Runaway Social Selection for Displays of Partner Value and Altruism

13 Pages·2007·0.14 MB·English

Runaway Social Selection for Displays of Partner
Value and Altruism
RandolphM.Nesse Abstract
DepartmentsofPsychiatry,Psychology, Runaway social selection resulting from partner choice may
andResearchCenterforGroupDynamicsatthe have shaped aspects of human cooperation and complex so-
InstituteforSocialResearch cialitythatareotherwisehardtoaccountfor.Socialselection
UniversityofMichigan
isthesubtypeofnaturalselectionthatresultsfromthesocial
AnnArbor,MI,USA
behaviors of other individuals. Competition to be chosen as
[email protected]
asocialpartnercan,likecompetitiontobechosenasamate,
result in runaway selection that shapes extreme traits. Peo-
plepreferpartnerswhodisplayvaluableresourcesandbestow
them selectively on close partners. The resulting phenotypic
covariancebetweendisplaysandpreferencesgivesfitnessad-
vantagestoboth,creatingrunawayselectionthatcouldshape
awholesuiteofprosocialtraitsincludingaltruism,moralca-
pacities,empathy,andtheoryofmind.Eventhoughtheygive
a net fitness benefit, traits at the endpoint of runaway social
selectioncanhavesubstantialdeleteriouseffectsonothertraits
suchasviability,abilitytoaccumulateresources,orvulnera-
bilitytomentaldisorders.Socialselectionforcesarisingfrom
self-interested partner choices may be an invisible hand that
shapedcapacitiesforcommitment,altruism,andotherproso-
cialcapacitiesofthehumansocialbrain.
Keywords
altruism,commitment,cooperation,evolution,morality,part-
nerchoice,relationships,runawaysocialselection
February12,2007;revisedandacceptedMay26,2007
BiologicalTheory2(2)2007,143–155.(cid:1)c2007KonradLorenzInstituteforEvolutionandCognitionResearch 143
RunawaySocialSelectionforDisplaysofPartnerValueandAltruism
Thediscoveryofevolutionaryexplanationsforcooperationis to suggest that the fitness benefits of being chosen as a part-
oneofthegreatachievementsoflate20th-centurybiology.As ner may shape extreme displays of partner value, including
most readers know, benefits to the group rarely explain ten- capacitiesforgenuinealtruism,thatareotherwisedifficultto
denciestohelpothers(Williams1966;Dawkins1976),bene- explain.
fits to kin explain altruism in proportion to the coefficient of
relatedness(Hamilton1964),andmutualbenefitsandrecipro-
SocialSelection
calexchangesexplainmuchcooperationbetweennonrelatives
(Trivers 1971). Subsequent theoretical and empirical studies Socialselectionisthesubtypeofnaturalselectioninwhichfit-
have blossomed into a body of knowledge that can explain nessisinfluencedbythebehaviorofotherindividuals(West-
much social behavior (Wilson 1975; Trivers 1985; Dugatkin Eberhard 1979, 1983; Wolf et al. 1999; Frank 2006). Al-
1997;Alcock2001;Hammerstein2003). though well established in biology, the term social selection
Controversiescontinue,however.Somearisefromapro- isslightlyproblematicbecauseepidemiologistsusethesame
fusion of models for cooperation that use inconsistent termi- phrasetodescribetheentirelydifferentphenomenonofsome
nology and tend to emphasize one explanation when several social groups having a higher proportion of individuals with
mayapply(Frank1998;Hirshleifer1999;Hammerstein2003; some condition. For instance, the proportion of people with
Lehmann and Keller 2006; Nowak 2006; West et al. 2007). schizophrenia is higher in inner cities simply because many
Othercontroversiesreflectimpassioneddebatesabouthuman cannotaffordtoliveelsewhere.Alsopotentiallyconfusingis
nature(Midgley1994;Wright1994;Ridley1997;Segerstra˚le the idiosyncratic use of social selection as an alternative to
2000; de Waal et al. 2006; Dugatkin 2006). However, some sexualselection(Roughgardenetal.2006),wheninfactitisa
controversies persist because no explanation seems entirely subtype. These potential confusions aside, social selection is
satisfactoryforsomephenomena,especiallyhumancapacities thestandardtermforfitnesschangesresultingfromthesocial
foraltruismandcomplexsociality. behaviorsofotherindividuals.
Whilekinselectionandvariationsonreciprocityexplain Sexualselectionbyfemalechoiceisthebest-knownsub-
most human capacities for cooperation, some observations typeofsocialselection.Femalebiasesformatingwithorna-
don’t fit the usual models. In behavioral economics labora- mentedmalesselectformoreelaboratemaledisplays,andthe
toryexperimentsandineverydaylife,peopletendtobemore advantagesofhavingsonswithextremedisplays(andperhaps
altruistic than predicted (Gintis 2000; Fehr and Rockenbach advantagesfromgettinggoodgenes)selectforstrongerprefer-
2004; Brown and Brown 2006; de Waal et al. 2006). They ences(Grafen1990;Kokkoetal.2003).Theresultingpositive
alsotendtopunishdefectorsevenwhenthatiscostly(Henrich feedbackmakesdisplaysandpreferencesmoreandmoreex-
and Boyd 2001; Boyd et al. 2003). People follow rules and treme until genetic variation is exhausted, or until the fitness
theyarepreoccupiedwithmoralsandmores,monitoringand increase from more matings equals the fitness decrease from
gossiping about even minor deviations (Axelrod 1986; Katz lowered competitive ability and earlier mortality (Andersson
2000;Krebs2000;deWaaletal.2006).Perhapsmostinterest- 1994;Kokkoetal.2006).Sexualselectionissocialselection
ingofall,closefriendstakepainstoavoidmakingexchanges becauseindividualfitnessisinfluencedbythechoicesandbe-
explicitbecausecallingattentiontothemharmsrelationships haviors of other individuals. West-Eberhard made the point
(Batson1991;MillsandClark1994;Dunbar1996;Toobyand succinctlyitinoneofthefirstpapersonthetopic:
Cosmides1996;Nesse2001,2006;BrownandBrown2006).
Sexualselectionreferstothesubsetofsocialcompetitioninwhich
Friendships exist, but they remain in want of a satisfactory
the resource at stake is mates. And social selection is differential
evolutionaryexplanation(Smuts1985;Silk2003).
reproductivesuccess(ultimately,differentialgenereplication)dueto
Thisarticlearguesthatwell-establishedmodelsofsocial
differential success in social competition, whatever the resource at
selectionmayexplainhowpartnerchoicecouldshapeextreme
stake.(1979:158).
prosocialtraitsinhumans.Itbeginsbyreviewingearlydescrip-
tionsofsocialselection(West-Eberhard1979,1983)andmore Social selection arising from conspecific choices and behav-
recentformalmodelsthatillustratethevalueofcalculatingthe iorshasbeendescribedindetail(Crook1972;West-Eberhard
fitnesscomponentsfromsocialselectionseparatelyfromthose 1975; West-Eberhard 1983; Tanaka 1996; Wolf et al. 1999;
thatarisefromtherestofnaturalselection(Tanaka1996;Frank Frank2006).Suprisingly,however,itsfullpowerisonlynow
1998,2006;Wolfetal.1999).Next,itreviewstherecentrecog- beingrecognized.Theperspectiveofsocialselectionshiftsat-
nitionofthepowerofpartnerchoice(Noe¨ andHammerstein tentionawayfromindividualstrategiesiniteratedexchanges,
1994)andconnectstheseinsightswithrecentmodelsofhow andtowardthepriorandlargerfitnesschallengesofidentifying
covariance of partner phenotypes can lead to runaway social thebestavailablepartnersanddoingwhateverwillgetthemto
selection(Tanaka1996;BredenandWade1991).Theselines chooseoneasapartner.Informalmodels,thismeanspartition-
of work come together with recent work on human altruism ingtheforceofsocialselectionresultingfromthecovariance
144 BiologicalTheory2(2)2007
RandolphM.Nesse
of partners’ phenotypes separately from other forces of nat- plays will both be present in the same individuals. Also,
ural selection. Following Queller (1992) and Frank (1997), benefits to others pay off not only directly, but also because
Wolfetal.(1999)describesocialselectionbysaying“factors benefitstopartnerseventuallyresultinbenefitstotheselfvia
other than one’s own phenotype may affect an individual’s interdependence(Rothstein1980;Humphrey1997;Brownand
fitness...individual variation in fitness can be attributed to Brown 2006). At equilibrium, many individuals will be pre-
variation in the value of traits expressed by an individual’s sentingandassessingexpensivedisplaysinacompetitionthat
socialpartners.”(1999:255–256). results in partnerships between individuals of similar partner
Building from Lande and Arnold’s model of sexual se- value.
lection (1983), Wolf. et al. (1999: 256) partition relative Insexualselection,runawayoccursonlywhenthecovari-
fitness ω into one component from social selection and a ance of the trait and the display is greater than the viability
separate component from the rest of natural selection: ω= decreasefromthedisplay.Atequilibrium,furtherincreasesin
α+β z +β z(cid:1) +ε,whereβ isthenaturalselectiongradi- female preference would lower fitness because of decreased
N i S j N
ent,β isthesocialselectiongradient,z isthetraitintheindi- viability of sons (Kokko et al. 2006). However, “even small
S i
vidualandz´ isacovaryingtraitinthepartner(theprimesign changes in female behavior (which cost little) can generate
j
indicatesthatthetraitisinthepartner,αisfitnessuncorrelated strong selection when a male’s fitness depends primarily on
withthetraits,andεiserror).Theythenderiveageneralized his mating success” (Kokko et al. 2006: 59). In selection for
phenotypic version of Hamilton’s rule to show that selection social partners, the cost of choosing partners with extremely
favors an altruistic trait z whenever Cij(cid:1)β +β >0, where highvaluehaslittleornodisadvantagecomparabletothedis-
Cij(cid:1) isthephenotypiccovaijriancebetwePeiintShetraNitintheindi- advantageexperiencedbyfemaleswhochoosemateswiththe
vidualandthepartner,andP isthecharacter’svariance.Here, most extreme displays. Displays of partner value will, there-
ii
β istheselectioncostforanaltruistictrait(andwilltherefore fore,continueunderdirectionalselectionuntiltheirmarginal
N
be <0), and β is the benefit to the partner (which will be benefitsimposeequalcoststootherfitnesscomponents,such
S
>0), so the altruistic trait will be selected only if its covari- asabilitytoaccumulatematerialresources.Thus,socialselec-
ance with the associated trait is large compared to the trait’s tionforpartnerscan,likesexualselection,explainextremely
variance. The model, very similar to Frank’s (1997, 1998) costlytraits.
andalsodrawingonFisherandPrice,isbasedonphenotypes In a model of social selection that emphasizes signaling
and does not require covariance of genes within individuals. submissionandrealfightingability,Tanaka(1996)addresses
Partner choice creates phenotypic covariance that can shape the possibility of runaway social selection more directly. As
extreme traits such as displays of one’s value as a partner. intheWolfetal.model,fitnessispartitionedintocomponents
Howfarwillsocialselectionpushsuchtraitsattheexpenseof from social selection that are distinct from the rest of natu-
othercomponentsofnaturalselection?Ananswertothisim- ral selection in order to assess where the equilibrium for a
portantquestionrequiresdetailedanalysisofsocialselection signallies.Thatpointoftenisreached,heconcludes,byrun-
bypartnerchoice. away selection that quickly arrives at the equilibrium where
Whileallsocialbehavioraltendenciescanbeinterpreted themarginalbenefitsoffurtherincreasingthesignalarebal-
asproductsofsocialselectionbecausetheyinvolvechoiceby anced by its direct costs. Crespi (2004) has argued that such
otherindividuals(Wolfetal.1999;Frank2006),theemphasis positive feedback cycles are much more common in nature
here is on forces of selection that arise from choices about thanisusuallyrecognized.Deceptionandcheatinghavebeen
relationshippartnersandgroupmembership.Ifpotentialpart- majorthemesinreciprocityresearch,andtheyapplyinsocial
ners or group members vary in resources and tendencies to selectionmodels,buttheireffectsarelimitedbyinexpensive
reliablybestowthemonclosepartners,thenapreferencefor gossipaboutreputationsandbythedifficultyoffakingexpen-
resource-rich, selectively altruisticpartners willgive aselec- siveresourcedisplays(Tanaka1996).
tiveadvantage.Beingpreferredasapartnergivesfitnessadvan-
tages because itgives morepossiblepartnerstochoose from
SocialSelectioninNature
(Noe¨ and Hammerstein 1994). This will select for displays
ofresourcesandselectivealtruismthatreflectanindividual’s If the above models are correct, then examples of nonsexual
potentialvalueasapartner. socialselectionshouldbeobservedinthenaturalworld.Some
The nonrandom association of individuals with extreme examples of traits shaped by preferences in one species for
displays and those with strong preferences can result in run- displaysinanotherspeciesillustraterunawayselectionwithout
away social selection increasing both traits to extremes that genetic covariation in the same genome. As Darwin noted
decrease other fitness parameters (Breden and Wade 1991; (1871),flowershaveelaborateanddiverseformsbecausethey
Tanaka 1996; Wolf et al. 1999). This model differs from compete to satisfy pollinator preferences. Flowers preferred
sexual selection because in most cases preferences and dis- by pollinators contribute more genes to future generations,
BiologicalTheory2(2)2007 145
RunawaySocialSelectionforDisplaysofPartnerValueandAltruism
so floral displays become increasingly extravagant until the Domestication
marginalbenefitsfromattractingmorepollinatorsarematched
by costs to other aspects of fitness, such as investment in Domestication illustrates how social preferences can shape
leaves and roots (Armbruster et al. 2005). Benefits can also profoundlyprosocialtraits.Itrequiresnoconsciousbreeding,
come from not being chosen. Staying near the center of a onlypreferencesthatinfluencefitnessamongindividualsfrom
selfishherdisshapedbypredatorpreferences.Stottingprotects the other species who vary on traits that matter to humans
gazellesbecauseitisanhonestsignalofvigorthatdiscourages (Price 1984; Diamond 2002). For instance, wolves with less
predatorsfromuselesschases. fear of humans and lower levels of aggression were able to,
Signals between members of the same species are and allowed to, stay closer to ancestral human camps where
shapedbythesamemechanisms(Grafen1984;Bradburyand thefitnessvalueoffoodscrapswasadomesticatingselection
Vehrencamp1998).Socialcoordinationsignalsareubiquitous. force.Inturn,thosehumanswhohadtendenciestobealtruis-
Forinstance,abirdonanestmakesdistinctivemovementsto tictowarddog-progenitorsreceivedfitnessbenefits—initially
signal to its partner that it is ready to trade roles. The signal warningsofdanger,butlater,helpinthehuntandprotection.
benefitsbothparties,sothereisnoselectionforanextremesig- Thisprocessselectedforgenesthatincreasehumanaltruism
nal. In competitive situations, amplified signals are common toward dogs, and it shaped dogs who behave in ways that
(Takana1996).Whenawolfbaresitsthroattosignalyielding pleasehumansenormously.
in a fight, both parties benefit by avoiding the danger of an Humans also show many characteristics of being
escalated fight; a prominent submission display that creates domesticated—lowratesofaggression,increasedcooperation,
realvulnerabilitypaysoffbyavoidinguselessfighting.Status eagernesstopleaseothers,andevenchangesinbonestructure
displays in lieu a fight are likely to be extreme because only similartothosecharacteristicofdomesticatedanimals(Leach
expensive honest signals will influence the competitor. Note 2003).Itseemsplausiblethathumanshavebeendomesticated
thatsuchsignalingbehaviorsgivebenefitsonlybecausethey by the preferences and choices of other humans. Individuals
interact with the phenotypes of other individuals who have whopleaseothersgetresourcesandhelpthatincreasefitness.
beenprimedbyselectiontobeinfluenced. Aggressiveorselfishindividualsgetnosuchbenefitsandare
Someexamples,suchasmalescompetingforaterritory, atriskofexclusionfromthegroup,withdireeffectsonfitness.
blur the boundary between sexual and other social selection. Theresultisthoroughlydomesticatedhumans,someofwhom
Othersarisemoreclearlyfromnonsexualsocialselection,such canbeenormouslypleasing.
as the huge brightly colored beaks of both male and female Thisprocessdoesnotdependonthesuccessofthegroup.
toucans. They do not result from sexual selection; nonsocial Instead,individualsconstantlymakesmallself-interestedso-
toucanspecieshavelessexaggeratedandmoresexuallydimor- cial choices that shape the behaviors of others who learn to
phic beaks. They are more likely honest signals of ability to dowhateverworks.Theresultingeffectsonfitnessshapethe
defend a nesting territory (West-Eberhard 1983). Bright col- speciesbysocialselection.Thisprocessoffersadramaticex-
oration in both sexes is also prominent in territorial lizards ampleofaBaldwineffect,inwhichlearningshapesadaptive
and some mammals, especially lemurs. Social selection has behaviorpatternsthatcreatenewselectionforcesthatrapidly
also been proposed as the explanation for bright coloration facilitatebetterabilitytoexploitthenewniche(Dennett1995;
of reef fish. West-Eberhard offers a wealth of examples, and Laland et al. 2000; Weber and Depew 2003; West-Eberhard
reasons why species recognition hypotheses are insufficient 2003; Ananth 2005). Once the benefits of relationships in-
(1983). She also notes that Wynne-Edwards (1962) provides creasedaboveacrucialthreshold,theycreatedanewlycom-
additionalexamples,evenifhewaswrongabouthowselection plexsocialenvironmentwhereindividualswithspecialsocial
shapedthem.Thisisespeciallyimportantbecauseithighlights skillsgotincreasingfitnessadvantages shapedmoreextreme
the power of social selection to account for phenomena that cognitive and prosocial traits (Humphrey 1976; Byrne and
might otherwise appear to be products of group selection. Whiten1988;Alexander2005).
While the sources of female ornamentation remain an active HerbertSimon,ina1990articleon“socialselectionand
research focus, a recent review endorses the importance of successful altruism,” described how selection for “docility”
socialselection: couldgiverisetobehaviorsthatbenefitothersmorethanthe
self. Simon defined docility as, “persons who are adept at
Almost20yearsago,West-Eberhardarguedthatmonomorphicshowy
sociallearningwhoacceptwelltheinstructionsocietyprovides
plumagewasassociatedwithaggressivesocialdisplays(overterrito-
them”(p.1666).Hismodelisbasedonthefitnessbenefitsof
riesorotherresources)bybothsexes.Herargumentwassupportedby
general social learning, and the assumption that “limits on
examplesfromseveraltaxaincludingtoucans,parrotsandhumming
rationality in the face of environmental complexity” result
birds.West-Eberhard’ssuggestionsresultedinsurprisinglylittleem-
pirical research in the following years. However, among published in individuals behaving altruistically for the good of society
studies,mostseemtosupportherview.(Amundsen2000:151) withoutrecognizingthe“tax”theyarepaying.Incontrast,the
146 BiologicalTheory2(2)2007
RandolphM.Nesse
modeldevelopedinthisarticleviewsaltruismasaresultofthe actedtocreatecapacitiesforsocialcognition.Inmoststudies,
fitnessbenefitsofsocialselection,notasaresultofcognitive anonymous agents are randomly paired, information is only
constraints. aboutpriorbehaviorwithoneagentorthesumofallagents’
behavior,thesamealgorithmisusedforinteractionswithall
otherplayers,andonlytwooutcomesarepossible,cooperate
SocialSelectionforCooperation
ordefect.Reputationandpunishmenthaveincreasinglybeen
Indirect benefits to kin are one powerful force that shapes added to such models (Fehr and Fischbacher 2003; Axelrod
conspecificcooperation.Abilitytorecognizekin,andprefer- et al. 2004; Henrich et al. 2006). However, few reciprocity
ences for helping them, give benefits to genes in kin that are models have all of the ingredients that are important to hu-
identical by descent to those in the helper (Hamilton 1964; mancooperationincloserelationships:reputation,communi-
Dugatkin 1997; Frank 1998; Queller and Strassmann 1998; cation,agreements,promises,threats,thirdpartyenforcement,
Westetal.2002).Thisprocesshasbeendescribedandstudied and especially, opportunities to use extensive information to
soextensivelythatthereisnoneedtorepeatthedetailshere. choosepartnersfromaselectionofpossibilities(Kitcher1993;
Onesubtype,“green-beardeffects”hasbeencontroversial,but Hammerstein 2001; Nesse 2001; Noe¨ 2001). While varia-
it now appears that selection does sometimes shape kinship tions in tendencies to cooperate or defect in discrete inter-
cues that facilitate kin altruism (Queller et al. 2003). Pheno- actionswithrapidlyshiftingpartnerscertainlycreateselection
type variability can also be shaped by social interactions in- forces, they explain only some aspects of some human rela-
volvedinreproductivecompetitions,atleastinwasps(Tibbetts tionships(FehrandHenrich2003;BarclayandWiller2007).
2004). Nonetheless, such models have been a boon for the study of
Preferencesforhelpingnonrelativeswhowillhelpinre- cooperation.
turnarealsoobviouslyvaluable(Trivers1971).Thechallenge Anotherdifficultyisthatthekindsofreciprocalexchange
ishowtogetthebenefitsoftradingfavorswithoutbeingex- modeledbytheprisoner’sdilemmaseemtoberareinnature.
ploited(KrebsandDawkins1984;Alexander1987;Cosmides Most apparent examples of reciprocity identified by field re-
1989; Fehr and Fischbacher 2003). Following Price (Frank searchnowappeartobebetterexplainedbykinshipormutual
1997)andQueller(1992),Frank(1997,2006)pointsoutthat benefits(Connor1995;Stevensetal.2005).Cooperativehunt-
suchcooperationcanbemodeledascorrelatedbehaviors,an ingisaprimeexample.Participantsallgain,sodefectiondoes
informationproblemequivalenttothatofkinselection.Inkin notpay.Impalagroomingisareciprocalexchange,butofthe
selection, a behavior increases inclusive fitness if its cost to mostminimalkind.Groomingboutsaretradedbackandforth
theselfislessthanthebenefittotheothertimesthecoefficient inparcelssosmallthattheexampleblurstheborderbetween
ofrelatedness,r.Incorrelatedbehaviors,thecostisthedirect reciprocity and mutualism (Connor 1995), although groom-
effects of the behavior on the individual’s fitness, the benefit ingmaybetradableforotherresources(Mansonetal.2004).
is the indirect benefit from others (holding constant individ- Another example, previously thought to exemplify recipro-
ualbehavior),andrreflectsthesimilarityofothers’behavior, cal exchange between nonrelatives, is vampire bats sharing
thatis,theinformationanindividualhasaboutbenefitsothers blood with others who did not succeed in that night’s hunt
will likely offer. Both kin selection and correlated behavior (Wilkinson 1984). However, it turns out the sharing almost
can thus be analyzed by partitioning fitness into direct costs, alwaysisbetweenkin.Coalitionsofmalebaboonswerealso
indirectbenefits,andascalingfactorthatreflectsrelatedness thoughttodemonstratereciprocity,butonreexamination,the
in the former case, and information about other’s anticipated malesdonotsharematingopportunitiestoanygreatextent.A
behaviorinthelatter(Frank1997;Wolfetal.1999). reviewbyStevensetal.(2005)assessestheevidenceforreci-
The iterated prisoner’s dilemma has long been the dom- procity in nature and concludes that there are few examples,
inant model for cooperation based on reciprocity (Axelrod perhaps,theysay,becausemostanimalshaveseverecapacity
1984;Sigmund1993;Axelrod1997).Inthismodel,themaxi- constraintsformemoryandcognition.
mumjointbenefitfortwoplayerscomesfromrepeatedcooper- Where reciprocal helping does exist, it is usually main-
ation,butanindividualcangetagreaterpayofffromdefecting tainedbysystemsforassessingpotentialpartnersorwithdraw-
onanymovewhentheothercooperates.Tit-For-Tat(starting ingresourcesfromdefectors(Sachsetal.2004).Parceling,as
with a cooperative act and then doing what the other person inreciprocalgrooming,distributesresourcesinsmallpackets
didonthepreviousmove)isaremarkablyrobuststrategythat sodefectionisnotanissue(Connor1995).Anotherstrategyis
nicely models some human interactions. The tractability of to distribute resources selectively depending on the behavior
modelsbasedontheprisoner’sdilemmahasfosteredscoresof ofothers.Forinstance,yuccaplantsabandonflowerswithtoo
valuablestudies(Axelrod1997). many moth larvae. This can be viewed as a punishment that
It is less clear, however, that prisoner’s dilemma models selectsformothswholimiteggdeposition.However,abandon-
accuratelyreflectthekindsoftraitvariationonwhichselection ingtheflowerswithtoomanylarvaeisinthedirectself-interest
BiologicalTheory2(2)2007 147
RunawaySocialSelectionforDisplaysofPartnerValueandAltruism
oftheyuccaplant,andthismakesitadvantageousformoths hassuggestedthatsexualselectionmayaccountformanyex-
tolimitthenumberofeggslaidinanyoneflower. tremehumancognitiveandbehavioraltraitsthatareotherwise
Imagescoring(NowakandSigmund1998;Wedekindand difficulttoexplain,especiallyaltruism.Hecitesevidencethat
Milinski 2000) and other reputation-based strategies such as both women and men prefer to marry kind reliable partners,
indirectreciprocity(Alexander1987),offerinformationabout giving a fitness advantage via sexual selection to individuals
anindividual’sreliabilityasapartnerandcanleadtomutually of both sexes with these heritable personality traits. Sexual
profitableexchangesevenintheabsenceofrepeatedinterac- selectioncouldthusshapeextremealtruism.Thispotentially
tions (Riolo et al. 2001). This is not the place to analyze the important hypothesis has not been emphasized in recent lit-
diversity of cooperation models, but it is important to recog- erature,perhapsbecauseitisdifficulttostudy.Miller(2000)
nize that delayed reciprocal exchange of resources is as rare acknowledgesthatotherformsofsocialselectionmaybeim-
in other animals as it is ubiquitous in humans. Furthermore, portant, but mostly, he says, “because they change the social
humanculturesvarysubstantiallyintheirlevelsofindividual scenerybehindsexualselection.”
cooperation, with much of the variance attributable to vari- Mate choices create potent selection forces, but so do
ations in the patterns of economic exchange (Henrich et al. choices of relationship partners. The fitness benefits from
2005), further demonstrating that human cooperation strate- choosing social partners are more distant from direct repro-
giesaremarshaledtosuitthecircumstances. duction,buttheycaninfluencefitnessnearlyeverydayandat
Theroleofpartnerchoiceinfacilitatingcooperationhas allages.Ifpartnershipsyieldanetgainforbothparties,then
longbeenrecognized(BullandRice1991),buthasbeenem- fitnessincreaseswiththeincreaseinthenumberofotherswho
phasized only recently (Noe¨ and Hammerstein 1994; Noe¨ wantyouasapartner,atleastforthefirstfewpartners.Ifpart-
2001; Sachs et al. 2004). When there is choice, potential nersvaryinvalue,thenfitnesswillbeincreasedbybehaving
partners must compete in markets that change the dynamics inwaysthatincreasethenumberofotherswhowantyouasa
of cooperation. Between-species partner choice is illustrated partner(Noe¨andHammerstein1994).Agoodwaytoincrease
by symbioses in which the slower-evolving organism selects the number of available number of partners is to advertise,
amongindividualsinafasterevolvingspeciestogetthemost and to usually provide, more benefits than others can or will
valuablepartners,forinstance,theplantsymbioseswithbac- provide (Roberts 1998; Barclay and Willer 2007; Hardy and
teriaandfungi(SimmsandTaylor2002;KummelandSalant VanVugt2006).
2006).Choiceofconspecificpartnersmaybefarmorepower- Such“competitivealtruism”hasbeenthetopicofseveral
ful(Roberts1998;Noe¨ andHammerstein1994). descriptions and studies (Roberts 1998; Barclay and Willer
2007; Hardy and Van Vugt 2006). The latter two studies are
especiallygermanebecausetheymodelandprovidedatathat
SocialSelectionforProsocialTraitsinHumans
demonstratecompetitivealtruisminhumans.Competitiveal-
Thepossibilitythatsocialselectionshapedhumancapacities truismgivesanadvantagewhenextremegenerosityresultsin
for altruism and complex sociality was suggested in West- disproportionate payoffs from pairing with the best partners.
Eberhard’sseminalpublicationonthetopic(1979:228): InBarclayandWiller’sstudyusingaprisoner’s-dilemma-like
task, generosity levels increased dramatically when partici-
Itistemptingtospeculatethattheexplosiveevolutionaryincreasein
pantsknewtheirbehaviorswereobservableandcouldbeused
the proto hominid brain size, which had the appearance of a “run-
byotherschoosingpartners.Theeffectwasrobusteventhough
away”process,wasassociatedwiththeadvantageofintelligencein
the experiment was anonymous. Hardy and Van Vugt also
themaneuveringandplasticityassociatedwithsocialcompetitionin
demonstrated increased altruism when behavior is observed,
primates.
andtheyfoundthatthemostaltruisticindividualsgainedthe
The complexity of the social environment is widely recog- highest status and were preferred as partners, thus gaining
nized as a selection force likely to be important for explain- benefits.
ing human social abilities (Humphrey 1976; Alexander and Theresultingpositivefeedbackprocesscanshapecostly
Borgia1978;Alexander1979;ByrneandWhiten1988).The displays, and preferences for partners who present such dis-
fullimplicationsforhumanprosocialtraitshaveyettobede- plays.Displaysofresources,talentandotherindicatorsofpart-
veloped, although one wide-ranging treatment suggests that ner value are prominent aspects of human cultures (Barkow
socialselectionmayhaveenormousscopeforexplaininghu- 1989;Dunbaretal.1999;Miller2000;SchallerandCrandall
mancapacitiesforartandliterature,aswellascapacitiesfor 2003;Alexander2005).Conspicuousconsumption,frompot-
intelligenceandcooperation(Alexander2005). latchestoRolexes,hasbeeninterpretedaswastefulstatusdis-
Acloselyrelatedmodelfortheevolutionofhumanaltru- plays(Veblen1899),butsuchdisplaysnotonlyenticemates,
ismisbasedonsexualselection.GeoffreyMiller(2000;2007) theyalsoadvertiseanindividual’sdesirabilityasarelationship
148 BiologicalTheory2(2)2007
RandolphM.Nesse
partneroragroupmember.Competitionsinsuchdisplaysre- profitablepartnersexertsapowerfulselectionforcefortrans-
wardonlythemostextremeandremarkableperformancesand action quality and the ability to conceal and detect defection
creations(Veblen1899;Frank1999;Alexander2005). (Frank1988,Trivers2000).Thisshapesmarketefficiencyand
People advertise their reputations as much as their re- integrity,eventotheapparentlymaladaptiveextremeofguar-
sources, and displays of moral character are an equally im- anteesthat“thecustomerisalwaysright.”Suchguaranteesare
pressive aspect of human cultures (Katz 2000). Reputation exploitable and costly, but competition for customers keeps
displaycompetitionsmaybeimportantforexplaininghuman themprevalent.
moral capacities and altruistic behaviors that are not reliably The argument that social selection shapes extreme traits
reciprocated. Recent models suggest that altruism itself may forwinningcompetitionsforrelationshippartnerscanberead-
be an honest advertisement based on the handicap principle ily expanded to encompass parallel processes at the group
(Gintis et al. 2001; Pilot 2005; Hardy and Van Vugt 2006; level. Individuals in groups assess the qualities of potential
BarclayandWiller2007). future members and admit those who offer the most while
Strong reciprocity is closely related (Gintis et al. 2001). demanding the least. Conversely, prospective new members
AsFehrandHenrichputit(2003:57),“Theessentialfeatureof assesswhichgroupoffersthemthemostattheleastcost.The
strongreciprocityisawillingnesstosacrificeresourcesinboth resultisasortingofindividualsbytheirabilitiestocontribute
rewardingfairbehaviorandpunishingunfairbehavior,evenif resources,creatinggroupsreadilyrankedinquality.However,
thisiscostlyandprovidesneitherpresentnorfutureeconomic becausebeingabigfishinasmallpondcanpayoffbetterthan
rewardsforthereciprocator.”Theyarguethatthisapparently beingasmallfishinabigpond,thepartnervalueofmembers
“excess”altruismisnotamistake,butanadaptationthatarises willoverlapbetweengroups(Frank1985).
because even small amounts of conformist transmission give Skewtheory(ReeveandShen2006)mayclarifythedy-
advantages to cooperate-punish strategies that result in their namicsofindividualscompetingforresourcesotherthanac-
spread in cultural groups. The previous argument about the cess to reproduction in social groups. Individuals in groups
Baldwin effect and emergent forces of selection in groups is should value new members proportional to their effects on
similar, but focuses more attention on behaviors at the level group members’ ability to get resources. Potential members
of the individual. Prior work on the evolution of capacities display both their resources and their willingness to share
for commitment (Nesse 2001) is also related, although com- them. After an individual joins a group, the dynamics shift
mitment strategies rely more on intensive communication of tothosebasedonthecostsandbenefitsofallowingamember
threatsandpromises,andwaystomakethembelievableeven tostay,andcompetitionforalliesandpositionwithinagroup.
when fulfilling the commitment would not be in the actor’s Socialselectionfromcompetitionstojointhebestgroupsmay
interests.Asnotedalready,researchoncooperationisvulner- bemorepowerfulthancompetitiontobechosenasanindivid-
able to confusion because it probably is shaped by multiple ualpartner,butthecomplexitiesmakeitwisetofocushereon
selectionpressuresthatarehardtodisentangle. simplerpartnerships.
Theassortmentthatbringscooperatorstogetherneednot It is important to note that the behaviors of individu-
bebasedonrecognitionoridentitytags;simpleenvironmen- als groups can create emergent forces of natural selection in
tal or partner preferences are sufficient (Pepper and Smuts groups that shape otherwise inexplicable traits such as gen-
2002; Pepper 2007). Any mechanism that associates coop- uine altruism, group loyalty and boundaries that define the
erators gives advantages to those with prosocial traits (Wolf in-groupanddevalueout-groups(AlexanderandBorgia1978;
et al. 1999; Frank 2006). The results of such selective asso- BoydandRicherson1985).Suchforcesmayemergereliably
ciation of cooperators can be framed as trait-group selection fromindividualsandpartnershipspursingtheirowninterests.
(WilsonandSober1994),butsuchmodelsareverydifferent Whilesuchemergentselectionforceswouldnotexistwithout
fromoldgroupselection,sotopreventconfusion“analterna- thegroup,theyareverydifferentfromgroupselectioninthat
tive is to state as simply as possible what they are—models theydonotdependonthesuccessofthegroup.
of nonrandom assortment of altruistic genes” (West et al.
2007:11).
Models
Theopportunitytochoosefromavarietyofpartners,and
thepossibilityofnegotiatingcontractsandprices,suggestsap- Mostmodelspartitionfitnesseffectsintosocialselectionand
plyingmarketmodelstotheproblemofcooperation(Noe¨and naturalselectioncomponents,anddescribehowcovariancebe-
Hammerstein1995;Hammerstein2001;Noe¨etal.2001).Con- tweentraitsinassociatedpartnerscanaccountforthestrength
sumersandproducers,whetherhumans,otheranimals,plants ofsocialselection(West-Eberhard1975;West-Eberhard1983;
orfungi,selectamongavailablepartnersbasedupontheirutil- Tanaka 1996; Wolf et al. 1999; Frank 2006). It is difficult,
ity,availability,andprice.Replacementofcheaterswithmore however,forsuchmodelstodescribethedynamicprocessof
BiologicalTheory2(2)2007 149
RunawaySocialSelectionforDisplaysofPartnerValueandAltruism
Correlations of G with C: Fixed Partnerships
1.00 1.01 1.02 1.03 1.04 1.05 1.10
0.80
0.60
0.40
n 0.20
o
ti
a
el 0.00
r
r
o
C -0.20
-0.40
-0.60
-0.80
10 20 30 40 50
Iteration
Figure1.
CorrelationofGwithCforfixedpartnershipsover50iterationsforsixlevelsofR.
choosingrepeatedlyamongmanypossiblepartnersasafunc- availablepartnerwhooffersthebestcombinationofresources
tionofbehaviorsthatchangeovertime. andgenerosity.Moregenerousagentsstillaccumulatecapital
Anagent-basedshared-investmentmodelmayhelptoil- moreslowlythantheirlessgenerouspartners,butthesorting
lustratesomeoftheseprocesses.Asimpleinitialmodelassigns process greatly increases the maximum correlation and how
eachagentarandomlydistributedgenerosityparameter,G,that quicklyitbecomespositive.AsillustratedinFigure2,G×C
ranges from 0.0 to 1.0. Each of 100 agents is endowed with becomespositiveatthe9thiterationifR=1.05,andatthe5th
capital,C=100.Ineachiteration,pairsofagentsinvestaper- iteration if R=1.10. Both continue on to correlations much
centageoftheirtotalresources,(G∗C)and(G(cid:1)∗C(cid:1))respectively higherthaninthemodelwithoutpartnerchoice.
(theprimemarkindicatesthepartner’sparameters).Bothpart- These simple models illustrate how partner choice can
nersreceiveapayoffequaltohalfoftheirtotaljointinvestment shapeincreasedgenerosity.Themodelcouldeasilybeelabo-
timesR,therateofreturn:Payoff =R∗((G∗C)+(G(cid:1)∗C(cid:1)))/2. ratedbyallowingreproductionasafunctionofcapitalaccumu-
Ifthismodelisrunwithoutsorting,agentsremaininfixed lation,orbyusingageneticalgorithmtoseewhatparameters
pairs.TheagentwithahigherGdoesworsebecauseitinvests areoptimalandwhetherdifferentsubtypesofagentsfindevo-
more than the partner on each move, but they share the pay- lutionarilystablealternativestrategies.Suchmodelscouldalso
offsequally.Despitethehigherpayoffsforthelessgenerous userandomnormaldistributionsofRinordertostudythein-
agentineachpair,whenall100agentsareconsidered,more fluenceofstochasticpayoffs.Itwillbeinterestingtodiscover
generousagentsonaveragehavesuperiorpayoffsasreflected theoptimallevelsofgenerosityacrossdifferentlevelsofother
inincreasingcorrelationsofGwithCwitheachiteration.The parameters and whether populations of agents go to a stable
correlation of G with C increases with each iteration. How equilibriumoriftheycycle.Futuremodelsalsoneedtoincor-
fastitbecomespositivedependsonR.AsshowninFigure1, poratethepossibilityofdeception,althoughcontinuingchoice
whenR=1.03,correlationsbecomepositivebyiteration40. amongknownpotentialpartnersmakesdeceptionlessimpor-
ForR=1.05thecorrelationbecomespositiveatiteration25, tantthaninmostreciprocitymodels.Socialselectionmodels
butreachesonly0.40atiteration50.WhenR=1.10,thecor- lendthemselvestoinvestigationsofhowhierarchyinfluences
relation becomes positive at iteration 12 and approaches an cooperation.
asymptoteof0.60.
Model2isthesameexceptthatateachiterationtheagents
TheInvisibleHand
aresortedaccordingtoG×C,thetotalinvestmentmadeonthe
previousmove.Thisincreasinglypairsmoregenerousagents AdamSmith([1759]1976)waspreoccupiedwithfindingex-
as if each one were watching all others and pairing with the planationsforsympathy,andhisfollowersarguethathewould
150 BiologicalTheory2(2)2007
RandolphM.Nesse
Correlations of G with C: Partner Choice
R=1.05 R=1.10
1.2
1
0.8
0.6
n
o
0.4
i
t
a
l 0.2
e
r
r
o 0
C 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
-0.2
-0.4
-0.6
-0.8
Iteration
Figure2.
CorrelationofGwithCgivenpartnerchoiceover50iterationsfortwolevelsofR.
be dissatisfied with current evolutionary theories of altruism guiltandshamewhenwecauseotherspainordisappointment
(Khalil2004).Inhisbookonthemoralpassions,Smithmen- (Gibbard1990).
tionedtheinvisiblehandonlyonce,andthiswaswithrespect
to the division of resources. The idea of the invisible hand
CaveatsandConclusions
seemsequallygermane,however,totheoriginsofmoralemo-
tions.Individualspursuetheirinterestsbytryingtoattractthe Severalcaveatsandlimitationsshouldbekeptinmind.First,as
best possible partners. To succeed, they must offer to fulfill alreadynoted,multiplemechanismsofselectionshapecapaci-
thewishesandexpectationsofpotentialpartnersatthelowest tiesforcooperation.Whilethisarticleemphasizestheeffectsof
possibleprice.Thisusuallyrequirescarryingoutmanyexpen- runawaysocialselectionresultingfromsocialpartnerchoice,
siveactionsthathelpandpleaseothers.Self-interestedpartner several other forces are involved, including sexual selection,
choicemaybetheinvisiblehandthatshapedhumancapacities thebenefitsofmutualisms,andplainreciprocity.
forsympathy. Second, and closely related, the fitness benefits of so-
Socialexchangewithpartnerchoicegivesrisetoemergent cialselectionareintimatelyinvolvedwithreciprocityandkin
forcesofnaturalselectionthatcanshapesocialtraitsfarmore selection.Inonesensethisisnotanissue.Otherdifferentper-
sophisticatedthangenericsympathy.Theseforcesshouldgive spectives,suchasreciprocityandkinselection,canbemodeled
fitness advantages to those who pay close attention to what in a common framework. The social selection perspective is
otherswant,somethingverymuchliketheoryofmind.They distinctive, however, because it shifts the focus of attention
couldalsoshapeempathicconcernforthewelfareofpartners away from decisions to cooperate or defect and abilities to
and strong motives to make reparations, not only for actual detectcheating,andtowardthequitedifferenttasksofselect-
defections, but for even hints of possible lack of attention to ing carefully among a variety of potential partners, trying to
theother’sneeds(WuandAxelrod1995).And,theycanshape discernwhattheywant,andtryingtoprovideit,sooneismore
love,spite,contempt,andthewholerangeofsocialemotions likelytobechosenandkeptasapartner.
(Nesse1990).Mostglobally,thesesocialmarketforcesshape Reciprocity and social selection models of cooperation
desires to please others in general, and desires to avoid any differ not only because they partition fitness effects differ-
cause of displeasure. Indeed, powerful internal mechanisms ently, but also because social selection gives rise to runaway
reward us for helping others (Brown et al. 2003), and cause processesthatcanaccountfortraitsthatdecreasesurvivalor
BiologicalTheory2(2)2007 151
RunawaySocialSelectionforDisplaysofPartnerValueandAltruism
competitiveness,suchasextremealtruism.Forthesamerea- ferent parameters for payoffs and noise, and the possibility
son, the benefits of socially selected traits may come at the that viscosity or other grouping mechanisms may maintain
costofincreasedvulnerabilitytoseriousmentaldisorders.For differentequilibria.
instance, rapid selection for complex social capacities may No definitive experiment is likely to prove the role of
havepushedsometraitsclosetoafitness“cliff-edge”beyond socialselectioninshapinghumancapacitiesforcooperation,
which lies catastrophic cognitive failure of the sort seen in and,forthereasonsjustnoted,cross-speciescomparisonswill
schizophrenia(Nesse2004). not be very useful. Nonetheless, just as a reciprocity models
Socialselectioncallsattentiontothelocusofselection’s suggestedlookingforspecializedcheaterdetectioncapacities,
action:heritablevariationsinsocialtraitsthatinfluenceabil- socialselectionmodelssuggestlookingforspecializedcapac-
ities to get and maintain relationships with preferred social itiesfordeterminingwhatotherswant,formonitoringwhether
partners.Empathy,self-esteem,guilt,anger,andtendenciesto oneispleasingthem,andforpresentingasocialselfthatwill
displaymoraltraitsandtojudgeothersmaybeshapeddirectly makeonedesirableasasocialpartner.Ofcourse,wealready
bysocialselection.Insteadofdescribingastableequilibrium, knowquitealotabouttheoryofmindandtheevolutionofself-
socialselectionfocusesattentiononthedynamicprocessthat esteem (Leary and Baumeister 2000), so to demonstrate that
shapessocialtraits. they were shaped by social selection will require predicting
Third,humannatureisnotunitary.Somepeoplearepro- unnoticedaspectsandlookingtoseeiftheyarethere.
foundly prosocial, others lack all sympathy. Do individuals Insum,partnerchoicecancreaterunawayforcesofsocial
who lack sympathy have a genetic defect? Or, did they miss selection that may have shaped human prosocial tendencies
someearlyexperiencenecessarytodevelopmentofthecapac- andcapacitiesforadvancedsocialcognitionthatareotherwise
ity? Or, is sympathy a facultative trait expressed only in cer- difficulttoexplain.Whetherthisturnsouttobecorrectawaits
tain social circumstances? Or is selection for such capacities additional modeling, experiments, field studies, and further
so recent that gene frequencies are changing rapidly? Or are syntheseswiththeprinciplesofmicroeconomics.
they maintained in some frequency dependent equilibrium?
(Mealey 1995). These are important questions, as yet unan-
Acknowledgments
swered. While finding the mean values and distributions for
Thanksforveryhelpfulcommentsfromtwoanonymousreviewers,andto
anytraitinanyspeciesisvaluable,attemptstoessentializehu-
membersofmylaboratorygroup,andtocolleagueswhoofferedvaluablead-
mannatureareatoddswithbothobservationofhumanvaria- vicealongthewayincludingRobertAxelrod,LeeDugatkin,SteveFrank,Kern
tionandanevolutionaryviewofhowhumannaturecametobe. Reeve, Bobbi Low, Richard Nisbett, Peter Railton, Mary Rigdon, Stephen
Forcesofsocialselectionmayalsovarysignificantlybe- Salant,StephenStearns,BarbaraSmuts,MaryJaneWest-Eberhard,andto
Lucyforinspiration.
tweendifferentgroups(Henrichetal.2005).Evenwithinone
society, different subgroups show different social patterns. It
also seems possible that the benefits of partner choice may References
be much larger in some settings compared to others. For in-
AlcockJ(2001)TheTriumphofSociobiology.NewYork:OxfordUniversity
stance, if most economic activity requires little cooperation
Press.
andnotrading,thenattendingcloselytoother’sneedswillbe AlexanderRD(1979)DarwinismandHumanAffairs.Seattle:Universityof
of little value as compared to a situation in which competi- WashingtonPress.
tivepresentationsofselfinfluencefitnessstrongly.Highrates AlexanderRD(1987)TheBiologyofMoralSystems.NewYork:Aldinede
Gruyter.
of narcissism may be a reliable product of certain social and
AlexanderRD(2005)Evolutionaryselectionandthenatureofhumanity.In:
economicstructures(Lasch1979).
DarwinismandPhilosophy(Ho¨sleV,IlliesC,eds),301–348.NotreDame,
Arelatedconcerniswhethertheopportunitiesforpartner
IN:UniversityofNotreDamePress.
choicehaveinfluencedfitnesslongenoughtocreateforcesof Alexander RD, Borgia G (1978) Group selection, altruism, and the levels
social selection sufficient to shape complex social traits. To of organization of life. Annual Review of Ecology and Systematics 9:
find out this will require anthropological data interpreted in 449–474.
AmundsenT(2000)Whyarefemalebirdsornamented?TrendsinEcology
this framework. The possibility that capacities for profound
andEvolution15:149–155.
sociality arose from culture without influences from natural
AnanthM(2005)Psychologicalaltruismvs.biologicalaltruism:Narrowing
selectionseemsunlikely.Humansclearlyhavesocialcapaci- thegapwiththeBaldwineffect.ActaBiotheoretica53:217–239.
tiesthatarequalitativelydifferentfromotheranimals(Kitcher AnderssonM(1994)SexualSelection.Princeton:PrincetonUniversityPress.
1993;Dunbar1998;Tomasello1999). Armbruster WS, Antonsen L, Pe´labonb C (2005) Phenotypic selection on
Dalechampia blossoms: Honest signaling affects pollination success.
Finally, words hide all manner of imprecision that is re-
Ecology86:3323–3333.
vealed only by transforming them into mathematical state-
AxelrodR(1984)TheEvolutionofCooperation.NewYork:BasicBooks.
ments.Themathematicalmodelsinthispaperarerudimentary.
Axelrod R (1986) An evolutionary approach to norms. American Political
Amongotherfactorsthatneedexplorationaredeception,dif- ScienceReview80:1095–1111.
152 BiologicalTheory2(2)2007