theywould have been if they had not attempted toderive benefits from the collective effort. Thepresence of punitive strategists (punitive coopera-tors) changes the payoff structure for free riders,influencing their choices (to the extent theyrespond to anticipated payoffs). When disposi-tional free riders (individuals psychologicallydesigned to not contribute) detect that one ormore punitive individuals are to be in an n-personexchange interaction, they should avoid the inter-action}which will be low paying and henceaversive for them. When facultative (tactical)free riders detect interactions involving punitivestrategists, they should avoid the interactionor should tactically contribute. At present, theweight of evidence appears to support the viewthat the primary function of punitive sentiment isas an anti-exploitation defense (Price et al., 2002;Price, in press A, B). Nevertheless, more work

will be required to comprehensively rule out thehypothesis that punitive sentiment was secondarilydesigned by selection to serve as a system forrecruiting additional labor into n-personexchanges.

What benefits do punitive strategists derive fromtheir evolved motivational design and the expen-ditures of effort it causes them to make? Whywould a cooperator equipped with punitivemotivation toward exploiters be fitter than co-operators without punitive motivation? Punitivestrategists switch on the productive possibilities ofthe groups they are in, unleashing collective effortsthat would otherwise be inhibited by the presenceof free riders. This happens because the presenceof punitive strategists in potential exchange inter-actions repels free riders, causing them either toavoid such interactions or to become (faculta-tively, in the presence of punitive strategists)behavioral cooperators. Because free riders avoidpunitive strategists, punitive strategists will farmore often find themselves in groups without freeriders. Punitive strategists, unlike those who aresimply strictly conditional cooperators, do nothave to forgo large numbers of opportunities toharvest a valuable joint gain. Even more impor-tant, the advantage of a punitive strategy appliesat small scales even when it is rare. For example, itwill be less often cheated even in dyads, becausethe cheating triggers punishment, and so thetemptation to defect is lowered by the anticipationof punishment7. Punitive strategists can evenafford to be more trusting, and can more success-fully enter into one-shot dyadic exchanges (pro-vided their interactants can anticipate that theirpartners will have a subsequent opportunity topunish). In sum, anti-free rider punitive strategiesunlock access to a wide range of n-personexchange interactions that would otherwise havegone unrealized.

A number of scholars, such as Boyd andRicherson, argue that punishment emergedthrough cultural or biological group selection oras a product of gene–culture coevolutionarydynamics}i.e. that punishment of norm violationis altruistic in the biological sense (Boyd andRicherson, 1992; Boyd et al., 2003; Henrich, 2004).The central argument is that mutant punitivecooperators and nonpunitive cooperators wouldshare the same group-wide benefits of punishment,but that only the punitive would bear the costsof inflicting punishment. So, on this view,

J. TOOBY ET AL.122


nonpunitive cooperators would outreproduce pu-nitive strategists in the same group in the same waythat free riders displace indiscriminate coopera-tors. The punishment of free riders is viewed as acollective action problem (Yamagishi, 1986), andthe evolution of punishment is considered to sufferfrom a second-order free rider problem. Therefore,according to this line of thinking, motivationalmechanisms designed to punish free riders cannotevolve without group selection or gene–culturecoevolutionary dynamics (Boyd and Richerson1992; Boyd et al., 2003; Henrich, 2004; see alsoPanchanathan and Boyd, 2004).

The objection that punishment in the context ofcollective actions suffers from a second-order freerider problem is an important argument, worthy ofserious attention. However, the validity of thisargument depends on the nature of the ancestraldistribution of opportunities for cooperation, onother pre-existing components in our evolvedsocial psychology, and on the exact nature of thecomputational procedures that implement thepunitive psychology. We think, for example, thatit is highly likely that the ancestral frequency ofopportunities for exchange was a declining func-tion of the number of individuals involved. That is,there were ubiquitous opportunities for dyads,frequent opportunities for triads, fewer for tetrads,even fewer for pentads, and so on. In normalhunter–gatherer social ecologies, twenty peoplerarely shared joint attention on a common project,while ten people sometimes did, while five peopleoften did, and two people commonly did. In such asize-skewed social ecology, selection for a punitivestrategy could easily proceed (1) when the benefitto punitive strategists of increased gains fromparticipating in n-person exchanges sufficientlyexceeds the costs of punishment, and (2) whenthe costs of punishment for punitive strategists aresufficiently less in their successful exchange inter-actions than the costs nonpunitive cooperatorsincur from more often finding themselves in failedexchange interactions (Tooby and Cosmides,2002). Nonpunitive cooperators will frequentlyfind themselves in exchange interactions that failedbecause there were no punitive strategists in theinteraction to defend against free riders and theundercooperation they trigger. Punitive coopera-tors, of course, always find themselves in interac-tions that include a punitive cooperator, whilenonpunitive cooperators only sometimes do. If allexchange interactions involved very large numbers

of individuals, then mutant punitive strategistswould indeed have a hard time outcompetingnonpunitive cooperators, because both would getthe benefits while only the punitive would pay thecosts. But because ordinary sociality consists ofnumerous daily interactions composed of dyads,triads, and so on, nonpunitive individuals willcommonly find themselves in potential exchangeinteractions without the punitive individuals, evenwhen the relative frequency of punitive individualsis high. For example, whenever a nonpunitivecooperator is in a dyad with a free rider, they willnot be in a dyad with a punitive strategist.Nonpunitive cooperators will commonly betrapped in unproductive triads without a punitivestrategist and with at least one free rider, untilpunitive strategists become very common.

In short, the existence of numerous small-scaleexchange interactions which fail without punitivecooperators but succeed with them can drive astrategy of punitive cooperation to high frequen-cies just by individual selection (in the ordinarysense8). Punitive cooperators outcompete bothnonpunitive cooperators and free riders becausethey far more commonly find themselves inpotential cooperative interactions without freeriders and their inhibiting effects. Consequently,punitive cooperators are successful at realizing afar broader range of gains from n-party exchangesthan are rival strategies. They are not vulnerable tothe abiding weakness of a strictly conditionalexchange strategy, because interactive contextsthat initially include one or more free riderscannot inhibit them from harvesting valuable jointgains. The selection pressure created by free riderspersists across evolutionary time because ourevolved psychology of effort minimization makesa reversion toward free riding a recurrent phenom-enon in the absence of corrective social feedbackprovided by punitive strategists. We think, there-fore, that this history of selection has constructed asophisticated cooperative psychology that includespunitive motivation against exploiters.

Of course, our evolved punitive psychologycannot be indifferent to the costs of inflictingpunishment, or to the efficiency with which effortdevoted to punishment affects the landscape ofexploitation. Our evolved punitive psychologyshould be designed to be sensitive to both ofthese variables. Moreover, the efficiency of punish-ment will be affected by how many others inthe exchange will participate in the punitive



enforcement of the exchange. The greater thenumber of individuals that collaborate in punish-ment, the smaller the cost of punishment for eachindividual, and the more efficient punishment willbe. Therefore, our punitive circuits should also bedesigned to invite others to participate in punitiveenforcement, by preceding punishment with ex-posures of exploitation, expressed disapproval oroutrage, solicitations of social support to respondto exploitation, and monitoring for a sufficientmutuality of sentiment to make punitive actioncost-effective.9 Social life is riddled with confiden-tial complaints about the misbehavior of thirdparties, as individuals sound each other out aboutshared views, and implicitly about various poten-tial collective punitive enterprises. Moreover, thesize of the exchange interaction will be animportant variable impacting the cost-effectivenessof individual punitive action. A single blindlyactivated punitive strategist in a large groupcontaining many free riders is unlikely to beeffective, and therefore would be selected against.Therefore, early in our social evolution, whenpunitive strategists were rare, such a strategywould have only been favored to the extent thatcircuitry was designed to activate punishment onlywhere individual action would be efficacious}thatis, in groups that were sufficiently small. However,as punitive strategists became more common, thecosts of punishing would have been shared amongthem, allowing the threshold of activation to berelaxed so that punishment was deployed in largerand larger groups. Still, other things being equal,punitive sentiment should be stronger (i.e. theamount of cost an individual is willing to incur inorder to punish should be larger) in smallerexchange interactions than in larger exchangeinteractions. Equally, the greater the number ofindividuals who collaborate in punitive enforce-ment, the more readily individuals will becomepunitively activated, other things being equal.

In short, our evolved punitively cooperativepsychology should be designed to be sensitive to anumber of variables related to its function,including: the costs of punishment to the punitivestrategist; the number of individuals who willshare the costs of punishment; the number of freeriders requiring punishment; how cost-effectivepunishment will be in eliminating situations ofexploitation (including how effective punishmentwill be in reducing the excess benefits accruing tofree riders); and the communicative efficiency with

which coordination can be achieved in collabora-tive punishment.

Although our psychology of punitive coopera-tion evolved in the context of small-scale interac-tions, and derives much of its fitness advantagefrom the enduring pervasiveness of small-scalesocial interaction, the cultural emergence of rarer,larger scale cooperative contexts need not haveselected against punitive cooperation designs. Onereason arises from the fact that individuals aredesigned to be selective about who they encourageto fill the limited social niches in their lives, such asfriends, cooperators, allies, and exchange partners(see, e.g. Tooby and Cosmides, 1996). Also, it islikely that the tendency to free ride or cheat insome contexts is predictive of the tendency to beexploitive in other contexts (for many reasons,such as the existence of stable individual differ-ences in impulse control). This means thatobservations of how others behave in large-scaleexchanges will provide some information abouthow they will act in small-scale exchanges. Publicdereliction in large-scale exchanges should nega-tively influence the willingness of others to includehighly delinquent individuals in advantageous,smaller scale exchanges (see Panchanathan andBoyd, 2004, for a related argument).

Asymmetric beneficiaries, the evolution of leader-ship, and the emergence of morality. There is aneven more important dimension of n-personexchanges that makes the evolution of punitivestrategies by individual selection nonproblematic.The economic theory of public goods and collec-tive action is based on assuming an idealization inwhich the benefits that individuals receive fromcollective action are exactly equal. For a broadarray of ancestral activities, this assumption wasvery unlikely to have held. Individual differences inkin arrays, status, alliances, reproductive value,strength, health, and other circumstances wouldcommonly have created a spectrum of benefitmagnitudes that differed from individual toindividual for any given joint enterprise. In sucha world, there would be no barrier to the evolutionof a conditional punitive cooperator strategy thatoperated according to the following rule: Initiateand enforce an n-person exchange, using punitivethreats against potential exploiters, whenever thecost of inflicting punishment will be less than theexcess benefit the punitive strategist derivesthan what nonpunitive cooperators derive

J. TOOBY ET AL.124


(Tooby and Cosmides, 2002, forthcoming; Priceet al., in preperation). The punitive strategist iscompensated for her excess expenditure of punitiveeffort by the fact that she derives a greater benefitfrom the interaction than the nonpunitive coop-erators do. Rationally or evolutionarily, theywould not be designed to resist the punitivestrategist to the extent that the punitive strategistselects n-person exchanges that are in theirinterest, too. In short, by limiting the expenditureof punitive efforts only to those common enter-prises that the punitive strategist has a strongerrelative interest in, the punitive strategy cannot beoutcompeted by a strategy of second-order freeriding. Such a conditional punitive cooperatorstrategy reliably gains net benefits in socialinteraction that alternative strategies such asnoncooperation, strict conditional cooperation,and dispositional free-riding do not. In sum, weconclude that a psychology of punishment inexchange became part of our species-typical designover evolutionary time, although it is expected tobe differentially activated in different individualsbecause of different individual circumstances anddifferent developmental trajectories.

What does the hypothesized presence of such apsychology predict? Those contributors to an n-person exchange who stand to gain the most fromits success will be differentially involved inorganizing and expending coordinative effort:punishment, encouragement, persuasion, and soon. Equally, those for whom the cost of inflictingpunishment is relatively less (i.e. more powerful orformidable individuals) will also be more likely tofind themselves in the envelope of conditionswhere punishment is cost-effective. These providesome of the kernels out of which a theory of thenature of leadership can be developed.

As discussed, leadership spontaneously emergesfrom a differential ability to solve the coordinationproblems involved in n-person exchanges, includ-ing the threat of dissolution through individualschoosing to undercontribute. Consequently, oneelement of leadership is the ability to effectivelydeploy punishment to prevent potential partici-pants from defecting on a joint cooperativeventure. To the extent that an individual is moreformidable (i.e. for whatever reason, can inflictnecessary punishment at lower costs), she is betterpositioned to become a leader. To the extent anindividual is a more effective communicator (andso can communicate the benefits and requirements

of a coordination), she is better positioned tobecome a leader. To the extent that an individualderives an asymmetrically greater benefit from agiven collective effort, she will be more motivatedto expend coordinative effort leading the group toachieve the group gain.10 Equally, a leader needsthe knowledge and intelligence to make thosecoordinative decisions that lead to success in ajoint enterprise. That is, she must be cognitivelyequipped to determine what causal steps areneeded to successfully orchestrate the intendedjoint outcome. Gains from n-person enterprisescan be achieved or lost depending on whether theplan corresponds to reality. Finally, a leader needsto choose, out of the available array of alternativeplans of action, those coordinated efforts that (1)benefit her, while also (2) sufficiently benefiting alarge enough subset of the potential participantsthat they (3) support the joint project stronglyenough for it to succeed. Social groups areconfronted with an indefinitely large array ofpotential joint enterprises, each with its ownmatrix of payoffs. A leader must be able tocompute how different proposals will differentiallyimpact the perceived welfare of potential partici-pants. A leader who proposes enterprises thatdifferentially benefit herself at the expense ofothers will not be supported by others, while anindividual who sacrifices her own interests toostrongly for the enterprise will not be motivated toexpend effort to act as leader. To be successful, aleader must be able to anticipate, perceive andrepresent the perspective-specific values of theinteractants, in order to converge on thosecollective enterprises with sufficient political sup-port. Persuasion involves the ability to discoverand represent the values of others, and to awakena sense of how personally valuable the anticipatedoutcome of the enterprise will be to the audience.Of course, to the extent that an individualnaturally and inextricably benefits from the sameset of outcomes that most other group members do(although perhaps to a greater degree), she isbetter positioned to become a leader. Of course,leadership is not an all or nothing matter: Allparticipants actively or passively play a role indirecting coordinative effort, instigating or inhibit-ing action, and otherwise exercising social influ-ence on the direction that the collective enterpriseswe participate in will take.

Finally, it is worth briefly considering how theseproposals concerning the evolved psychology of



n-person exchange provides a theory of thetransmission of moral precepts in organizationsand communities. We think that the willingness tofollow many classes of moral rules within com-munities and organizations is implicitly treated bythe mind as a form of n-person exchange.Logically speaking, there need be no relationshipbetween whether some individuals follow a ruleand whether others do, if the goal is to be ethical.Yet, we implicitly treat many rules as if followingthem were somehow conditional on others’ con-duct. Many people feel sentiments according to thefollowing social exchange logic: I will give up thebenefits of violating this moral rule if others in mysocial world do. If I followed the rule, and you didnot, I have been cheated by you. The more otherscheat on a rule I follow, the more exploited I feel,and the more tempted I am to discontinuefollowing the rule when it is costly to do so.Whether or not we act on it, we feel a link betweenour motivation to follow moral rules and others’adherence to them. The hypothesis that morality istreated psychologically as a collective exchangealso provides an explanation for our otherwiseirrational response to hypocrisy. Although thevalue of an argument is logically unrelated to whosays it, individuals act as if their willingness toadopt a moral proposal depends on whether theproposer follows it herself. Nothing destroys aleader’s credibility faster than the discovery thatshe has been a hypocrite. It is as if a cheaterproposes a social exchange: Our psychology isdesigned to avoid entering into exploitive relation-ships, and adopting a moral rule from one who isnot following it maps into our exchange psychol-ogy as an instance of allowing oneself to becheated. This hypocrisy circuit makes no senselogically, but makes evolutionary psychologicalsense to the extent that morality in organizationsand communities is one expression of our ex-change psychology.

CONCLUSIONS

The picture of n-person exchange psychology thatwe have sketched here is of course highlysimplified both computationally and game theore-tically. For example, we have not been able todiscuss a number of game theoretic complexitiesinvolving the strategy sets against which punitive

cooperator strategies prosper, the likely effects offrequency-dependent dynamics on the evolution ofthe mechanisms, and the necessary additionalcognitive abilities and motivational circuits hu-mans must have for conditional cooperation,punishment, leadership, and morality to emerge.Nevertheless, we think that this treatment canprovide some insights into how humans arepsychologically designed behave in organizations,and what variables will govern the dynamics,success, and failure of various organizationalforms.

Acknowledgements

Thanks to Pat Barclay, Will Brown, Oliver Curry, Ed Hagen,Lin Ostrom, Jade Price, Steve Rothstein, Don Symons,Masanori Takezawa, Robert Trivers, members of the Beha-vioral Science Group at the Santa Fe Institute, and members ofthe Center for Evolutionary Psychology Research Seminar.Funding was provided by the Indiana University Workshop inPolitical Theory and Policy Analysis, and by a grant to theSanta Fe Institute from the James S. McDonnell Foundation-21st Century Collaborative Award for the Study of ComplexSystems.

NOTES

1. It is better not to think of economic rationality as anexplanation for behavior}that is, as a set of causesof anything. That is, theories of economic rationalityare high level post hoc descriptive idealizations ofoutputs, rather than a model of a real process thatproduces or explains behavior. As economics, orga-nizational behavior, and psychology mature andintegrate, the ground level theories will be detaileddescriptions of the information-processing architec-tures of the mechanisms in the human brain, and howthey interact in sets in structured environments.Using rationality as an account of choice is parallelto using animism as an account for the ability ofanimals to move}admirably descriptive, while fail-ing to track or connect with the science of actualcausation.

2. One can view the phenomenon of increasing con-servatism with increasing scale either negatively orpositively, as either cognitive inertia or the coordi-native facilitation of pre-existing coordinative struc-tures. On the plus side, far more large-scalecooperation is realized than otherwise would exist ifhumans had to build exchange structures de novo foreach individual opportunity for an interaction torealize a gain in trade. On the downside, the fact thatthe revision of pre-existing practices is increasinglyretarded the larger the interaction structure meansthat larger organizations increasingly grandfather inobsolete and nonfunctional features. Conversely, thesmaller the interaction set, the more dynamic it will

J. TOOBY ET AL.126


be, and the more rapidly it will be tailored to meetchanging conditions. The larger the structure, themore processes of cultural transmission will beimportant, while the smaller the structure, the moregame-theory like interactive responses to existingconditions will predominate. This principle explainswhy small, young companies more effectively exploitnew economic opportunities, despite the capitaliza-tion advantages large companies enjoy.

3. By parallel logic, it should be easier to grow acoordinated n-person exchange structure from asmaller to a larger size than to create it de novo at alarge size.

4. This is true even of riots, which usually outgrow theinitial focus or pretext and become a wide-rangingopportunity to realize those joint values of the riotersthat are usually individually deterred by a lack ofsufficiently coordinated collective effort.

5. Socially marginal individuals should be more eager tocovertly feed disputes among others.

6. Cults, for example, place a strong emphasis oncutting ties to families and other outsiders (e.g.Lalich, 2004), and basic training isolates new recruitsfrom family members for the first several months oftraining.

7. We consider anger to be the expression of an evolvedemotion program (a functionally structured neuro-computational system) whose design features andsubcomponents evolved to advantageously regulatethinking, motivation, and behavior in the context ofresolving conflicts of interest in favor of the angryindividual. Two negotiating tools regulated by thissystem are the threat of inflicting costs (aggression)and the threat of withdrawing benefits (the down-regulation of cooperation). Humans have a systemthat, in each individual, recognizes the welfare trade-off ratio expressed in the actions another party takeswith respect to oneself. Anger is conceptualized as amechanism whose functional product is the recali-bration in the mind of another of this other person’swelfare trade-off ratio with respect to oneself (orother valued agent). That is, the goal of the system isto change the targeted persons’ disposition to makewelfare trade-offs so that they more strongly favorthe angered individual (or their group) in the presentand the future. As in animal contests, the target ofanger may relinquish a contested resource, or maysimply in the future be more careful to help or toavoid harming the angered individual. In cooperativerelationships, where there is the expectation that thecooperative partner will spontaneously take thewelfare of the individual into account, the primarythreat from the angered person that potentiallyinduces recalibration in the targeted individual isthe signaled possibility of the withdrawal of futurehelp and cooperation if the welfare trade-off ratio isnot modified. In the absence of cooperation, theprimary threat is the infliction of damage. Conceptsthat are anchored in the internal psychologicalvariable welfare trade-off ratio include respect, con-sideration, deference, status, rank, and so on. An-ger is an ancient system, and provides much of the

evolved infrastructure out of which the punitivesentiment system evolved. Indeed, to the extent thatour evolved psychologies are designed to recognizethat free riding inhibits n-person exchanges andthe gains they generate (while simultaneously bene-fiting the free rider) the anger system wouldprocess this as a situation in which the free rider istrading off the welfare of others for her welfare. Tothe extent that the exploiter is inflicting unacceptablyhigh costs in pursuit of insufficient personalgains, this should trigger anger at exploitation.That is, anger and punitive sentiment are closelyrelated functional systems, using much of the sameinfrastructure.

8. Individual selection and group selection have come toacquire such a broad variety of meanings that it isdifficult to speak briefly about them with any clarity.David Sloan Wilson, for example, uses such a broaddefinition of group selection that dyadic reciproca-tion and kin selection both become examples of‘group selection’ because they involve interactions in‘groups’ (Wilson, 1980). We prefer to restrict the termgroup selection to those cases where the higher level(‘group’ derived) component of the fitness of theorganism derives from an organism’s stable member-ship in a single group for a majority of its life history,rather than from its transient participation inthousands of different ‘groups’ or (as we would putit) interactions. In our view of the evolution of n-person exchange, different individuals (embodyingdifferent designs) accumulate fitness differences be-cause of how they act across large numbers ofdifferent social interactions. That is, a person mightengage in 20 dyadic interactions in a day, and 8triadic interactions, and 1 twenty-person interaction.Confusion arises by labeling every social interaction agroup.

9. A strategy is defined computationally by the detailedspecification of a behavior control program, includ-ing how it represents conditions in order to regulatebehavior. The nature of the implementation matters:Consider, for example, a program design for apunitive strategist that sums both punitive effortand cooperative effort together into one unitaryaccount as its definition of the contribution itconsiders either self or others makes to the collectiveenterprise. This particular design will not suffer froma second-order free rider problem. The punitivestrategist will contribute less nonpunitive effort tomake up for the additional punitive effort it iscontributing. Alternatively, it could insist that non-punitive cooperators make a larger nonpunitivecontribution to make up for the additional punitiveeffort it is expending. Indeed, the second-order freerider problem only exists to the extent that partici-pants do not compute punishment as itself acontribution. But punishment is in reality a contribu-tion, and punitive strategists (as well as rationalcooperators, if any) will view it as such. In any case, amutant design that considers punishment a coopera-tive contribution cannot be exploited by second-orderfree riders. It may prosper, or not, depending on the



details of how conditional cooperation is implemen-ted in nonpunitive cooperators.

10. It is a commonplace of democratic politics thatmany proposals that would modestly benefit largemajorities of the population are not acted on, infavor of other proposals that strongly benefit farsmaller fractions of the population.

REFERENCES

Axelrod R, Hamilton WD. 1981. The evolution ofcooperation. Science 211: 1390–1396.

Baron-Cohen S, Leslie AM, Frith U. 1985. Does theautistic have a ‘theory of mind’? Cognition 21: 37–46.

Boyd R, Gintis H, Bowles S, Richerson PJ. 2003. Theevolution of altruistic punishment. Proceedings of theNational Academy of Sciences (USA) 100: 3531–3535.

Boyd, R, Richerson PJ. 1992. Punishment allows theevolution of cooperation (or anything else) in sizablegroups. Ethology and Sociobiology 13: 171–195.

Brase GL. 1998. Human reasoning about social groups:evidence of evolved mechanisms for a coalitionalpsychology. Dissertation Abstracts International:Section B: The Sciences and Engineering 58(9-B):5150 (UMI# 9809613).

Brewer MB. 1991. The social self: on being the same anddifferent at the same time. Personality and SocialPsychology Bulletin 17: 475–482.

Cosmides L. 1989. The logic of social exchange: hasnatural selection shaped how humans reason? Studieswith the Wason selection task. Cognition 31: 187–276.

Cosmides L, Tooby J. 1989. Evolutionary psychologyand the generation of culture, Part II. Case study: acomputational theory of social exchange. Ethologyand Sociobiology 10: 51–97.

Cosmides L, Tooby J. 1992. Cognitive adaptations forsocial exchange. In The Adapted Mind: EvolutionaryPsychology and the Generation of Culture, Barkow JH,Cosmides L, Tooby J (eds). Oxford University Press:New York; 163–228.

Cosmides L, Tooby J. 2000. Consider the source: theevolution of adaptations for decoupling and metar-epresentation. In Metarepresentations: A Multidisci-plinary Perspective, Sperber D (ed.). OxfordUniversity Press: New York; 53–115.

Cosmides L, Tooby J. 2005. Neurocognitive adaptationsdesigned for social exchange. In The Handbook ofEvolutionary Psychology, Buss DM (ed.). Wiley: NewYork.

Darley JM, Latane B. 1968. Bystander intervention inemergencies: diffusion of responsibility. Journal ofPersonality and Social Psychology 8: 377–383.

Dumas A. 1844/2001. The Three Musketeers. RandomHouse: New York.

Ehrhart K-M, Keser C. 1999. Mobility and Cooperation:On the Run. Scientific Series. CIRANO: Montreal.

Fehr E, Fischbacher U. 2003. The nature of humanaltruism. Nature 425: 785–791.

Fehr E, Gachter S. 2000. Cooperation and punishmentin public goods experiments. American EconomicReview 90: 980–994.

Fischbacher U, Gachter S, Fehr E. 2001. Are peopleconditionally cooperative? Evidence from a publicgoods experiment. Economics Letters 71: 397–404.

Fiske, AP. 1991. Structures of Social Life: The FourElementary Forms of Human Relations. Free Press:New York.

Henrich J. 2004. Cultural group selection, co-evolutionary processes and large-scale cooperation.Journal of Economic Behavior and Organization 53:3–35.

Henrich J, Smith N. 2004. Comparative experimentalevidence from Machiguenga, Mapuche, Huinca andAmerican populations. In Foundations of HumanSociality, Henrich J, Boyd R, Bowles S, Camerer C,Fehr E, Gintis H (eds). Oxford University Press:Oxford; 125–167.

Hoffman E, McCabe KA, Smith VL. 1998. Behavioralfoundations of reciprocity: experimental economicsand evolutionary psychology. Economic Inquiry 36:335–352.

Kurzban R, McCabe K, Smith VL, Wilson BJ. 2001a.Incremental commitment and reciprocity in a real timepublic goods game. Personality and Social PsychologyBulletin 27: 1662–1673.

Kurzban R, Tooby J, Cosmides L. 2001b. Can race beerased?: Coalitional computation and social categor-ization. Proceedings of the National Academy ofSciences 98(26), 15387–15392 (December 18, 2001;Epub 2001 Dec 11. PMID: 11742078).

Lalich J. 2004. Bounded Choice: True Believers andCharismatic Cults. University of California Press:Berkeley.

Ledyard JO. 1995. Public goods: a survey of experi-mental research. In The Handbook of ExperimentalEconomics, Kagel JH, Roth AE (eds). PrincetonUniversity Press: Princeton; 111–194.

Lubell M, Scholz JT. 2001. Cooperation, reciprocity andthe collective-action heuristic. American Journal ofPolitical Science 45: 160–178.

Madison J. 1788/1982. Federalist paper no. 10. In TheFederalist Papers (2nd ver), Hamilton A, Jay J,Madison J (eds). Bantam Classics; Reissue edition:New York.

Masclet D, Noussair C, Tucker S, Villeval M. 2003.Monetary and nonmonetary punishment in thevoluntary contributions mechanism. AmericanEconomic Review 93: 366–380.

Maynard Smith J. 1982. Evolution and the Theory ofGames. Cambridge University Press: Cambridge, UK.

Olson M. 1965. The Logic of Collective Action: PublicGoods and the Theory of Groups. Harvard UniversityPress: Cambridge.

Orbell J, Dawes R. 1991. A ‘cognitive miser’ theory ofcooperators’ advantage. American Political ScienceReview 85: 515–528.

Orbell J, Dawes R. 1993. Social welfare, co-operators’ advantage, and the option of notplaying the game. American Sociological Review 58:787–800.

J. TOOBY ET AL.128


Ostrom E. 1990. Governing the Commons: The Evolutionof Institutions for Collective Action. CambridgeUniversity Press: New York.

Ostrom E. 1998. A behavioral approach to the rationalchoice theory of collective action. PresidentialAddress, American Political Science Association,1997. American Political Science Review 92: 1–22.

Panchanathan K, Boyd R. 2004. Indirect reciprocity canstabilize cooperation without the second-order freerider problem. Nature 432: 499–502.

Price ME. 2003. Pro-community altruism andsocial status in a Shuar village. Human Nature 14:191–208.

Price ME. in press A. Monitoring, reputation and‘greenbeard’ cooperation in a Shuar work team.Journal of Organizational Behavior.

Price ME. in press B. Punitive sentiment among theShuar and in industrialized societies: cross-culturalsimilarities. Evolution and Human Behavior.

Price ME, Cosmides L, Tooby J. 2002. Punitivesentiment as an anti-free rider psychological device.Evolution and Human Behavior 23: 203–231.

Price ME, Tooby J, Cosmides L. in preparation. Theevolutionary psychology of work teams.

Sidanius J, Pratto F. 1999. Social Dominance: AnIntergroup Theory of Social Hierarchy and Oppression.Cambridge University Press: New York.

Stone V, Cosmides L, Tooby J, Kroll N, Knight R.2002. Selective impairment of reasoning about socialexchange in a patient with bilateral limbic systemdamage. Proceedings of the National Academy ofSciences 99(17):11531–11536.

Tiger L. 2004. Men in Groups (3rd edn). TransactionPublishers: New York.

Tooby J, Cosmides L. 1992a. The psychologicalfoundations of culture. In The Adapted Mind: Evolu-tionary Psychology and the Generation of Culture,Barkow JH, Cosmides L, Tooby J (eds). OxfordUniversity Press: Oxford; 19–136.

Tooby J, Cosmides L. 1992b. Ecological rationality andthe multimodular mind: grounding normative theories

in adaptive problems. Center for Evolutionary Psy-chology Technical Report 92-1 (Reprinted in: Tooby J,Cosmides L. (in press) Evolutionary Psychology:Foundational Papers. With a foreword by StevenPinker. MIT Press: Cambridge).

Tooby J, Cosmides L. 1996. Friendship and the Bank-er’s Paradox: Other pathways to the evolution ofadaptations for altruism. In Evolution of SocialBehaviour Patterns in Primates and Man, RuncimanWG, Maynard Smith J, Dunbar RIM (eds). OxfordUniversity Press: Oxford; 119–143.

Tooby J, Cosmides L. 2000. Cognitive adaptations forkin-based coalitions: human kinship systems at theintersection between collective action and kin selec-tion. Current Anthropology 41: 803–804.

Tooby J, Cosmides L. 2002. The evolution of collectiveaction: an adaptationist dissection. Human Behaviorand Evolution Society Meetings, Rutgers University,Newark, NJ, 19–23 June.

Tooby J, Cosmides L. 2005. The functional architectureof human motivation. Paper presented at the HumanBehavior and Evolution Society Meetings, Austin, TX.

Tooby J, Cosmides L. forthcoming. Human coalitionalpsychology. Oxford Handbook of Evolutionary Psy-chology. Dunbar RIM, Barrett L (eds). OxfordUniversity Press: New York.

Tooby J, Cosmides L, Barrett HC. 2005. Resolving thedebate on innate ideas: learnability constraints and theevolved interpenetration of motivational and concep-tual functions. In The Innate Mind: Structure andContents, Carruthers P, Laurence S, Stich S (eds).Oxford University Press: Oxford.

Trivers RL. 1971. The evolution of reciprocal altruism.The Quarterly Review of Biology 46: 35–57.

Wilson DS. 1980. The Selection of Populations andCommunities. Addison-Wesley: Menlo Park.

Williams GC. 1966. Adaptation and Natural Selection.Princeton University Press: Princeton.

Yamagishi T. 1986. The provision of a sanctioningsystem as a public good. Journal of Personality andSocial Psychology 51: 110–116.



Cognitive Adaptations for n-person Exchange: The ... · Cognitive Adaptations for n-person...

Documents

Transcript of Cognitive Adaptations for n-person Exchange: The ... · Cognitive Adaptations for n-person...