doi:10.1111/j.1420-9101.2006.01258.x MINI REVIEW Social semantics: altruism, cooperation, mutualism, strong reciprocity and group selection S. A. WEST,* A. S. GRIFFIN* & A. GARDNER*(cid:2) *InstituteofEvolutionaryBiology,SchoolofBiologicalSciences,UniversityofEdinburgh,King’sBuildings,Edinburgh,UK (cid:2)DepartmentsofBiologyandMathematics&Statistics,Queen’sUniversity,Kingston,ON,Canada Keywords: Abstract directfitness; From an evolutionary perspective, social behaviours are those which have Hamilton’srule; fitnessconsequencesforboththeindividualthatperformsthebehaviour,and inclusivefitness; another individual. Over the last 43 years, a huge theoretical and empirical kinselection; literaturehasdevelopedonthistopic.However,progressisoftenhinderedby reciprocalaltruism; poorcommunicationbetweenscientists,withdifferentpeopleusingthesame socialevolution; termtomeandifferentthings,ordifferenttermstomeanthesamething.This socialselection. canobscurewhatisbiologicallyimportant,andwhatisnot.Thepotentialfor such semantic confusion is greatest with interdisciplinary research. Our aim hereistoaddressissuesofsemanticconfusionthathavearisenwithresearch ontheproblemofcooperation.Inparticular,we:(i)discussconfusionoverthe terms kin selection, mutualism, mutual benefit, cooperation, altruism, reciprocal altruism, weak altruism, altruistic punishment, strong reciprocity, group selection and direct fitness; (ii) emphasize the need to distinguish between proximate (mechanism) and ultimate (survival value) explanations ofbehaviours.Wedrawexamplesfromallareas,butespeciallyrecentworkon humans andmicrobes. Thedifferenceofapproachispurelydidactic–thereisno large body of theory and a rich experimental literature disagreement about matters of substance Grafen (1984, have arisen on this problem in the wake of Hamilton’s p.82) pioneering papers. To a large extent, semantic confusion should not be Introduction a problem in this field. Several evolutionary theoreti- cians have developed comprehensive overviews of the Scientific progress depends upon precise, reliable com- area, and there is relatively general agreement between munication between scientists (Brown, 1983). This can them (Grafen, 1985; Queller, 1985; Taylor, 1996; be hindered if people use the same term to mean Frank, 1998, 2003; Rousset, 2004; Sachs etal., 2004; different things, or different terms to mean the same Foster & Wenseleers, 2006; Gardner etal., 2007; thing.Intheextreme,thiscanleadtodebatesordisputes Grafen, 2006; Lehmann & Keller, 2006). Furthermore, when in fact there is nodisagreement, or the illusion of several possible points of confusion have long been agreement when there is disagreement. Here, we are solved (Grafen, 1984; Frank, 1998; Reeve & Keller, concerned with issues of semantic confusion that have 1999; Rousset, 2004). However, despite this, semantic arisenwithresearchontheproblemofcooperation.The problems are still managing to creep in from a number problem is why should an individual carry out a of directions. cooperative behaviour that appears costly to perform, Inparticular,severalfactorshaveledtouswritingthis butbenefitsotherindividuals(Hamilton,1963,1964)?A paper.First,extensionsofsocialevolutiontheorytonew taxa seem to allow old semantic problems to reoccur or Correspondence:InstituteofEvolutionaryBiology,SchoolofBiological new variants to arise. Two areas where this is currently Sciences,UniversityofEdinburgh,King’sBuildings,Edinburgh, important,thatweshalldiscuss,areworkoncooperation EH93JT,UK. Tel.:01316505496;fax:01316506564;e-mail:[email protected] inhumansandmicrobes.Otherrecentexamplesinclude ª2006THEAUTHORS20(2007)415–432 JOURNAL COMPILATION ª2006EUROPEANSOCIETYFOREVOLUTIONARY BIOLOGY 415 416 S. A. WEST ET AL. Box.1 Glossary. As stated above, the problem of cooperation is why should an individual carry out a cooperative behaviour Actor:focalindividualwhoperformsabehaviour. that appears costly to perform, but benefits other Altruism:abehaviourwhichiscostlytotheactorandbeneficialtothe individuals (Hamilton, 1963, 1964)? Theoretical expla- recipient;inthiscaseandbelow,costandbenefitaredefinedonthe nations for the evolution of cooperation (or any basisofthelifetimedirectfitnessconsequencesofabehaviour. Cheaters:individualswhodonotcooperate(orcooperatelessthantheir behaviour) can be broadly classified into two categories: fairshare),butarepotentiallyabletogainthebenefitofothers direct fitness benefits or indirect fitness benefits (Fig.1; cooperating. Hamilton, 1964; Brown & Brown, 1981; Grafen, 1984; Cooperation:abehaviourwhichprovidesabenefittoanotherindividual Taylor, 1996; Lehmann & Keller, 2006; West et al., (recipient),andwhichisselectedforbecauseofitsbeneficialeffecton 2006b). This follows from Hamilton’s insight that therecipient. individuals gain inclusive fitness through their impact Directfitness:thecomponentoffitnessgainedthroughtheimpactofan on the reproduction of related individuals (indirect individual’sbehaviourontheproductionofoffspring. fitness effects) as well as directly through their impact Inclusivefitness:‘theeffectofoneindividual’sactionsoneverybody’s on their own reproduction (direct fitness effects) (Ham- numbersofoffspring…weightedbytherelatedness’(Grafen,1984); ilton,1964;Grafen,1984).Thetermsdirectandindirect thesumofdirectandindirectfitness;thequantitymaximizedby Darwinianindividuals. fitness were introduced by Brown & Brown (1981), Indirectfitness:thecomponentoffitnessgainedfromaidingthe although Fisher (1930, chapter 2) discussed indirect reproductionofrelatedindividuals. effects in a similar context. Kinselection:processbywhichtraitsarefavouredbecauseoftheir Thefirstclassofexplanationsforcooperationisthatit beneficialeffectsonthefitnessofrelatives. mayprovideadirectfitnessbenefittotheindividualthat Localgroup:asubsetofthepopulationwhoareinteracting;thelocal performs the behaviour, which outweighs the cost of groupmayvaryfromtheperspectiveofdifferentbehavioursortraits. performing the behaviour (Sachs etal., 2004). One Mutualbenefit:abehaviourwhichisbeneficialtoboththeactorandthe possibility is that individuals have a shared interest in recipient. cooperation.Forexample,inmanycooperativebreeding Mutualism:cooperationbetweenspecies. species,largergroupsizemayprovideabenefittoallthe Neighbour-modulatedfitness:totalpersonalfitness,includingtheeffects ofone’sownbehaviourandthebehavioursofsocialpartners. members of the group through factors such as greater Recipient:anindividualwhoisaffectedbythebehaviourofthefocal survival or higher foraging success – in this case, individual. individuals can be selected to help rear offspring that Relatedness:ameasureofgeneticsimilarity. arenottheirown,inordertoincreasegroupsize(Kokko Selfishness:abehaviourwhichisbeneficialtotheactorandcostlytothe et al., 2001). Another possibility is that there is some recipient. mechanism for enforcing cooperation, by rewarding Spite:abehaviourwhichiscostlytoboththeactorandtherecipient. cooperators or punishing cheaters (Trivers, 1971; Frank, 2003). This could happen in a variety of ways, which have been termed punishment, policing, sanctions, agronomy (reviewed by Denison etal., 2003) and para- reciprocal altruism, indirect (reputation based) recipro- sitology (reviewed by Read, 1994; West etal., 2001, city andstrong reciprocity (see below). 2003). Secondly, serious semantic misunderstandings The second class of explanations for cooperation is canstilloccurintheevolutionaryliterature(e.g.Zahavi, that it provides an indirect benefit because it is directed 2003, 2005; Wilson & Ho¨lldobler, 2005; Wilson, 2005). towards other individuals who carry the cooperative Thirdly, even the leading social evolution theoreticians gene (Hamilton, 1964, 1970, 1975). The easiest and do not use fundamental terminology consistently, as most common way in which this could occur is if genes illustrated by a recent target review on the topic are identical by descent – by helping a close relative (Lehmann & Keller, 2006) and its associated commen- reproduce,anindividualisstillpassingonitsowngenes taries,ora recentedited volume(Hammerstein, 2003). to the next generation, albeit indirectly. Hamilton Ouraimherewastodiscusssomemajortermsthatare (1964) pointed out that this could occur via two oftenmisusedormisunderstood(Box 1).Specifically:(i) mechanisms: (i) kin discrimination, when cooperation kin selection; (ii) mutualism; (iii) cooperation; (iv) is preferentially directed towards relatives; (ii) limited altruism; (v) group selection; (vi) direct fitness (vii) dispersal (population viscosity) keeping relatives distinguishingbetweenultimateandproximateexplana- together, allowing cooperation to be directed indiscrim- tions of behaviour. We then end with the caveat that: inately towards all neighbours (this will be favoured as (viii)classifyingbehaviourswillnotalwaysbetheeasiest those neighbours tend to be relatives). The second way ormost useful thingto do. to obtain an indirect fitness benefit is if cooperation is directed towards nonrelatives who share the same cooperative gene. This assortment or ‘greenbeard’ Social evolution theory mechanism requires a single gene (or a number of Before discussing possible points of confusion, it is tightly linked genes) that both causes the cooperative useful to provide a basic summary of relevant theory. behaviour and can be recognized by other individuals ª2006THE AUTHORS20(2007)415–432 JOURNALCOMPILATIONª2006EUROPEANSOCIETY FOREVOLUTIONARYBIOLOGY Socialsemantics 417 - Fig.1 Aclassificationoftheexplanationsforcooperation.Directbenefitsexplainmutuallybeneficialcooperation,whereasindirectbenefits explainaltruisticcooperation.Withinthesetwofundamentalcategories,thedifferentmechanismscanbeclassifiedinvariousways–herewe followWestetal.(2006;seealsoSachsetal.,2004;Lehmann&Keller,2006).Thesepossibilitiesarenotmutuallyexclusive,forexamplea singleactofcooperationcouldhavebothdirectandindirectfitnessbenefits.Wehavelistedsomeofthemanydifferenttermsthathavebeen usedtodescribethemechanismsforenforcingcooperationtoemphasizethatreciprocity(reciprocalaltruism)isonlyoneofmanywaysto obtaindirectfitnessbenefitsthroughcooperation.Theseenforcementmechanismscanalsoaltertheindirectbenefitsofabehaviour(Lehmann &Keller,2006),anddeterminingtherelationshipsbetweenthesetermsremainsanimportanttask.Kinselectionhasbeenusedtoreferto(i) justthoseindirectbenefitsinvolvingcoancestry(i.e.limiteddispersalandkindiscrimination),or(ii)allindirectbenefits(i.e.alsoincluding greenbeardeffects). due to a distinctive phenotypic marker, such as a green using kin selection wherever genetic relatedness beard (Hamilton, 1964, 1975; Dawkins, 1976; Jansen & between social partners occurs, irrespective of the van Baalen, 2006). An alternative is to conceptualize causes of relatedness. Here, we shall follow the broader greenbeards as a form of kin discrimination. definition, whilst also noting that this distinction is usuallynotimportantbecausekinshipisbyfarthemost common reason for indirect fitness benefits. Kin selection It is also useful to make the distinction between Maynard Smith (1964) coined the term ‘kin selection’ process and maximand. Natural selection is the process to describe how indirect fitness benefits arise from by which fitness is maximized. Inclusive fitness is a helping relatives reproduce. Since then, the phrase kin form of analysis of social traits, and a generalization of selection has been used in two different ways. The Darwinian fitness that takes account of social interac- narrower use of kin selection works upon interactions tions – natural selection leads to organisms acting as if between individuals who are genetically related due to they are maximizing their inclusive fitness (Grafen, common ancestry – i.e. indirect benefits due to limited 2006). Hamilton (1964) did not provide a specific term dispersal or kin discrimination. The broader use of kin for the process by which inclusive fitness was maxim- selection works upon interactions between individuals ized – his aim was to show that natural selection would who share the gene of interest, regardless of whether maximize inclusive fitness. Consequently, the term kin thisisduetocoancestryorsomeothermechanism–i.e. selection appears to have been adopted for this role, of also includes greenbeardeffects. The difference between what provides a generalized description of natural these usages is therefore whether kinship and related- selection on social interactions. The key point here is ness are defined on the basis of average genetic that kin selection is the process by which inclusive similarity over most of the genome (narrow definition), fitnessismaximizedonlyifweareusingkinselectionin or at the particular locus of the behaviour being its broadest sense. In some of his later papers, after the examined (broad definition). Hamilton (1975) favoured term kin selection had been adopted elsewhere, Ham- the narrower definition of kin selection, arguing that ilton (1971, 1972) seems to refer to social selection as inclusivefitness(andHamilton’srule)wasmoregeneral theprocessbywhichinclusivefitnessismaximized,and and should be distinguished from kinship effects. How- distinguishes between classical theory and social theory ever, most people have used the broader definition, (see also Frank, 2006). ª2006THEAUTHORS20(2007)415–432 JOURNAL COMPILATION ª2006EUROPEANSOCIETYFOREVOLUTIONARY BIOLOGY 418 S. A. WEST ET AL. 2004). Indeed, a major question in studies of cooper- Mutualism and mutual benefit ative breeding species is to determine the extent to Fromanevolutionarypointofview,abehaviourissocial which cooperation is explained by direct (+/+) or ifithasfitnessconsequencesforboththeindividualthat indirect ()/+) fitness benefits (Clutton-Brock, 2002; performs that behaviour (the actor) and another indi- Griffin & West, 2002). ‘Helping’ is another possible vidual (the recipient). Hamilton (1964) classified social term for +/+ behaviours, but it suffers from the same behavioursaccordingtowhethertheconsequencesthey problems as cooperation (Rousset, 2004). We return to entailfor the actor and recipient are beneficial (increase the specific definition of cooperation below. directfitness)orcostly(decreasedirectfitness)(Table1). Another possible term for +/+ behaviours is ‘mutual- Whetherabehaviourisbeneficialorcostlyisdefinedon ism’ (Krebs & Davies, 1993; Emlen, 1997; Foster et al., the basis of: (i) the lifetime consequences of the beha- 2001; Clutton-Brock, 2002; Gardner & West, 2004a; viour (i.e. not just the short-term consequences) and Ratnieks,2006).Therearealsoexamplesofpeopleusing (ii) the absolute fitness effect – for example, does it cooperation and mutualism interchangeably (Brown, increase or decrease the actor’s number of offspring 1983; Alcock, 1989). However, this use of mutualism surviving to adulthood (i.e. not just relative to the can be confusing because many also use mutualism to individualsorsocial group withwhich theactor directly refer to the more specific case of cooperation between interacts). In his original papers, Hamilton (1964) pro- species(Wilson,1975b;Herreetal.,1999;Yu,2001;West videdtermsfortwoofthefourpossibilities.Hetermeda et al., 2002b; Kiers etal., 2003; Ridley, 2004; Futuyma, behaviour which is beneficial to the actor and costly to 2005; Foster & Wenseleers, 2006). This is also the therecipient(+/))asselfishness,andabehaviourwhich common use of mutualism in the ecological literature. iscostlytotheactorandbeneficialtotherecipient()/+) The two ideas here are quite distinct, as mutually asaltruism.Later,hetermedabehaviourwhichiscostly beneficial social behaviour is a description of the effect to both the actor and the recipient ()/)) as spite of a single behaviour on the actor and the recipient, (Hamilton, 1970). We do not discuss the semantics of whereas interspecific mutualism describes the impact thetermspiteinthispaper,becausewehavedonesoin thateachpartyhasontheother.Usingthesamewordto detailelsewhere (A.Gardner, I.C.W.Hardy, P.D. Taylor, indicatebothmaycauseconfusion,forexampleitiseasy S.A. West, unpublished data). In this section, we are to see how a mutually beneficial behaviour can evolve, concernedwithdefiningabehaviourwhichisbeneficial butthisdoesnotmeanthatinterspecificmutualismsare toboth theactor andthe recipient (+/+)? soeasilyexplained.Wethinkthatthetermmutualismis One possible term for +/+ behaviours is ‘cooperation’ best reservedfor itsecological usage. (Trivers, 1985; Bourke & Franks, 1995; Rousset, 2004; In order to solve this problem we suggest the term Lehmann & Keller, 2006). However, this use of ‘mutual benefit’ to describe +/+ behaviours. This term cooperationimpliesthatcooperationisalwaysexplained states as simply as possible, and without extra conno- by direct fitness benefits. This can lead to confusion tations, that a behaviour is beneficial to both the actor because others use cooperation more generally for and the recipient. We emphasize that we think rede- behaviours that are beneficial to the recipient, but can fining terms is usually counter-productive, and can add be either beneficial (+/+) or costly ()/+) to the actor more confusion than it solves. However, we think it is (Hamilton, 1972; Axelrod & Hamilton, 1981; Frank, useful in this case, because we already have the 1995b; Maynard Smith & Szathmary, 1995; Sachs et al., situation where multiple terms are being used to define 2004; Foster etal., 2006; West et al., 2006b). Further- the same thing, and those terms also have other uses. more, to use cooperation for only +/+ behaviours, and Wehavefoundtwosimilarusesofmutualbenefit.First, not )/+ behaviours (indirect benefits) contradicts the Maynard Smith (1982, p. 167) appears to use mutual popular use of cooperation in the empirical literature. benefit in the same way as us, commenting that For example, referring to cooperative breeding in cooperation that can be explained by either interactions vertebrates or insects does not necessarily mean that between kin or mutual benefits to cooperating individ- direct fitness benefits are the explanation (Rousset, uals. Secondly, Maynard Smith & Szathmary (1995, p. 261) use it in a similar, but not identical way. They use mutual benefit to describe ‘synergistic’ benefits to Table1 SocialBehaviours.AHamiltonianclassificationschemefor cooperation, and distinguish it from when individuals behavioursthathavebeenselectedforbynaturalselection. force others to cooperate, which they term enforce- ment. This classification can be confusing because: (i) it Effectonrecipient is not clear how they classify scenarios when cooper- + ) ation evolves through punishment or policing, which is enforced, but leads a mutual benefit in the long term; Effectonactor (ii) they seem to classify some cases which rely on kin + MutualBenefit Selfishness selection into their mutual benefit category (e.g. Nowak ) Altruism Spite & May, 1992). ª2006THE AUTHORS20(2007)415–432 JOURNALCOMPILATIONª2006EUROPEANSOCIETY FOREVOLUTIONARYBIOLOGY Socialsemantics 419 historical factors can be importantfor social evolution is Cooperation demonstrated by the example that it can be hard for Wenowconsidertheformaldefinitionfor‘cooperation’. eusocialitytoevolveinspecieswithmultiplemating,due It is extremely useful to have a term that describes to reduced relatedness, but that multiple mating can ‘cooperative behaviours’ that provide a benefit to the evolve in species which are already eusocial (Hamilton, recipient,butcouldbebeneficial(+/+,favouredbydirect 1964; Boomsma & Ratnieks, 1996). We admit that fitness benefit) and/or costly ()/+, favoured by indirect determiningwhethercertain+/+behavioursarecooper- fitnessbenefit)totheactor.Cooperationandhelpingare ation may be hard, but this emphasizes that it is a key both frequently used in this context. For example, as question. discussedabove,atopicofdebateistheextenttowhich Theuseofdistinguishingsocialbehavioursonthebasis cooperative breeding, or helping behaviours in cooper- of the selective forces maintaining them has been ative breeders, can be explained by direct or indirect demonstrated by the analogous distinction made in the fitnessbenefits.Consequently,onepossibilityistodefine communication literature between a ‘signal’ and a ‘cue’ ‘cooperation’asabehaviourthatprovidesabenefittothe (MaynardSmith&Harper,2003).Asignalis‘anyactor recipient – this therefore includes both +/+ and )/+ structurewhichaltersthebehaviourofotherorganisms, behaviours (Sachs etal.,2004). which evolved because of that effect, and which is However,thisdefinitionofcooperationmaybeoverly effective because the receiver’s response has also inclusive. For example, when an elephant produces evolved’, whereas a cue is ‘a feature of the world, dung,thisisbeneficialtotheelephant(emptyingwaste), animateorinanimate,thatcanbeusedbyananimalasa butalsobeneficialtoadungbeetlethatcomesalongand guidetofutureaction’(MaynardSmith&Harper,2003). uses that dung. It does not seem useful to term behav- This distinction has been extremely useful, allowing a iours such as this, which provide a one-way byproduct clear and general conceptual overview to be developed. benefit, as cooperation. Consequently, we prefer that a Although,wenotethatthisdefinitionofasignaldoesnot behaviourisonlyclassedascooperationifthatbehaviour excludeactionsthatoperatebecauseoftheirsubstantive is selected for because of its beneficial effect on the effects rather than their information content, so for recipient.Wedonotwishtoimplythatthebehaviouris example,itcouldincludereciprocity,wherecooperation selected for purely because of its beneficial effect on the is conditional upon the cooperative behaviour of others recipient,justthatithasatleastpartiallydoneso.Thisis (A. Grafen, Personal Communication; see also Grafen, easily illustrated with an example. Suppose that two 1990). bacterialspecies(AandB)areinteracting,andthateach feedsuponawasteproductoftheother.Thiswouldbea Altruism mutually beneficial (+/+) behaviour, but we would not class it as cooperation. However, now suppose that a The above sections have emphasized how terms such as higher production of species A’s waste product evolved altruismhaveveryspecificmeanings,whichcanconvey because this benefited species B, and hence led to a usefulinformation.Consequently,whenthesetermsare higherlevelofwasteproductionbyspeciesB,whichwas misused, or redefined, it can lead to confusion. In this beneficialtospeciesA.Thiskindofinteractionhasbeen sectionwediscussthreecasesinwhichthishasoccurred termed byproduct reciprocity or invested benefits (Con- with the term altruism. A general point here is that ner,1995a;Sachsetal.,2004),andwewouldclassthisas altruism is defined: (i) with respect to the lifetime cooperation. Our definition of cooperation therefore consequences of a behaviour; (ii) on absolute fitness includesallaltruistic()/+)andsomemutuallybeneficial effects(i.e.doesitincreaseordecreasetheactor’sfitness, (+/+) behaviours. and not relative to just some subset of the population). This distinction over whether a trait is favoured for Forexample,ifacooperativebehaviourwascostlyinthe that purpose relates to the standard text book definition short term, but provided some long-term (future) bene- of adaptation (Rose & Lauder, 1996). A common fit, which outweighed that, it would be mutually bene- definition for an adaptation is a ‘trait that enhances ficial and not altruistic. This does not mean that such a fitness and that arose historically as a result of natural behaviour is somehow less interesting – determining selection for its current role’ (Rose & Lauder, 1996). A whetherandhowacooperativebehaviourprovidesshort difference here is that our definition of cooperation or long term direct fitness benefits remains a major considers the selective forces maintaining the trait, and problem (Clutton-Brock, 2002; Griffin & West, 2002). not just those that led to its initial evolution – this is Instead, our aim is to distinguish the fundamental sometimes termed ‘aptation’, as opposed to adaptation difference between direct andindirect fitnessbenefits. (Rose & Lauder, 1996). Although we suspect this distinctionwillbeunimportantformostrealcases,there Reciprocal altruism arefieldsinwhichitisimportant,suchasdistinguishing between the evolution and maintenance of sex and Trivers (1971) suggested that cooperation could be recombination (West etal., 1999; Burt, 2000). That favouredbetweennonrelatives,inreciprocalinteractions. ª2006THEAUTHORS20(2007)415–432 JOURNAL COMPILATION ª2006EUROPEANSOCIETYFOREVOLUTIONARY BIOLOGY 420 S. A. WEST ET AL. Theideahereisthatindividualscantaketurnsinhelping formofcooperation’–havingpreviouslydefinedcooper- each other, for example by preferentially aiding others ation as a +/+ behaviour and altruism as a )/+ who have helped them in the past. Trivers termed this behaviour), and at times uses phrases which do not ‘reciprocalaltruism’.Thisworkwashighlyinfluentialin invoke altruism, such as ‘return effects’, ‘return-benefit’ showing that cooperation could be favoured between and‘reciprocity’(p.47).However,altruismisstillusedin nonrelatives, and stimulated a huge amount of theoret- other ways at other times, in a manner that is based on ical and empirical research. Current theoretical over- short term rather than long term cost and benefits. For views place reciprocal altruism as just one of many example: ‘In effect, two individuals trade altruistic acts. mechanisms by which cooperation between unrelated This can be called reciprocity or reciprocal altruism’ (p. individuals can be favoured (Krebs & Davies, 1993; 48), and by referring to kinship and reciprocity as two Griffin & West, 2002; Frank, 2003; Sachs etal., 2004; ways to explain altruism(p. 49). Lehmann & Keller, 2006; West etal., 2006b). Further- more,ithasbeensuggestedthatthespecificmechanism Weak altruism of reciprocity is unlikely to be of general importance outside of humans, because the conditions required can Thedifferentuseofthewordaltruisminthe‘new’group beextremelyrestrictive(Conner,1995b;Dugatkin,1997; selectionliteraturehasalsoledtoconfusion.Wilsonand Clutton-Brock, 2002; Hammerstein, 2003; Stevens & colleagues (Wilson, 1975a, 1977; Colwell, 1981) rede- Hauser,2004;Stevensetal.,2005). fined altruism to refer to the fitness of an individual However, reciprocal altruism is not altruistic – it relative to the individuals that it interacts with, in its provides a direct fitness advantage to cooperating. If an group.Theydefineabehaviouras‘weaklyaltruistic’ifit individual does not pay the cost of cooperation in the leads to a decrease in the fitness of the focal individual, shorttermthenitwillnotgainthebenefitofcooperation relative to the other members of its group. This means in the long term (although things could get more that behaviours which provide a benefit to everyone complicated if reciprocity was between relatives). Con- within the local group, including the actor, such as the sequently, following Hamilton’s original scheme, it is a productionof apublicgood,can bedefinedas altruistic. mutuallybeneficial(+/+)behaviourandnotanaltruistic These are sometimes termed ‘whole-group’ (Pepper, behaviour ()/+). It is presumably for this reason that 2000) or ‘group beneficial’ traits (Dugatkin etal., 2003, Hamilton(1996,p.263)thoughtthatreciprocalaltruism 2005) – as opposed to ‘other-only’ traits which do not was misnamed, and that he and others have used provide a benefit to the actor (Pepper, 2000). Whole- alternative terms such as ‘reciprocity’ (Alexander, group behaviours can have both a direct and indirect 1974), or ‘reciprocal cooperation’ (Axelrod & Hamilton, benefit,andasdiscussedabove,thismeansthatwhether 1981). Unfortunately, the term reciprocal altruism has they are altruistic or mutually beneficial will depend been in use so long that we do not expect its use be upon the relative cost and benefits of the behaviour, as changed, although it would be preferable to use well as population structure (Pepper, 2000; Rousset, reciprocity or reciprocal cooperation. It is not clear how 2004).Thisleadstotheconfusingsituationwhereatrait muchconfusionthisdifferentuseofaltruismhasledto– could be favoured because it selfishly increases an althoughwesuspectitisatleastpartiallyresponsiblefor: individual’s direct fitness, but will be weakly altruistic (i) the frequent and incorrect assumption that kin by Wilson’s definition (Dawkins, 1979; Grafen, 1984; selection and reciprocal altruism are the two leading Harvey etal.,1985)! This alsoemphasizes that there is a explanations for cooperation or altruism and (ii) the fundamental difference between altruism ()/+) and confusing use of altruism in the human literature (see weak altruism (which can be +/+ or )/+), in contrast to below). the suggestion ofFletcher& Doebeli(2006). Trivers (1971) originally redefined altruism in a We illustrate this point in Table2, with the simplest different way to Hamilton. Specifically, he defined it possible case, where groups consist of only two individ- with respect to inclusive fitness, apparently in the short uals. We compare the fitness of cooperators (C) who term, rather than direct fitness in the long term: perform some cooperative behaviour, and defectors (D) ‘Altruistic behavior can be defined as a behavior that whodonot.Thecooperativebehaviourisassumedtobe benefits another organism, not closely related, while costly to the individual who performs it (cost¼ x), but beingapparentlydetrimentaltotheorganismperforming provide a benefit to all the members in the group. We the behavior, benefit and detriment being defined in assumethatthebenefittocostratioisthree,andsoeach termsofcontributiontoinclusivefitness’(Trivers,1971). cooperativebehaviourbringsabenefitof3xtothegroup, Itisclearthatthisformof‘altruism’couldnotbeselected and hence 3x/2 to each individual. This cooperative for, unless benefit and cost are only measured in the behaviourwouldbeclassedas‘weaklyaltruistic’.Table2 short term, and that there is some longer-term benefit illustratesthatfromtheselfishperspectiveofanindivid- (i.e. it is mutually beneficial and not altruistic). In his ual, C always leads to a higher fitness, irrespective of later book, Trivers (1985) returns to Hamilton’s defini- whetheritisinagroupwithaCoraD.Thisshowsthat tions associated with Table1 (‘the altruism becomes a weakly altruistic behaviours can be selected for because ª2006THE AUTHORS20(2007)415–432 JOURNALCOMPILATIONª2006EUROPEANSOCIETY FOREVOLUTIONARYBIOLOGY Socialsemantics 421 Table2 Thefitnessofcooperativeindividualswhoperforma‘weaklyaltruistic’trait(C),anddefectorswhodonot(D). Nocoopera- Group Twocooperators Onecooperator tors TypeofIndividual C C C D D D BaselineFitness 1 1 1 1 1 1 Individualcostofcooperating x x x 0 0 0 Benefitofcooperation(sharedwithingroup) 6x ¼ 3x 3x 3x 3x 0 0 2 2 2 Benefit–cost 3x)x¼2x 2x 3x(cid:2)x ¼ x 3x 0 0 2 2 2 Fitness 1+2x 1+2x 1þx 1þ3x 1 1 2 2 Thecalculationassumesthatthecooperators(C)investxresourcesincooperation,thebenefittocostratioisthree,andthatbenefitsareshared amongstallgroupmembers.Fromtheselfishperspectiveofanindividual,Calwaysleadstoahigherfitness,irrespectiveofwhetheritisina groupwithaCoraD.Thisshowsthatabehaviourwhichwouldbeclassedasweaklyaltruisticcanbeselectedforbecauseitincreasesan individua’lsdirectfitness.Thistablewasinspiredbyananalogousone,inthesexratioliterature(Harveyetal.,1985). theyincreaseanindividual’sdirectfitness,andarehence Boyd etal., 2003) – this is the same as weak altruism mutually beneficial. More generally, the fundamental described above. In both cases we are left with the point is that the spread of a gene is determined by its possibility that cooperation can be labelled as altruistic, fitness‘relativetoothersinthebreedingpopulation,and evenwhenitprovidesadirectbenefitthatcanoutweigh nottootherswithwhichithappenstointeract’(Grafen, its cost, and hence can be selected for from selfish 1984, 2002, 2006; Harvey etal., 1985). A more compli- individualinterests(i.e.mutuallybeneficialcooperation, catedcaseillustratingthesamepointisgiveninTable3. witha+/+payoff).Wefocusontherelevanttheoretical literatureinthissection,andwillreturntotheempirical literature later. Altruism inhumans andstrong reciprocity The usage of altruism in the human cooperation Altruismhasbeenredefinedinanumberofwaysinthe theoretical literature is best illustrated by examining a literature on cooperation in humans. One approach in specific model. Gintis (2000) compared the relative theempiricalsideoftheliteraturehasbeentodescribea fitness of two different strategies: ‘self-interested agents’ costly behaviour as altruistic if it benefits another who do not punish or cooperate and altruistic ‘strong individual, but then to seemingly only measure the cost reciprocators’ who cooperate and punish noncoopera- to the actor over the short term or relative to who they tors. He labels strong reciprocators as altruistic because interact with (Fehr & Fischbacher, 2003). This clearly they ‘increase the fitness of unrelated individuals at a includesthepossibilityforeithermutuallybeneficial(+/ costtothemselves’.However,inthisandrelatedmodels, +)oraltruistic()/+)behaviours,andisanalogoustothe cooperationisindividuallycostlywithinthesocialgroup, definition of cooperation (or helping). In contrast, the but provides a benefit to all the members of the group approachtakenintherelevanttheoreticalliterature,isto (including the cooperative individual), through mecha- usealtruismforbehaviourswhicharecostlybutprovide nismssuchasincreasedproductivityorreducingtherate a benefit to all individuals in the group (Gintis, 2000; ofgroupextinction(Gintis,2000;Henrich&Boyd,2001; Table3 Therelativefitnessofanindividuals Onecooperator,and whocooperates(C)anddefectorswhodonot Group N-1defectors Ndefectors (D),whenmorecooperativegroupsaremore likelytosurviveandreproduce. Typeofindividual C D D Baselineresources 1 1 1 Publicgoodcontribution 1 0 0 Benefittogroup m 0 0 Individualresourcesafterpublicgoodsgame m 1þm 1 N N TotalGroupResources m+N)1 N ProportionofGroupResources m=N 1þm=N 1 mþN(cid:2)1 mþN(cid:2)1 N Relativereproductivesuccessofgroups 1 P FitnessofIndividuals m=N 1þm=N P mþN(cid:2)1 mþN(cid:2)1 N WecomparethefitnessofacooperatorinagroupwithN)1defectors,withthefitnessofa defectorinagroupwithNotherdefectors.Seemaintextfordetails.Undercertainconditions, fromtheselfishperspectiveofanindividual,Cleadstoahigherfitness.Thisshowsthata behaviourwhichsomewouldclassasaltruisticandnotintheindividualsselfinterest,canbe selectedforbecauseitincreasesanindividual’sdirectfitness,andhenceisactuallymutually beneficial. ª2006THEAUTHORS20(2007)415–432 JOURNAL COMPILATION ª2006EUROPEANSOCIETYFOREVOLUTIONARY BIOLOGY 422 S. A. WEST ET AL. Table4 Therelationshipbetweensomeofthetermsthatwehave adefectorinagroupofNdefectors.Individualshaveone discussed.Thedifferentcategoriescanbegroupedaccordingto unitofresourcetocontributetothepublicgoodsgame– whethertheyhavebeneficial(+)orcostly())fitnessconsequences cooperators contribute this whole unit, defectors con- fortheactorandrecipient. tributenothing.Foreachunitinvestedinthegroup,the whole group obtains a return of m units of resources to Consequences Consequences Term foractor forrecipient shareequallyamongstthegroup(cooperationprovidesa group benefit: m> 1). Consequently, as long as m <N, Altruism ) + the cooperator ends up with less resources than they Reciprocalaltruism + + startedwith(comparedwithTable2,wherem >N).We Weakaltruism )or+ + then assume that the group with greater resources is Strongreciprocity )or+ )or+ morelikelytosurviveorreproduce,suchthatthegroup Cooperation )or+ + withgreaterresourceshasarelativereproductivesuccess Mutualbenefit + + of 1.0, and the other group has a relative reproductive successofP(P< 1).Iftherelativefitnessofindividualsis Bowles et al., 2003; Boyd etal., 2003; Gintis, 2003; their share of the group resources multiplied by the Bowles & Gintis, 2004). Consequently, cooperation can groupproductivity,thenthefitnessofthecooperatorwill provideadirectfitnessbenefit,aswellasthepotentialfor begreaterthanthatofadefectorinagroupofNdefectors indirect benefits due to individuals who share the if m > P(cid:2)NP. This demonstrates that cooperation can be P(cid:2)1 cooperativegene. mutually beneficial, and hence favoured by selfish The problem here is that the definition of altruism is interests (a direct fitness benefit), even in cases when relative to the local group, and not the population as a playing the public goods game leads to the cooperator whole (as with ‘weak altruism’). As discussed above, having less resources (m< N), and a lower fitness than naturalselectionselectsforageneifitcausesabehaviour the other members of the group ( m=N < 1þm=N ). mþðN(cid:2)1Þ mþðN(cid:2)1Þ that leads to that gene increasing in frequency in the This is because the benefit at the level of the group, of population,notsomeotherarbitrarilydefinedscalesuch having just one individual cooperate, can outweigh this as social partners (Grafen, 1984) (see Table 2). Another cost, and allow a cooperator to invade a population of way of looking at this problem is that by examining defectors. altruism relative to the group this means that any Punishment can be selfish or altruistic (like cooper- benefits of the behaviour which are equally spread ation) or even spiteful, and so without detailed analysis throughoutthegroupareignored–traitswhichprovide ofparticularsituations,thewordpunishmentshouldnot abenefitatthegrouplevelwillthereforeseemaltruistic be given a prefix such as ‘altruistic’. Punishment may because the benefits are ignored. Consequently, the provideadirectbenefitbecauseitcancausehigherlevels model of Gintis (2000) leads to the confusing situation of cooperation within the punisher’s group (Gardner & wherecooperationcanbefavouredbecauseitprovidesa West,2004b;Lehmann&Keller,2006).Ifthisisthecase direct benefit to the cooperator, because it increases the in the models discussed here, then ‘selfish punishment’ chancetheyandtherestoftheirgroupsurvive,butthat would have been the appropriate term, rather than this is defined as altruistic and not in their self interest. ‘altruisticpunishment’.Thealternativepossibilityisthat The same issue reoccurs in a number of related models punishment is favoured because it leads to an indirect (Bowles etal., 2003; Boyd et al., 2003; Gintis, 2003; fitness benefit, by: (i) making the punished individual Bowles & Gintis, 2004). The potential direct fitness morelikelytocooperatewithrelativesofthepunisher– benefit of cooperation is illustrated clearly by the model altruisticpunishment(Gardner&West,2004b;Lehmann ofBoydetal.(2003),wheregroupscompeteforterritor- &Keller,2006);or(ii)reducingthefitnessofindividuals ies in pairs. In their model, the territory is won by the who are competing with relatives – spiteful punishment groupwiththemostcooperators,andsoitisclearthata (Gardner & West, 2004b). In the human models single individual could potentially gain a huge direct discussed above, the simulation approaches used mean fitness advantage by cooperating, and hence making its that the relative importance of direct and indirect group much more successful. The role of direct fitness selection in favouring punishment are not clear. benefits in this model is further emphasized by the fact A detailed analysis of this problem would be extremely that it assumes that social behaviours are culturally useful (Gardner & West, 2004b; Lehmann & Keller, transmitted, by imitation on the basis that they benefit 2006). theindividual. We illustrate how cooperation can provide a direct More altruism? fitnessbenefitinsuchmodelsinTable3.Weassumethat groupsofNindividualsplayapublicgoodsgamewithin Wehavenotexhaustedtheredefinitionsofaltruism.For groups, and that the more productive groups are more example, altruism has also been to used to describe likelytosurviveorreproduce.Wecomparethefitnessof behavioursthatbenefitotherindividuals(e.g.Kaushik& acooperatorinagroupwithN )1defectors,withthatof Nanjundiah, 2003; Zahavi, 2003, 2005) – this could ª2006THE AUTHORS20(2007)415–432 JOURNALCOMPILATIONª2006EUROPEANSOCIETY FOREVOLUTIONARYBIOLOGY Socialsemantics 423 clearly include behaviours which are altruistic ()/+) or Maynard Smith (1976) showed that group selection mutually beneficial (+/+). However, we think that the would not work if the number of individuals who exampleswehavegivenabovearesufficienttoillustrate disperse and reproduce elsewhere (successful migrants) thegeneralpoints.Inparticular,thatredefiningaltruism is greater then one per group. Empirical work showed can obscure, and lead to confusion about the selective that individuals were reproducing at the rate that forces at work. We stress that by pointing out when maximized theirlifetimereproductive success, andwere behaviours are not altruistic, we do not hope to imply notpractisingreproductiverestraint(Lack,1966;Krebs& they are less interesting or easier to explain. Indeed, Davies,1993).Itisthisformofgroupselectionthatleads demonstrating direct long-term fitness benefits of a peopletothefalseconclusionthatindividualsbehavefor cooperativebehaviourcanbeharderthandemonstrating the good ofthe population orspeciesorecosystem. kin selected benefits (Griffin & West, 2002, 2003). Inthe1970sand1980sanewformofgroupselection Explaining cooperation remains one of the greatest was developed, based on a different conception of the challenges for evolutionary biology, irrespective of group (Wilson, 1975a, 1977; Colwell, 1981; Wilson & whether itisaltruisticormutually beneficial. Colwell,1981).Theideaherewasthatatacertainstages of an organism’s life cycle, interactions take place between only a small number of individuals. It can be Group selection shown that under these conditions, cooperative beha- The group selection literature has generated a huge viour can be favoured. This ‘new group selection’ is amount of semantic confusion (reviewed by Dawkins, sometimesreferredtoas‘trait-groupselection’or‘demic 1979;MaynardSmith,1983;Grafen,1984;Trivers,1998; selection’ or ‘intrademic selection’. One way of concep- Foster etal., 2006). Although this debate was solved tualizing the difference between the old and new group decisively during the 1960s to 1980s, by evolutionary selection models is that the new group selection models biologists, it seems to reoccur and lead to confusion as rely on within-population (intrademic) group selection, new fields embrace the relevant aspects of social evolu- whereasoldgroupselectiontheoryworkedonbetween- tion theory (Reeve & Keller, 1999). We have already population(interdemic)groupselection(Fig.2;Reeve& discussed the phrase ‘weak altruism’. In this section we Keller, 1999). Another difference is that the old group brieflydiscusssomeoftheotherpointsofconfusionthat selection approach argued that selection at the group have arisen. An interested reader is directed elsewhere level was the driving force of natural selection, whereas for general reviews (Grafen, 1984; Dugatkin & Reeve, the new group selection emphasizes that there are 1994),ormoretechnicalsummaries(Wade,1985;Frank, multiple levels of selection, and these can vary in their 1986a; Queller, 1992; Gardner etal., 2007). Readers importance. Another way of looking at this is that the familiar with thisworkcould easilyskip thissection. new group selection approach looks at the evolution of individual characters in a group structured population, whereas the old group selection approach looks at the The oldand the new evolution of groupcharacters (Fig.2;Okasha,2005). Before discussing the semantic issues, it is useful to It has since been shown that kin selection and new distinguish between two different types of group selec- groupselectionarejustdifferentwaysofconceptualizing tion,andexplaintheirrelationtosocialevolutiontheory thesameevolutionaryprocess.Theyaremathematically moregenerally(Grafen,1984;Trivers,1998).Duringthe identical, and hence are both valid (Hamilton, 1975; 1960s, Wynne-Edwards (1962) argued for the import- Grafen, 1984; Wade, 1985; Frank, 1986a, 1998; Taylor, ance of group selection in its original or ‘old’ form. He 1990; Queller, 1992; Bourke & Franks, 1995; Gardner considered relatively cooperative behaviours such as etal., 2007). New group selection models show that reproductive constraint, as follows. In groups consisting cooperation is favoured when the response to between- of selfish individuals (who reproduce at the maximum group selection outweighs the response to within-group rate), resources would be over exploited, and the group selection, but it is straightforward to recover Hamilton’s would go extinct. In contrast, groups consisting of rulefromthis.Bothapproachestellusthatincreasingthe cooperative individuals who restricted their birth rate group benefits and reducing the individual cost favours would not over exploit their resources, and not go cooperation. Similarly, group selection tells us that extinct. Hence, by a process of differential survival of cooperation is favoured if we increase the proportion of groups, behaviour evolved that was for the good of the genetic variance that is between-group as opposed to group. within-group, but that is exactly equivalent to saying Duringthe1960sand1970salargebodyoftheoretical that the kin selection coefficient of relatedness is andempiricalworkwaspiledupagainstthisidea.Theory increased (Frank, 1995a). In all cases where both showedthatthistypeofgroupselectionwouldonlywork methods have been used to look at the same problem, underextremelyrestrictiveconditions,andsoitsimport- they give identical results (Frank, 1986a; Bourke & ance would be rare or nonexistent (Maynard Smith, Franks, 1995; Wenseleers etal., 2004; Gardner etal., 1964, 1976; Williams, 1966; Leigh, 1983). For example, 2007). ª2006THEAUTHORS20(2007)415–432 JOURNAL COMPILATION ª2006EUROPEANSOCIETYFOREVOLUTIONARY BIOLOGY 424 S. A. WEST ET AL. A B Fig.2 Thedifferencebetweenoldandnewgroupselection.PanelAshowstheoldgroupselection,withwell-definedgroupswithlittlegene flowbetweenthem(solidoutline).Thewhitecirclesrepresentcooperators,whereasthegreycirclesrepresentselfishindividualswhodonot cooperate.Competitionandreproductionisbetweengroups.Thegroupswithmorecooperatorsdobetter,butselfishindividualscanspread withingroups.PanelBshowsthenewgroupselection,witharbitrarilydefinedgroups(dashedlines),andthepotentialformoregeneflow betweenthem.Thedifferentgroupsmakedifferentcontributionstothesamereproductivepool(althoughthereisalsothepossibilityoffactors suchaslimiteddispersalleadingtomorestructuring),fromwhichnewgroupsareformed. 1997, 1998;Taylor etal.,2007); (ii)in someofthe most Avoidingconfusion successfulareasofsocialevolution,suchassplitsexratios In addition to the problems associated with the term insocialinsectsorextensionsofHamilton’s(1967)basic ‘weak altruism’, the group selection literature has pro- localmatecompetitiontheory,predictionsariseelegantly duced three other sources of semantic confusion. First, from kin-selection models, whereas the corresponding the different types of group selection can be mixed up group selection models would be either unfeasible or so (Grafen, 1984; Trivers, 1998; Okasha, 2005). This can complex that they have not been developed (Frank, give the impression that the validity of the new group 1986b, 1998; Boomsma & Grafen, 1991; Queller, 2004; selection justifies the application of the old group Shukeretal.,2005);(iii)kinselectionmethodologiescan selection(Trivers,1998).Thiscanbeaparticularproblem usuallybelinkedmoreclearlytoempiricalresearch,both with nonspecialists or nontheoreticians, and we believe empirically (Queller & Goodnight, 1989) and concep- that it plays a major role in explaining why old group tually–‘knowingthatr ¼0.22givesmanybiologistsan selection lingers in some fields, such as areas of micro- understanding of the genetic closeness described; the biology (e.g. Shapiro & Dworkin, 1997; Shapiro, 1998; knowledgethatn¼10andv/v ¼ 2.98is(atleastforthe b Bassler, 2002; Henke & Bassler, 2004), parasitology present)lessilluminating’(Grafen,1984);(iv)thegroup (reviewed by West etal., 2001, 2003), and agronomy selectionapproachtendstohidethedistinctionbetween (reviewed by Denison etal., 2003). Numerous examples direct and indirect benefits of behaviours, which many of this problem are also provided by Sober & Wilson find extremely useful; (v) the group selection approach (1998), who switch confusingly between old and new hasbeenlessusefulforidentifyingandquantifyingissues groupselection (Trivers,1998). of reproductive conflict, which have provided some of The potential for confusion is increased when the the most useful areas for empirical testing of theory impression is gained that evolutionary biologists do not (Trivers,1974;Ratnieksetal.,2006);(vi)inclusivefitness use the new group selection methodology because they theoryleadstotherecoveryofamaximizingprinciplefor think it is unimportant (as they do with the old). In social settings (individuals should behave as if maxim- reality,thekinselection(orinclusivefitness)approachis izing theirinclusive fitness), which isa useful reasoning generallypreferredovernewgroupselectionbecauseitis tool (it is easier to think of individuals optimizing usually easier to construct models, interpret the predic- something rather than the evolutionary dynamics), tions, and then apply these to real biological cases. For provides formal justification for the use of intentional example: (i) recent methodological advances mean that language (selfishness, altruism, conflict), and legitimizes kin selection and inclusive fitness models can be discussion of ‘function’ and ‘design’ ofsocial behaviours constructed and analysed much more simply, and for (Hamilton, 1964; Grafen, 1999, 2006) – whereas the much more general cases (Taylor & Frank, 1996; Frank, group selection view does not lead so easily to a ª2006THE AUTHORS20(2007)415–432 JOURNALCOMPILATIONª2006EUROPEANSOCIETY FOREVOLUTIONARYBIOLOGY