Rethinking Sociobiology

7/28/2019 Rethinking Sociobiology

1/23

Volume 82, No. 4 December 2007

327

The Quarterly Review of Biology, December 2007, Vol. 82, No. 4

Copyright 2007 by The University of Chicago. All rights reserved.

0033-5770/2007/8204-0001$15.00

The Quarterly Review

ofBiology

RETHINKING THE THEORETICAL FOUNDATIONOF SOCIOBIOLOGY

David Sloan WilsonDepartments of Biology and Anthropology, Binghamton University

Binghamton, New York 13903 USA

e-mail: [email protected]

Edward O. Wilson

Museum of Comparative Zoology, Harvard University

Cambridge, Massachusetts 02138 USA

keywordsaltruism, cooperation, eusociality, group selection, human evolution,

inclusive fitness theory, kin selection, major transitions, multilevel selection,

pluralism, sociobiology

abstractCurrent sociobiology is in theoretical disarray, with a diversity of frameworks that are poorlyrelated

to each other. Part of the problem is a reluctance to revisit the pivotal events that took place during the

1960s, including the rejection of group selection and the development of alternative theoretical frame-

works to explain the evolution of cooperative and altruistic behaviors. In this article, we take a back

to basics approach, explaining what group selection is, why its rejection was regarded as so important,

and how it has been revived based on a more careful formulation and subsequent research. Multilevel

selection theory (including group selection) provides an elegant theoretical foundation for sociobiology

in the future, once its turbulent past is appropriately understood.

D

ARWIN perceived a fundamental prob-lem of social life and its potential solu-

tion in the following famous passage from De-scent of Man(1871:166):

It must not be forgotten that although ahigh standard of morality gives but aslight or no advantage to each individualman and his children over the other men


2/23

328 Volume 82THE QUARTERLY REVIEW OF BIOLOGY

of the same tribe . . . an increase in thenumber of well-endowed men and an ad-vancement in the standard of moralitywill certainly give an immense advantageto one tribe over another.

The problem is that for a social group to func-tion as an adaptive unit, its members must dothings for each other. Yet, these group-advan-tageous behaviors seldom maximize relativefitness within the social group. The solution,according to Darwin, is that natural selectiontakes place at more than one level of the bio-logical hierarchy. Selfish individuals mightout-compete altruists within groups, but in-ternally altruistic groups out-compete selfishgroups. This is the essential logic of what hasbecome known as multilevel selection theory.

Darwins insight would seem to provide anelegant theoretical foundation for sociobiol-ogy, but that is not what happened, as anyonefamiliar with the subject knows. Instead,group selection was widely rejected in the1960s and other theoretical frameworks weredeveloped to explain the evolution of altru-ism and cooperation in more individualisticterms. The following passage from George CWilliamss book, Adaptation and Natural Selec-tion (1966:9293), illustrates the tenor of thetimes, which seemed to make the rejection ofgroup selection a pivotal event in the historyof evolutionary thought:

It is universally conceded by those whohave seriously concerned themselveswith this problem . . . that such group-related adaptations must be attributed tothe natural selection of alternative groupsof individuals and that the natural selec-tion of alternative alleles within popula-tions will be opposed to this develop-ment. I am in entire agreement with thereasoning behind this conclusion. Onlyby a theory of between-group selectioncould we achieve a scientific explanationof group-related adaptations. However, I

would question one of the premises onwhich the reasoning is based. Chapters 5to 8 will be primarily a defense of thethesis that group-related adaptations donot, in fact, exist. A group in this discus-sion should be understood to meansomething other than a family and to be

composed of individuals that need notbe closely related.

Forty years later, this clarity has been lost.

In the current sociobiological literature, it iseasy to find the following contradictory posi-tions, side by side in the same journals andbookshelves:

Nothing has changed since the 1960s. Multilevel selection theory (including

group selection) has been fully revived. There is a new multilevel selection the-

ory that bears little relationship to theold theory.

Group selection is not mentioned, as if itnever existed in the history of evolution-ary thought.

Part of this confusion can be explained interms of the diffusion of knowledge. It takestime for the newest developments in theo-retical biology to reach scientists who conductempirical research, and longer still to reachdiverse audiences who receive their informa-tion third, fourth, and fifth hand. However,part of the confusion continues to exist at thehighest level of scientific discourse, as we willshow.

We think that sociobiologys theoreticalfoundation can be as clear today as it ap-peared to be in the 1960s, but only if we go

back to the beginning and review the basiclogic of multilevel selection, what appearedto be at stake in the 1960s, and why the origi-nal rejection of group selection must be re-evaluated on the basis of subsequent re-search. Everyone can benefit from this backto basics approach, from the most advancedtheorists to students learning about socio-biology for the first time.

A Word About Tainted Words

It is a natural human tendency to avoid as-sociating oneself with people or ideas that

have acquired a bad reputation in the past.Thus, there are evolutionists who study socialbehavior, but avoid the term sociobiology,or who study psychology, but avoid the termevolutionary psychology, because of partic-ular ideas that were associated with theseterms in the past, including their supposed


3/23

December 2007 329THEORETICAL FOUNDATION OF SOCIOBIOLOGY

political implications. At a broader scale,there are people who avoid the word evolu-tion because of past negative associations,even though they are clearly talking aboutevolutionary processes. We think that thisvery understandable temptation needs to beresisted in the case of scientific terminology,because the short-term gain for the user(avoiding negative associations) results inlong-term confusion for the field as a whole(a proliferation of terms that mean the samething). The problem has been especially se-vere for multilevel selection theory becausemany evolutionists have felt that their very ca-reers would be jeopardized if they invokedgroup selection. In some cases, their fearswere well founded; we could provide numer-ous examples of colleagues whose articles andgrant proposals were rejected when stated interms of multilevel selection theory, and thenaccepted when restated using other terms. Inthis article, we define our terms at face value,regardless of past associations: sociobiology isthe study of social behavior from a biologicalperspective, group selection is the evolutionof traits based on the differential survival andreproduction of groups, and so on. Return-ing to face-value definitions is a first step to-ward resolving the confusion that plagues themodern sociobiological literature (see alsoFoster et al. 2007).

From an evolutionary perspective, a behav-ior can be regarded as social whenever it in-fluences the fitness of other individuals in ad-dition to the actor. Social behaviors need notbe prosocial; aggression fits the definition asdoes cooperation. Also, the interactions neednot be direct; a feeding behavior that reducesthe fitness of others by depleting their re-sources counts as social. Even genetic and de-velopmental interactions within a single in-dividual can be regarded as social, since theorganisms of today are now known to be thesocial groups of past ages, as we will describein more detail below. Narrower definitions of

social behavior might be useful for some pur-poses, but the important point to keep inmind is that the concepts reviewed in this ar-ticle apply to any trait that influences the fit-ness of others in addition to the actor, re-gardless of how social these traits mightappear in the intuitive sense.

The History and Basic Logic ofMultilevel Selection Theory

During evolution by natural selection, a

heritable trait that increases the fitness of oth-ers in a group (or the group at a whole) atthe expense of the individual possessing thetrait will decline in frequency within thegroup. This is the fundamental problem thatDarwin identified for traits associated withhuman morality, and it applies with equalforce to group-advantageous traits in otherspecies. It is simply a fact of social life thatindividuals must do things for each other tofunction successfully as a group, and thatthese actions usually do not maximize theirrelative fitness within the group.

Why is there usually a tradeoff? Because

there is usually a tradeoff between all adap-tations. Antipredator adaptations usually in-terfere with harvesting food, adaptations formoving through one medium (such as theair) usually interfere with moving through an-other medium (such as the water), and so on.The same principle applies to adaptations forfunctioning as a team player in a well-coor-dinated group, compared to maximizingones relative fitness within the group. Thisdoes not mean that the tradeoff must neces-sarily be severe. Benefiting others or onesgroup as a whole does not invariably requireextreme self-sacrifice, such as rushing into a

burning house to save a child, but it does re-quire some set of coordinating mechanisms,such as organizing and paying for a fire de-partment, passing and enforcing fire safetylegislation, and so on. It is unlikely that thesecoordination mechanisms evolve as a coin-cidental byproduct of organisms that areadapted exclusively to survive and reproducebetter than other members of their samegroup. That is why Darwin felt confident insaying that a high standard of morality givesbut a slight or no advantage to each individ-ual man and his children over the other men ofthe same tribe. As for human morality, so alsofor group-level adaptations in all species.

Something more than natural selectionwithin single groups is required to explainhow altruism and other group-advantageoustraits evolve by natural selection. For Darwin,in the passage quoted above, that some-


4/23


thing was between-group selection. Group-advantageous traits do increase the fitness ofgroups, relative to other groups, even if theyare selectively neutral or disadvantageouswithin groups. Total evolutionary change in apopulation can be regarded as a final vectormade up of two component vectors, within-and between-group selection, that often pointin different directions.

The basic logic of multilevel selection ap-plies to an enormous range of social behaviors,including the evolution of sexual reproduc-tion and sex ratio, distastefulness in insects,prudent use of resources, warning othersabout predators, social insect colonies as su-perorganisms, and more. The relevant group-ings are equally diverse, from a social insectcolony (as a superorganism) or an ephemeralflock of birds (for warning calls), to multige-nerational groups (for prudent use of re-sources), to entire species and clades (for sex-ual reproduction). Two related themes givethese examples conceptual unity. First, singletraits can evolve despite being locally disadvan-tageous wherever they occur. For this to hap-pen, an advantage at a larger scale (betweengroups) must exist to counteract the disadvan-tage at a smaller scale (within groups). Second,a higher-level unit (such as a social insect col-ony) can become endowed with the sameadaptive properties that we associate with sin-gle organisms. There can be such a thing asa superorganism. D S Wilson (1997) referredto these themes as altruism and organism.They are closely related but not entirely over-lapping, since becoming a superorganism in-volves more than the evolution of a singletrait.

Evolutionary theory was placed on a math-ematical foundation by the first populationgeneticists, in particular Ronald Fisher, Sew-all Wright, and J B S Haldane. Each consid-ered the problem of multilevel selection, butonly briefly, because it was not the most im-portant issue compared to even more foun-

dational issues such as the consequences ofMendelian genetics (reviewed by Sober andD S Wilson 1998). All three men shared Dar-wins perception that group-advantageoustraits seldom maximize relative fitness withingroups, thereby requiring a process ofbetween-group selection to evolve. Unfortu-

nately, many other biologists did not sharethis insight and uncritically assumed that ad-aptations evolve at all levels of the biologicalhierarchy without requiring a correspond-ing level of selection. When the need forbetween-group selection was acknowledged,it was often assumed that between-group se-lection easily trumped within-group selec-tion. The following passage from the text-book Principles of Animal Ecology (Allee et al.1949:729) illustrates what became known inretrospect as nave group selectionism:

The probability of survival of individualliving things, or of populations, increaseswith the degree with which they harmo-niously adjust themselves to each otherand to their environment. This principleis basic to the concept of the balance ofnature, orders the subject matter of ecol-ogy and evolution, underlies organismicand developmental biology, and is thefoundation for all sociology.

Another nave group selectionist was V CWynne-Edwards, who proposed that organ-isms evolve to assess and regulate their popu-lation size to avoid overexploiting their re-sources in his book, Animal Dispersion inRelation to Social Behavior (Wynne-Edwards1962, 1986). He was aware that group selec-

tion would be required and would often beopposed by selection within groups, but heassumed that group selection would usuallyprevail and proceeded to interpret a vast ar-ray of animal social behaviors according to histhesis without evaluating the levels of selec-tion in any particular case.

These issues began to occupy center stageamong evolutionary biologists in the 1960s,especially under the influence of George CWilliamss (1966) Adaptation and Natural Selec-tion. Williams began by affirming the impor-tance of multilevel selection as a theoreticalframework, agreeing with Darwin and the

population geneticists that group-level adap-tations require a process of group-level selec-tion. He then made an additional claim thatbetween-group selection is almost invariablyweak compared to within-group selection(both positions are represented in the above-quoted passage). It was this additional claim


5/23


that turned multilevel selection theory intowhat became known as the theory of individ-ual selection. Ever since, students have beentaught that group selection is possible in prin-ciple, but can be ignored in practice. Seem-ingly other oriented behaviors must be ex-plained as forms of self-interest that do notinvoke group selection, such as by helpingones own genes in the bodies of others (kinselection), or by helping others in expecta-tion of return benefits (reciprocity). The con-cept of average effects in population geneticstheory, which averages the fitness of allelesacross all genotypic, social, and environmen-tal contexts, was elaborated by both Williamsand Richard Dawkins (1976) into the geneseye view of evolution, in which everythingthat evolves is interpreted as a form of ge-netic selfishness.

The rejection of group selection in the1960s was based on three arguments, like thelegs of a stool: a) group selection as a signifi-cant evolutionary force is theoretically im-plausible; b) there is no solid empirical evi-dence for group selection as a distinctive,analytically separable process; and c) alter-native theories can explain the evolution ofapparent altruism without invoking groupselection. In the following sections, we willshow that all three arguments have failed,based on subsequent research. If this infor-mation had been available to Williams andothers in the 1960s, the history of sociobiol-ogy would have headed in a completely dif-ferent direction. The component vectors ofwithin- and between-group selection wouldneed to be calculated on a case-by-case basisto determine the final vector of evolutionarychange in the total population. Traits couldlegitimately be regarded as for the good ofthe group whenever they evolve by groupselection, in the same sense that an indi-vidual-level adaptation (such as the eye)is regarded as for the good of the individ-ual. Instead, sociobiology proceeded along

a seemingly triumphant path based entirelyon the calculus of individual and geneticself-interest, under the assumption thatgroup selection can be categorically ig-nored. It is precisely this branch point thatmust be revisited to put sociobiology back ona firm theoretical foundation.

The Theoretical Plausibility ofGroup Selection as a Significant

Evolutionary Force

The rejection of group selection was basedlargely on theoretical plausibility arguments,which made it seem that between-group se-lection requires a delicate balance of param-eter values to prevail against within-group se-lection. These early models were published ata time when the desktop computing revolu-tion, the study of complex interactions, andappreciation of such things as social control(e.g., Ratnieks and Visscher 1989; Boyd andRicherson 1992) and gene-culture coevolu-tion (Lumsden and E O Wilson 1981; Boydand Richerson 1985; Richerson and Boyd2005) were barely on the horizon. It should

surprise no one that the initial assessmentmust be revised on the basis of four decadesof subsequent research.

All of the early models assumed that altru-istic and selfish behaviors are caused directlyby corresponding genes, which means thatthe only way for groups to vary behaviorallyisfor them to varygenetically. Hardly anyone re-gards such strict genetic determinism as bio-logically realistic, and this was assumed in themodels primarily to simplify the mathematics.Yet, when more complex genotype-pheno-type relationships are built into the models,the balance between levels of selection can be

easily and dramatically altered. In otherwords, it is possible for modest amounts ofgenetic variation among groups to result insubstantial amounts of heritable phenotypicvariation among groups (D S Wilson 2004).

The early models also assumed that varia-tion among groups is caused primarily bysampling error, which means that it declinesprecipitously with the number of individualsthat independently colonize each group andmigration among groups during their exis-tence. This assumption must be revised onthe basis of agent-based models. When indi-vidual agents interact according to biologi-cally plausible decision rules, a spatial patch-iness emerges that has little to do withsampling error (e.g., Johnson and Boerlijst2002; Pepper and Smuts 2002; Pepper 2007).An example is a recent simulation model onthe kind of social signaling and population


6/23


regulation envisioned by Wynne-Edwards(Werfel and Bar-Yam 2004). Individuals cre-ate a local signal when crowded and curtailtheir reproduction accordingly. Their basereproductive rate and response to the signalare allowed to vary as independent continu-ous traits, including cheaters who repro-duce at the maximum rate and ignore thesignal altogether. Interactions occur on a two-dimensional lattice in which each cell repre-sents an area occupied by the resource alone,both the resource and consumers, or by nei-ther. Consumers that reproduce at the maxi-mum rate are selectively advantageous withingroups, but tend to drive their resource (and,therefore, themselves) extinct, exactly as en-visioned by Wynne-Edwards and the earlygroup selection models. More prudent con-sumers are maintained in the total popula-tion by spatial heterogeneity, which emergesspontaneously on the basis of complex inter-actions among the various traits. The localdisadvantage of curtailed reproduction doesnot entirely determine the outcome of selec-tion in the total population. In general, com-plex social and ecological interactions, cou-pled with limited dispersal, result in a kind ofspatial heterogeneity that is far outside theenvelope conceived by earlier models basedon sampling error in the absence of complexinteractions (see also Gilpin 1975; Aviles et al.2002; Aktipis 2004).

Another early conclusion was that group se-lection is weak for groups that last for multi-ple generations, because the generationtime is greater for groups than for individ-uals. Three examples will show how this con-clusion has been overturned by subsequenttheoretical models. First, even though altru-ists decline in frequency within each groupand ultimately go extinct after a sufficientnumber of generations, the differential fit-ness of groups also increases with each gen-eration, especially when the groups grow ex-ponentially at a rate determined by the

frequency of altruists. Simulations show thatgroup selection can remain a significant forceeven when the groups last 10 or 15 genera-tions between dispersal episodes (D S Wilson1987; Aviles 1993). Second, Gilpin (1975)showed that when predator/prey dynamicsare nonlinear, a small increase in predator

consumption rate can have a large effect onextinction rates, causing group selection tobe effective in multiple-generation groups.Third,Peck (2004) modeled altruism andself-ishness as suites of traits that must occur inthe right combination to function correctly,rather than as single traits. In this case, whena selfish individual migrates into an altruisticgroup, its genes do not spread because theybecome dissociated by sexual reproductionand no longer occur in the right combina-tion. An altruistic group can persist indefi-nitely, replacing less altruistic groups whenthey go extinct. These and other examples donot imply that group selection is invariablyef-fective in multigenerational groups, but theydo overturn the earlier conclusion that group

selection can be categorically ignored.Acknowledging the theoretical plausibilityof group selection as a significant evolution-ary force is not a return to the bad old daysof nave group selectionism. It has alwaysbeen the goal of population genetics theoryto provide a complete accounting system forevolutionary change, including selection,mutation, drift, and linkage disequilibrium.The question is whether group selection canbe categorically ignored when natural selec-tion is separated into within- and between-group components. Few theoretical biologistswould make this claim today, however reason-

able it might have appeared in the 1960s. Yet,these developments have not resulted in anappropriately revised theory, even amongsome of the theorists, nor have they spread tothe wider community of scientists interestedin the evolution of social behavior. There is aform of nave selectionism that needs to becorrected, as before the publication ofAdap-tation and Natural Selection, but today it is thenave assumption that group selection can beconsistently ignored.

Empirical Evidence forGroup Selection

The rejection of group selection in the1960s was not based upon a distinguishedbody of empirical evidence. Instead, Williams(1966) used the theoretical implausibility ofgroup selection as a significant evolutionaryforce to argue that hypotheses framed in


7/23


terms of individual selection are more parsi-monious and, therefore, preferable to hypoth-eses that invoke group selection. In this fash-ion, broad categories of behavior such asdominance and territoriality were interpretedindividualistically on the basis of plausibility ar-guments, without careful measurements ofwithin- versus between-group selection for par-ticular traits in particular species. Parsimonycan be a factor in deciding between alternativehypotheses, but it cannot substitute for an eval-uation of the data (Sober and D S Wilson 1998;Sober 2008). No population geneticist wouldargue that drift is more likely than selectionand no ecologist would argue that predationis more likely than competition on the basis ofparsimony. These alternatives are all plausibleand their relative importance must be deter-mined empirically on a case-by-case basis. Sim-ilarly, the direction and strength of within- andbetween-group selection must be determinedon a case-by-case basis if both are theoreticallyplausible.

The closest that Williams came to a rigor-ous empirical test was for sex ratio, leadinghim to predict that female-biased sex ratioswould provide evidence for group selection.The subsequent discovery of many examplesof female-biased sex ratios, as well as evidenceof group selection in the evolution of diseaseorganisms, brought him back toward multi-level selection in the 1990s (Williams andNesse 1991; Williams 1992).

Some of the best recent evidence for groupselection comes from microbial organisms, inpart because they are such efficient systemsfor ecological and evolutionary researchspanning many generations (Velicer 2003).The wrinkly spreader (WS) strain of Pseu-domonas fluorescens evolves in response to an-oxic conditions in unmixed liquid medium,by producing a cellulosic polymer that formsa mat on the surface. The polymer is expen-sive to produce, which means that nonpro-ducing cheaters have the highest relative fit-

ness within the group. As they spread, the matdeteriorates and eventually sinks to the bot-tom. WS is maintained in the total populationby between-group selection, despite its selec-tive disadvantage within groups, exactly as en-visioned by multilevel selection theory (Rai-ney and Rainey 2003).

As another example, Kerr et al. (2006)created a metapopulation of bacteria (theresource) and phage (the consumer) by cul-turing them in 96-well microtiter plates. Mi-gration between groups was executed by ahigh-throughput, liquid-handling robot ac-cording to a prespecified migration scheme.Biologically plausible migration rates en-abled prudent phage strains to out-compete more rapacious strains, exactly asenvisioned by Wynne-Edwards and subse-quent theorists such as Gilpin (1975) andWerfel and Bar-Yam (2004). As Kerr et al. putit, spatially restricted migration reduces theprobability that phage reach fresh hosts, ren-dering rapacious subpopulations more proneto extinction through dilution. Consequently,the tragedy of the commons is circumventedat the metapopulation scale in the Restrictedtreatment (Kerr 2006:77). More generally,the well-established fact that reduced viru-lence often evolves by group selection in dis-ease organisms (Bull 1994; Frank 1996) pro-vides a confirmation of Wynne-Edwardshypothesisnot for allspecies, but for at leastsomespecies.

Multilevel selection experiments in the lab-oratory have been performed on organismsas diverse as microbes, plants, insects, andvertebrates (Goodnight et al. 1992; Good-night and Stevens 1997). Phenotypic varia-tion among groups is usually considerable,even when the groups are founded by largenumbers of individuals, as expected on thebasis of the newer theoretical models. For ex-ample, microcosms colonized by millions ofmicrobes from a single well-mixed source nev-ertheless become variable in their phenotypicproperties within a matter of days. When mi-crocosms are selected on the basis of theseproperties and used to colonize a new gen-eration of microcosms, there is a response toselection (Swenson et al. 2000a,b).

Quantitative genetics models separate phe-notypic variation into additive and nonaddi-

tive components, with only the former lead-ing to a response to selection (narrow-senseheritability). Laboratory selection experi-ments show that the nonadditive componentof variation within groups can contribute tothe additive component of variation amonggroups, causing group-level traits to be more


8/23


heritable than individual-level traits. For ex-ample, selecting plants within a single groupon the basis of leaf area did not producemuch response to selection, but selectingwhole groups on the basis of leaf area pro-duced a strong response to selection. This re-sult makes sense theoretically when pheno-typic traits such as leaf area are influenced byinteractions among individuals within thegroup, rather than being directly coded bygenes (Goodnight 2000, 2005).

Field studies of social vertebrates are sel-dom as precise as laboratory experiments butnevertheless provide convincing evidence forgroup selection. The following description ofterritorial defense in lions (Packer and Hein-sohn 1996:1216; see also Heinsohn and

Packer 1995) is virtually identical to Darwinspassage about human morality that began thisarticle: Female lions share a common re-source, the territory; but only a proportion offemales pay the full costs of territorial de-fense. If too few females accept the responsi-bilities of leadership, the territory will be lost.If enough females cooperate to defend therange, their territory is maintained, but theircollective effort is vulnerable to abuse by theircompanions. Leaders do not gain additionalbenefits from leading, but they do providean opportunity for laggards to gain a freeride. In this field study, extensive efforts to

find a within-group advantage for territorialdefense failed, leaving between-group selec-tion as the most likelyand fully plausiblealternative.

To summarize, four decades of researchsince the 1960s have provided ample empiri-cal evidence for group selection, in additionto its theoretically plausibility as a significantevolutionary force.

Are There Robust Alternativesto Group Selection?

Inclusive fitness theory (also called kin se-lection theory), evolutionary game theory(including the concept of reciprocal altru-ism), and selfish gene theory were all devel-oped explicitly as alternatives to group selec-tion. In addition to these major theoreticalframeworks, there are numerous conceptssuch as indirect reciprocity (Nowak and Sig-

mund 2005; Nowak 2006), byproduct mutu-alism (Dugatkin 2002; Sachs et al. 2004), andcostly signaling (Lachmann et al. 2001; Cronk2005) that claim to explain the evolution ofcooperation and altruism without invokinggroup selection. Nevertheless, all evolution-ary models of social behavior share certainkey features, no matter what they are called.Recognizing the similarities can go a long waytoward establishing theoretical unity for thefield.

First, all models assume the existence ofmultiple groups. Why? Because social inter-actions almost invariably take place amongsets of individuals that are small compared tothe total population. No model can ignorethis biological reality. In N-person game the-

ory, N refers to the size of the group withinwhich social interactions occur. In kin selec-tion theory, r specifies that individuals are in-teracting with a subset of the population withwhom they share a certain degree of genea-logical, genetic, or phenotypic similarity (de-pending upon the specific formulation), andso on. The groups need not have discreteboundaries; the important feature is that so-cial interactions are local, compared to thesize of the total population.

Second, all models must converge on thesame definition of groups for any particulartrait. Why? Because all models must calculate

the fitness of individuals to determine whatevolves in the total population. With social be-haviors, the fitness of an individual dependsupon its own phenotype and the phenotypesof the others with whom it interacts. Theseother individuals must be appropriately spec-ified or else the model will simply arrive atthe wrong answer. If individuals interact ingroups of N 5, two-person game theory willnot do. Evolutionary theories of social behav-ior consider many kinds of groups, but that isonly because they consider many kinds oftraits. For any particular trait, such as inter-group conflict in humans, mat formation inbacteria, or territorial defense in lions, thereis an appropriate population structure thatmust conform to the biology of the situation,regardless of what the theoretical frameworkis called. That is the concept of the trait-group(D S Wilson 1975); the salient group (and


9/23


other aspects of population structure) for anyparticular trait.

Third, in virtually all cases, traits labeledcooperative and altruistic are selectively dis-advantageous within the groups and requirebetween-group selection to evolve. W D Ham-ilton (1975) realized this property of inclusivefitness theory when he encountered the workof George Price in the early 1970s (Price1970, 1972). Price had derived an equationthat partitions total gene frequency changeinto within and between group components.When Hamilton reformulated his theory interms of the Price equation, he realized thataltruistic traits are selectively disadvantageouswithin kin-groups and evolve only becausekin-groups with more altruists differentiallycontribute to the total gene pool. Hamiltonskey insight about the importance of geneticrelatedness remained valid, but his previousinterpretation of inclusive fitness theoryas analternative to group selection was wrong, as hefreely acknowledged (Hamilton 1996:173174; Schwartz 2000). The importance of ge-netic relatedness can be explained in termsof the parameters of multilevel selection,rather than requiring additional parameters(Michod 1982). For example, genetic relat-edness might be an important factor in theevolution of territorial defense in lions, butonly because it increases genetic variationamong groups, thereby increasing the impor-tance of between-group selection comparedto within-group selection. Much the sameconclusion has been drawn from social in-sects (e.g., Queller 1992; Bourke and Franks1995; Wenseleers et al. 2003), as we will de-scribe in more detail below.

For two-person game theory, the coopera-tive tit-for-tat strategy never beats its socialpartner; it only loses or draws. The only rea-son that tit-for-tat and other cooperative strat-egies evolve in a game theory model is be-cause groups of cooperators contribute moreto the total gene pool than groups of non-

cooperators, as Anatol Rapoport (1991)clearly recognized when he submitted the tit-for-tat strategy to Robert Axelrods famouscomputer simulation tournament. The pairsof socially interacting individuals in two-per-son game theory might seem too small orephemeral to call a group (Maynard Smith

2002), but the same dynamic applies to N-per-son game theory as a whole, including largeand persistent groups that are described interms of evolutionary game theory, but whichoverlap with traditional group selection mod-els. All of these models obey the followingsimple rule, regardless of the value of N, theduration of the groups, or other aspects ofpopulation structure: Selfishness beats altruismwithin single groups. Altruistic groups beat selfishgroups. The main exception to this rule in-volves models that result in multiple localequilibria, which are internally stable by def-inition. In this case, group selection can favorthe local equilibria that function best at thegroup level, a phenomenon sometimes calledequilibrium selection (Boyd and Richerson1992; Samuelson 1997; Gintis 2000; themodel by Peck 2004 described earlier pro-vides an example).

Dawkins (1976, 1982) envisioned selfishgene theory and the concept of extendedphenotypes as arguments against group selec-tion but, in retrospect, they are nothing of thesort. The concept of extended phenotypesnotes that genes can have effects that extendbeyond the body of the individual, such as abeaver dam. Genes that cause beavers to builddams are still at a local disadvantage com-pared to genes in beavers in the same pondthat do not build dams, so the concept of ex-tended phenotypes does nothing to preventthe fundamental problem of social life or toprovide a solution other than that providedby between-group selection. The concept ofgenes as replicators and the fundamentalunit of selection is identical to the conceptof average effects in population genetics,which averages the fitness of alleles across allgenotypic, environmental, and social con-texts. The average effect provides the bottomline of what evolves in the total population,the final vector that reflects the summationof all the component vectors. The wholepoint of multilevel selection theory is, how-

ever, to examine the component vectors of evo-lutionary change, based on the targets of se-lection at each biological level and, inparticular, to ask whether genes can evolve onthe strength of between-group selection, de-spite a selective disadvantage within groups.Multilevel selection models calculate the av-


10/23


erage effects of genes, just like any otherpopulation genetics model, but the final vec-tor includes both levels of selection and, byitself, cannot possibly be used as an argumentagainst group selection. Both Williams(1985:8) and Dawkins (1982:292298) even-tually acknowledged their error (reviewed inD S Wilson and Sober 1998; see also Okasha2005, 2006), but it is still common to read inarticles and textbooks that group selection iswrong because the gene is the fundamentalunit of selection.

A similar problem exists with evolutionarymodels that are not explicitly genetic, such asgame theory models, which assume that thevarious individual strategies breed true insome general sense (Maynard Smith 1982;

Gintis 2000). The procedure in this case is toaverage the fitness of the individual strategiesacross all of the social groupings, yielding anaverage fitness that is equivalent to the aver-age effect of genes in a population geneticsmodel. Once again, it is the final vector thatis interpreted as individual fitness and re-garded as an argument against group selec-tion, even though the groups are clearly de-fined and the component vectors are therefor all to see, once it is clear what to look for.

To summarize, all of the theories that weredeveloped as alternatives to group selectionassume the basic logic of multilevel selection

within their own frameworks.

Pluralism

The developments outlined above have ledto a situation that participants of the contro-versy in the 1960s would have difficulty rec-ognizing. The theories that were originally re-garded as alternatives, such that one might beright and another wrong, are now seen asequivalent in the sense that they all correctlypredict what evolves in the total population.They differ, however, in how they partition se-lection into component vectors along the way.The frameworks are largely intertranslatableand broadly overlap in the kinds of traits andpopulation structures that they consider. Tomake matters more confusing, each majorframework comes in a number of varieties(e.g., Fletcher and Zwick 2006; Okasha 2006;West et al. 2007; D S Wilson 2007a). Consid-

erable sophistication is required to interpretthemeaningsof terms such as altruism, self-ishness, relatedness, and individual selec-tion, depending upon the specific model be-ing employed.

This kind of pluralism is a mixed blessing.On the positive side, multiple perspectives arehelpful for studying any complex problem, solong as they are properly related to eachother (Sober and D S Wilson 2002; Foster2006). On the negative side, it is easy to losesight of the fundamental issues that made therejection of group selection appear so impor-tant in the first place. The central issue ad-dressed by Williams in Adaptation and NaturalSelection was whether adaptations can evolveat the level of social groups and other higher-level units. The problem, as recognized byDarwin and affirmed by Williams, was thattraits that are for the good of the group areusually not favored by selection withingroupswhat we have called the fundamen-tal problem of social life. When Williams andothers rejected group selection, they were re-jecting the possibility that adaptations evolveabove the level of individual organisms. Thisis not a matter of perspective, but a funda-mental biological claim. If true, it is every bitas momentous as it appeared to be in the1960s. If false, then its retraction is equallymomentous.

A sample of issues debated by contempo-rary theorists and philosophers of biology willshow that, whatever the merits of pluralism,they do not deny the fundamental problemof social life or provide a solution other thanbetween-group selection. Let us begin with in-clusive fitness theory. Hamilton (1963, 1964)originally interpreted the coefficient of relat-edness (r), as a measure of genealogical re-latedness, based on genes that are identicalby descent. When he reformulated his theoryin terms of the Price equation, he realizednot only that kin selection is a kind of groupselection, but also that r can be interpreted

more broadly as any positive correlationamong altruistic genesnot just based onidentity by descent (Hamilton 1975). Subse-quent theorists have broadened the interpre-tation of r still further. For example, altruisticgenes can evolve as long as they associate pos-itively with altruistic phenotypes, coded by the


11/23


same or different altruistic genes (Queller1985; Fletcher and Doebeli 2006). When in-dividuals benefit their entire group (includ-ing themselves) at their own expense, r canbe positive even in randomly formed groups(Pepper 2000; Fletcher and Zwick 2004).Models that were originally conceptualized asexamples of group selection, in contrast tokin selection, such as Maynard Smiths (1964)haystack model, can be reconceptualized asmodels of kin selection by noting that mem-bers of groups are more genetically similar toeach other than to members of the totalpopulation. Generality is a virtue, so it is un-derstandable that theorists might want topush the boundaries of inclusive fitness the-ory as far as possible. Nevertheless, when ev-erything that was ever called group selectioncan now be described in terms of inclusivefitness theory, it is time to take stock of theoriginal empirical issues at stake. Is the fun-damental problem of social life present in thebroadened form of inclusive fitness theory?Absolutely. Altruistic traits are locally disad-vantageous, just as they always were. Are theingredients of between-group selection re-quired to solve the fundamental problem ofsocial life? Absolutely. Altruistic traits stillmust be favored at a larger scale to counteracttheir local disadvantage. Does altruism evolveonly among immediate genealogical rela-tives? Absolutely not. In the passage quotedat the beginning of this article, Williams(1966) rejected group-level adaptations forany groups other than a family or com-posed of individuals that need not be closelyrelated, by which he meant genealogical re-latedness. Inclusive fitness theory refuted thisclaim as soon as r became generalized beyondimmediate genealogical relatedness (e.g.,Aviles 2002).

To pick a second example of pluralism,Kerr and Godfrey-Smith (2002a) outline twoequivalent frameworks that they call collectiveand contextual (similar to Dugatkin and

Reeves 1994 distinction between multilevelselection and broad-sense individualism). Inthe collective framework, groups are assignedfitnesses and individuals are assigned differ-ent shares of their groups fitness. In the con-textual framework, individuals are assignedfitnesses that are functions of the composi-

tion of their group. The distinction betweenthe two frameworks is similar to thinking ofgenotypes as individuals, as in standard popu-lation genetics theory, as opposed to environ-ments of genes, as in selfish gene theory. Kerrand Godfrey-Smith stress that the two frame-works are fully equivalent, which means thatany statement in one can be translated into astatement in the other. Equivalence alsomeans that neither is more correct in anycausal sense, although one might providemore insight than the other in any particularcase. Fair enough, but this kind of pluralismby itself does not address any particular em-pirical issue. When we begin to ask the em-pirical questions that endow the group selec-tion controversy with such significance, wediscover that the contextual approach doesnot avoid the fundamental problem of sociallife or provide a solution other than between-group selection. It merely describes theseprocesses in different terms. In this sensebroad-based individualism ( the contex-tual approach) is nothing like the theory ofindividual selection that claimed to be a gen-uine alternative to group selection, such thatone could be right and the other wrong (formore detailed discussion of this issue, seeKerr and Godfrey Smith 2002b; Sober andWilson 2002).

As a third example of pluralism, eventhough the Price equation elegantly parti-tions selection into within- and between-group components, it misclassifies certaincases. In particular, when individuals that dif-fer in their individual fitness (without behav-ing socially at all) are separated into groups,the between-group component of the Priceequation is positive, even though there is nogroup selection (Sober 1984). Another statis-tical method called contextual analysis avoidsthis problem, but it misclassifies other cases.Thus, there is no single statistical method thatcaptures all aspects of multilevel selectiontheory (van Veelen 2005; Okasha 2006). This

is interesting and important, but does not castdoubt on the basic empirical issues. In fact,the reason that we can spot classification er-rors in statistical methods such as the Priceequation is because we have such a strongsense of what multilevel selection means inthe absence of formal statistical methods.


12/23


In general, the issues discussed under therubric of pluralism are important but alsohighly derived, to the point of becoming de-tached from the issues that endowed multi-level selection with such importance in thefirst place. There is a need for all perspectivesto converge upon a core set of empiricalclaims, including the following:

1) There isa fundamental problem thatrequires a solution in order to explain theevolution of altruism and other group-leveladaptations. Traits that are for the good ofthe group are seldom selectively advanta-geous within groups. At worst, they arehighly self-sacrificial. At best, they providepublic goods at little cost to the actor, mak-ing them close to selectively neutral, orthey constitute a stable local equilibrium.Notice that the only way to evaluate thisclaim is by making a local relative fitnesscomparison. It is not enough to show thatan individual increases its absolute fitnessbecause it might increase the fitness of oth-ers in its own group even more (D S Wilson2004).

2) If a trait is locally disadvantageouswherever it occurs, then the only way for itto evolve in the total population is for it tobe advantageous at a larger scale. Groupswhose members act for the good of thegroup must contribute more to the totalgene pool than groups whose members actotherwise. This is the only solution to theproblem from an accounting standpoint,which is why the basic logic of multilevelselection is present in all theoretical frame-works, as we showed in the previous sec-tion. In biological hierarchies that includemore than two levels, the general rule isadaptation at any level requires a processof natural selection at the same level andtends to be undermined by natural selec-tion at lower levels. All students of evolu-tion need to learn this rule to avoid theerrors of nave group selectionism. Notice

that, so far, we are affirmingkey elements ofthe consensus that formed in the 1960s.3) Higher-level selection cannot be cat-

egorically ignored as a significant evolu-tionary force. Instead, it must be evaluatedseparately and on a case-by-case basis. Fur-thermore, all of the generalities about the

likelihood of group selection that becameaccepted in the 1960s need to be reexam-ined. Wynne-Edwards hypothesis has meritfor at least some species, group selectioncan be significant in groups that last formultiple generations, and so on. One ofthe biggest problems with the current lit-erature is that the early generalities remainunquestioned, as if there is an old groupselection that deserves to be rejected and anew form that bears little relationshipwith its own past (e.g., Keller 1999; West etal. 2006, 2007). This is a false portrayal andcannot be justified on the basis of plural-ism. Going back to basics requires acknowl-edgment that Williams and others wereright to criticize nave group selection, butjust plain wrong in their own assessment ofthe likelihood of group selection. New gen-eralities need to be formed on the basis ofongoing research.

4) The fact that a given trait evolves inthe total population is not an argumentagainst group selection. Evaluating levels ofselection requires a nested series of relativefitness comparisons; between genes withinindividuals, between individuals withingroups, between groups within a popula-tion of groups, and so on, each presentingtraits that are separate targets for selection.All theoretical frameworks include the in-formation for making these comparisons,as we have seen. In this sense, they are notpluralistic. They merely differ in the degreeto which they focus on the comparisons ontheir way toward calculating evolutionarychange in the total population. If we aremerely interested in whether a given traitevolves, then it is not necessary to examinelevels of selection, and multiple perspec-tives can be useful. If we want to addressthe particular biological issues associatedwith multilevel selection, then we are re-quired to examine the appropriate infor-mation and the perspectives converge with

each other.To summarize, it is possible to acknowledgethe usefulness of multiple perspectives with-out obscuring the fundamental biological is-sues that seemed so clear in the 1960s. Wethink that items 14 above can become thebasis for a new consensus about when adap-


13/23


tations evolve at any given level of the bio-logical hierarchy, restoring clarity and unityto sociobiological theory. We will now exam-ine three cases where higher-level selectionhas been exceptionally important: the evolu-tion of individual organisms, the evolution ofeusociality in insects and other taxa, and hu-man evolution.

Individuals as Groups

An important advance in evolutionary bi-ology began with Marguliss (1970) theory ofthe eukaryotic cell. She proposed that eu-karyotic (nucleated) cells did not evolve bysmall mutational steps from prokaryotic (bac-terial) cells, but by symbiotic associations of

bacteria becoming so integrated that the as-sociations qualified as single organisms intheir own right. The concept of groups ofor-ganisms turning into groups asorganisms wasthen extended to other major transitions dur-ing the history of life, including the origin oflife itself as groups of cooperating molecularreactions, the first cells, and multicellular or-ganisms (e.g., Maynard Smith and Szathmar y1995, 1999; Michod 1999; Jablonka and Lamb2006; Michod and Herron 2006).

Despite multilevel selection theorys tur-bulent history for the traditional study of so-cial behavior, it is an accepted theoretical

framework for the study of major transitions.There is widespread agreement that selectionoccurs within and among groups, that the bal-ance between levels of selection can itselfevolve, and that a major transition occurswhen selection within groups is suppressed,enabling selection among groups to domi-nate the final vector of evolutionary change.Genetic and developmental phenomena suchas chromosomes, the rules of meiosis, a singlecell stage in the life cycle, the early sequestra-tion of the germ line, and programmed deathof cell lineages are interpreted as mecha-nisms for stabilizing the organism and pre-venting it from becoming a mere group ofevolving elements. At the same time, within-group selection is never completely sup-pressed. There are many examples of intra-genomic conflict that prevent the higher-levelunits from functioning as organisms in the

full and truest sense of the word (Burt andTrivers 2006).

The concept of major transitions decisivelyrefutes the notion that higher-level selectionis invariably weaker than lower-level selection.The domain of multilevel selection theoryhas been expanded to include the internalorganization of individuals in addition to thesocial organization of groups. Ironically, therejection of group selection made it heresyto think about groups as like organisms, andnow it has emerged that organisms are literallythe groups of past ages. Okasha (2005:1008)eloquently summarizes the implications ofthese developments for sociobiological theoryas a whole:

Since cells and multi-celled creatures ob-

viously have evolved, and function well asadaptive units, the efficacy of group se-lection cannot be denied. Just as theblanket assumption that the individualorganism is the sole unit of selection isuntenable from a diachronic perspec-tive, so too is the assumption that groupselection is a negligible force. For by fra-meshifting our perspective downwards,it becomes apparent that individual or-ganisms are co-operative groups, so arethe productof group selection!

Eusociality as a Major Transition

Eusociality, found primarily in social insectsbut now known for other organisms such asmammals (Sherman et al. 1991) and crustacea(MacDonald et al. 2006), has always played apivotal role in the history of sociobiology. Theterm eusocial is applied to colonies whosemembers are multigenerational, cooperate inbrood care, and are separated into reproduc-tive and nonreproductive castes. For the firsthalf of the 20th century, following W MWheelers classic paper of 1911, eusocial col-onies were treated as superorganisms thatevolved by between-colony selection. Hamil-

tons (1964) inclusive fitness theory appearedto offer a very different explanation based ongenetic relatedness, especially the extra-highrelatedness among sisters in ants, bees, andwasps based on their haplodiploid genetic sys-tem. The focus on genetic relatedness there-after made it appear as if social insect evolu-


14/23


tion could be explained without invokinggroup selection, along with other examples ofapparent altruism. The following passagefromWest Eberhard (1981:12; parenthetical com-ments are hers) illustrates the degree to whichbetween-colony selection was rejected as an ex-planation of eusociality in insects: Despite thelogical force of arguments against group (orcolony) selection (e.g., Williams 1966) and theinvention of tidy explanations for collabora-tion in individual terms . . . the supraorganism(colony-level selection) still haunts evolution-ary discussions of insect sociality.

Four decades later, there is an urgent needto establish some fundamental biologicalclaims that have been obscured rather thanclarified by multiple perspectives. Beginningwith Wheelers original claim that eusocialcolonies are superorganisms, the evolution ofeusociality falls squarely within the paradigmof major transitions. Most traits associatedwith eusociality do not evolve by increasing infrequency within colonies, but by increasingthe colonys contribution to the larger genepool. Inclusive fitness theory is not a denialof this fact, although that is how it was origi-nally interpreted. Hamiltons rule calculatesthe conditions under which an altruistic actincreases the proportion of altruistic genes inthe total population, not a single colony.Showing that a trait evolves in the total popu-lation is not an argument against group se-lection, as we have already stressed. The Priceequation demonstrated to Hamilton that al-truism is selectively disadvantageous withinkin groups, just as in any other kind of group.The importance of kinship is that it increasesgenetic variation among groups, thereforethe importance of between-group selectioncompared to within-group selection. Thereare traits that evolve by within-colony selec-tion, but they are forms of cheating that tendto impair the performance of the colony,similar to intragenomic conflict within indi-vidual organisms (Ratnieks et al. 2006). All

social insect biologists should be able to agreeupon these facts, regardless of the theoreticalframework that they employ.

Another substantive biological question isthe role of genealogical relatedness in theevolution of eusociality. Hamiltons originaltheory was that the extra-high sociality of in-

sect colonies can be explained by the extra-high relatedness among workers, at least inhaplodiploid species, when groups arefounded by single queens who have matedwith a single male. More generally, Hamil-tons rule (br c, where b benefit to therecipient, r coefficient of relatedness, andc cost to the altruist) easily gives the im-pression that the degree of altruism shouldbe proportional to r. This perception was infact a principal reason for the erroneous earlyacceptance of collateral (indirect) kin selec-tion as a critical force in the origin of euso-ciality (E O Wilson 1971,1975).

Decades of research have led to a morecomplicated story in which genealogical re-latedness plays at best a supporting ratherthan a pivotal role. The haplodiploidy hy-pothesis has failed on empirical grounds. Inaddition to termites, numerous other diploideusocial clades in insects and other taxa havebeen discovered since the 1960s, enough torender the association of haplodiploidy andeusociality statistically insignificant (E O Wil-son and Holldobler 2005). Moreover, manyhaplodiploid colonies are founded by multi-ple females and/or females that mated withmultiple males, lowering genetic relatednessto unexceptional levels. Further, followingcolony foundation in primitively eusocialwasp species, the degree of relatedness tendsto fall, not rise or hold steady, at least in caseswhere it has been measured (e.g., Landi et al.2003; Fanelli et al. 2004). These facts arewidely acknowledged by social insect biolo-gists but it is still common to read in the widerliterature that genetic relatedness is the pri-mary explanation for insect eusociality. Infact, extra-high relatedness within coloniesmay be better explained as a consequencerather than a cause of eusociality (E O Wilsonand Holldobler 2005).

From a multilevel evolutionary perspec-tive, traits that cause an insect colony to func-tion as an adaptive unit seldom increase in

frequency within the colony and evolve onlyby causing the colony to out-compete othercolonies and conspecific solitaires, eitherdirectly or through the differential produc-tion of reproductives. If colonies are initi-ated by small numbers of individuals, mini-mally a single female mated with a single


15/23


male, then there is ample genetic variationamong groups and only modest genetic vari-ation within groups. However, this is only oneof many factors that can influence the bal-ance between levels of selection. Consider ge-netic variation for traits such as nest construc-tion, nest defense, provisioning the colony forfood, or raiding other colonies. All of theseactivities provide public goods at private ex-pense. All entail emergent properties basedon cooperation among the colony members.Slackers are more fit than solid citizens withinany single colony, but colonies with more solidcitizens have the advantage at the group level.The balance between levels of selection will beinfluenced by the magnitude of the group-level benefits and individual-level costs, in ad-dition to the partitioning of genetic variationwithin and among groups. For example, eco-logical constraints are more important thangenetic relatedness in the evolution of euso-ciality in mole-rats (Burland et al. 2002). Thesame is true of the eusocial invertebrates(Choe and Crespi 1997; E O Wilson andHolldobler 2005). The ancestors of most eu-social insects probably built nests and re-mained to feed and protect their broodthroughout larval development. Such a pro-gressive provisioning was evidently the keypreadaptation for the origin of eusociality inthe Hymenoptera. It is the multigroup popu-lation structure provided by this ecologicalniche and the magnitude of shared benefitsthat brought these species up to and over thethreshold of eusociality, more than excep-tional degrees of genetic relatedness.

It might seem that reproductive division oflabor must be a form of high-cost altruismthat requires a high degree of genetic varia-tion among groups (represented by high rvalues) to evolve. This is only true, however,if heritable phenotypic variation exists forworker reproduction and if reproductiveworkers are not suppressed by the queen orother workers. Reproductive suppression is

common in eusocial species, and to under-stand its evolution we need to study the polic-ing and reproduction traits in conjunctionwith each other (Ratnieks et al. 2006). Sup-pressing the reproduction of others can befavored by within-group selection, but it cantake many forms that vary in their conse-

quences for the reproductive output of thecolony, compared to other colonies. Between-group selection is required to evolve forms ofreproductive suppression that function wellat the colony level, but the amount of geneticvariation among colonies need not be excep-tional. That need is diminished further whenthe trait favored by group selection is a formof phenotypic plasticity that enables single ge-notypes to be reproductive or nonreproduc-tivewhich, in fact, is universal in the socialinsects (E O Wilson 1975; Holldobler and EO Wilson 1990).

In eusocial insects, it appears that the evo-lution of anatomically distinct worker castesrepresents a point of no return beyondwhich species never revert to a more primi-tively eusocial, presocial, or solitary condition(E O Wilson 1971; Maynard Smith and Sza-

thmary 1995; E O Wilson and Holldobler2005). At this point, the colony has become astable developmental unit and its persistencedepends on its ability to survive and repro-duce, relative to other colonies and solitaryorganisms. The hypothetical mutant repro-ductive worker that would be favored bywithin-colony selection simply does not occurat significant levels or at all, although, insome species, cheating by workers occursand is suppressed through policing by fellowworkers. This is similar to the evolution of sex-ual lineages that do not give rise to asexualmutants (Nunney 1999) and the evolution ofmechanisms that prevent intragenomic con-flict in individual organisms (Maynard Smithand Szathmar y 1995, 1999).

A common assumption of theoretical mod-els is that genes have additive effects on phe-notypes, so that phenotypic variation amonggroups corresponds directly to genetic vari-ation among groups, as we have alreadystressed. More complex genotype-phenotyperelationships enable small genetic differencesto result in large phenotypic differences, at thelevel of groups no less than individual organ-

isms (D S Wilson 2004). Even a single mutantgene in a colony founded by unrelated indi-viduals can have powerful effects on pheno-typic traits such as caste development or al-location of workers to various tasks, whichmight provide a strong advantage to thegroup, compared to other groups.


16/23


Single eusocial insect colonies often have apopulation structure of their own, which canbe spatial or based on kin recognition. Thereis a multiple-tiered population structure inwhich selection can occur between individ-uals within immediate families (such as ma-trilines or patrilines), between immediatefamilies within a single colony, and betweencolonies within the larger population. Inkeeping with the dictum adaptation at anylevel requires a process of natural selection atthe same level and tends to be underminedby natural selection at lower levels, kin selec-tion becomes part of the problem as far ascolony-level selection is concerned. Numer-ous examples of nepotism as a disruptiveforce have been documented, along withmechanisms that have evolved to suppressnepotism along with individual selfishness,enabling the multifamily colony to be the pri-mary unit of selection (Ratnieks et al. 2006;Wenseleers and Ratnieks 2006).

Social insect biologists spend much of theirtime studying the mechanisms that enable acolony to function as an adaptive unit. Thetitle of one book, The Wisdom of the Hive(See-ley 1995), alludes effectively to Walter Can-nons (1932) The Wisdom of the Body, which fa-mously described the complex physiologicalmechanisms of single organisms. The socialinteractions that enable an insect colony tomake complex decisions are even directlycomparable to the neuronal interactions thatenable individual organisms to make deci-sions (Seeley and Buhrman 1999). These in-teractions did not evolve by within-colony se-lection, but by colonies with the mostfunctional interactions out-competing othercolonies. A high degree of relatedness wasnot required and little insight is gained bynoting that individuals benefit as members ofsuccessful groups. The challenge is to under-stand the complex mechanisms that enable acolony to function as a single organism, exactlyas imagined by Wheeler so long ago.

Almost all of the spectacular evolutionaryefflorescence of the more than 12,000 knownant species, hence almost all the progressiveadvance of their communication and castesystems, life cycles, algorithms of colonial self-organization and caste-specific adaptive de-mographies, are manifestly the product of

group selection acting on the emergent, col-ony-level traits, which are produced in turnby the interaction of the colony members.

We will conclude this section by discussingthe extent to which pluralism has facilitatedor retarded the study of the landscape of eu-sociality during the last four decades. Thequestion is not whether everything that wehave recounted above can be stated withinthe rubric of inclusive fitness theory; it can.Moreover, we certainly do not deny the ad-vances in knowledge about social insects inrecent decades, some of which has been stim-ulated by inclusive fitness theory as the dom-inant paradigm. Nevertheless, we also thinkthat inclusive fitness theory has retarded un-derstanding in a number of other, important

respects. First, it initially gave the impressionthat eusociality can be explained as an indi-vidual-level adaptation, without distinguish-ing and invoking group ( between-colony)selection; this turned out to be a monumentalmistake. Second, it misleadingly suggestedthat genetic relatedness is the primary factorthat explains the evolution of eusociality, dis-tracting attention from other factors ofgreater importance. Third, the coefficient ofrelatedness was originally interpreted interms of genealogical relatedness, whereas to-day it is interpreted more broadly in terms ofany genetic or even phenotypic correlation

among group members (Fletcher et al. 2006;Fletcher and Zwick 2006; Foster et al.2006a,b). Inclusive fitness theory now com-pletely overlaps with multilevel selection the-ory, as we have already stressed. Multiple per-spectives are useful, as long as they areproperly related to each other, and we aresure that inclusive fitness theory will be usedto study eusociality in the future. However, wealso think that multilevel selection theory willprove to be both correct and more heuristic,because it more clearly identifies the colonyas the unit of selection that has driven theevolution of the social complexity.

Human Evolutionas a Major Transition

Anyone who studies humans must acknowl-edge our groupish nature and the impor-tance of between-group interactions through-


17/23


out human history. Ever since the 1960s,sociobiologists and evolutionary psycholo-gists have been burdened with the task of ex-plaining these obvious facts without invokinggroup selection. In retrospect, these expla-nations appear needlessly contorted. Instead,human evolution falls squarely within the par-adigm of major transitions (Lumsden and EO Wilson 1981; Boehm 1999; Richerson andBoyd 1999; D S Wilson 2002, 2006, 2007a,b;Hammerstein 2003; Foster and Ratnieks2005; Bowles 2006).

A key event in early human evolution was aform of guarded egalitarianism that made itdifficult for some individuals to dominateothers within their own group (Bingham1999; Boehm 1999). Suppressing fitness dif-ferences within groups made it possible forbetween-group selection to become a power-ful evolutionary force. The psychologicaltraits associated with human moral systemsare comparable to the mechanisms that sup-press selection within groups for other majortransitions, such as chromosomes and therules of meiosis within multicellular organ-isms and policing mechanisms within euso-cial insect colonies (D S Wilson 2002; Avileset al. 2004; Haidt 2007). The human majortransition was a rare event, but once accom-plished, our ability to function as team playersin coordinated groups enabled our species toachieve worldwide dominance, replacingother hominids and many other species alongthe way. The parallels with the other majortransitions are intriguing and highly instruc-tive (E O Wilson and Holldobler 2005).

A common scenario for human evolutionbegins with the evolution of sophisticatedcognitive abilities, such as a theory of mind,which in turn enabled widespread coopera-tion (Tomasello 1999). Now it appears morereasonable for the sequence to be reversed(Tomasello et al. 2005). Our capacities forsymbolic thought and the social transmissionof information are fundamentally communal

activities that probably required a shift in thebalance between levels of selection beforethey could evolve. Only when we could trustour social partners to work toward sharedgoals could we rely upon them to share mean-ingful information. The shift in the balancebetween levels of selection is reflected in an-

atomical features, such as the human eye asan organ of communication (Kobayashi andKohshima 2001), and basic cognitive abili-ties, such as the ability to point things out toothers (Tomasello et al. 2005) and to laughin a group context (Gervais and D S Wilson2005), in addition to more advanced cogni-tive and cultural activities associated withour species.

Group selection is an important force inhuman evolution in part because culturalprocesses have a way of creating phenotypicvariation among groups, even when they arecomposed of large numbers of unrelated in-dividuals. If a new behavior arises by a geneticmutation, it remains at a low frequency withinits group in the absence of clustering mech-anisms such as associations among kin. If anew behavior arises by a cultural mutation, itcan quickly become the most common be-havior within the group and provide the de-cisive edge in between-group competition(Richerson and Boyd 2005). The importanceof genetic and cultural group selection in hu-man evolution enables our groupish natureto be explained at face value. Of course,within-group selection has only been sup-pressed, not entirely eliminated. Thus multi-level selection, not group selection alone,provides a comprehensive framework for un-derstanding human sociality.

These ideas can potentially explain thebroad sweep of recorded history in addition tothe remote past. According to Turchin (2003,2005), virtually all empires arose in geograph-ical areas where major ethnic groups cameinto contact with each other. Intense between-group conflict acted as a crucible for the cul-tural evolution of extremely cooperative soci-eties, which then expanded at the expense ofless cooperative societies to become major em-pires. Their very success was their undoing,however, as cultural evolution within the em-pire led to myriad forms of exploitation, freeriding, and factionalism. That is why the cen-

ter of the former Roman empire (for exam-ple) is today a cultural black hole as far asthe capacity for cooperation is concerned.Turchin, a theoretical biologist who special-izes in nonlinear population dynamics, hasmarshaled an impressive amount of empiricalevidence to support his thesis about the rise


18/23


and fall of empires as a process of multilevelcultural evolution, with profound implica-tions for interactions among modern culturesand their consequences for human welfare inthe future.

A New Consensus and TheoreticalFoundation for Sociobiology

Making a decision typically involves en-couraging diversity at the beginning to eval-uate alternatives, but then discouraging di-versity toward the end to achieve closure andto act upon the final decision. It can be verydifficult to revisit an important decision thathas been made and acted upon, but that isprecisely what needs to be done in the caseof the 1960s consensus about group selection.

Historians of science have made a start, in-cluding a recent article appropriately titledThe Rise, Fall, and Resurrection of GroupSelection (Borrello 2005; see also Okasha2006), but the real need is for practicing so-ciobiologists to arrive at a new consensusbased on the many developments that havetaken place during the last four decades.

In concluding this article, it is interestingto revisit the contradictory positions that existin the current sociobiological literature:

Nothing has changed since the 1960s. Anexample is Alcocks (2005) influentialtextbook Animal Behavior: An EvolutionaryApproach, in which group selection is de-scribed as non-Darwinian and a near im-possibility because of the insuperableproblem of selection within groups.There is no excuse for this kind of treat-ment, given the developments over thelast four decades that we have reviewedin this article.

Multilevel selection theory (includinggroup selection) has been fully revived.It is important to stress once again thatthis is not a return to nave group selec-tionism. On the contrary, going back to

basics means affirming key elements ofthe consensus that formed in the 1960s,which insisted that higher-level adapta-tions require a process of higher-level se-lection and cannot be expected to evolveotherwise. The revival of multilevel selec-tion is based solely on rejecting the em-

pirical claim that higher-level selectioncan be categorically ignored as an im-portant evolutionary force. It is notablethat key figures such as Williams (for sexratio and disease virulence), Hamilton(in terms of the Price equation), andMaynard Smith (for major transitions ofevolution) easily reverted back to multi-level selection when they became con-vinced that group selection might be asignificant evolutionary force after all. Itis time for everyone to follow suit, for so-ciobiology as a whole rather than specificsubject areas.

There is a new multilevel selectiontheory that bears little relationship tothe old theory. According to RichardDawkins (quoted in Dicks 2000:35)[e]normous credit would accrue to any-one who could pull off the seemingly im-possible and rehabilitate group selection. . . [b]ut actually, suchrehabilitation cantbe achieved, because the great heresy re-ally is wrong. Yet, theoretical biologistswidely agree that modern multilevel selec-tion is a legitimate theory for accountingfor evolutionary change. The only way tomaintain these two positions is by claim-ing that modern multilevel selection the-ory bears no relationship to its own past(e.g., Keller 1999; West et al. 2006, 2007).We hope that our back to basics ap-proach has established the continuity ofideas, from Darwin to the present. More-over, other than avoiding nave group se-lection, all of the major conclusionsabout group selection that seemed toemerge during the 1960s, such as the re-jection of Wynne-Edwards hypothesis,need to be reconsidered on the basis ofongoing research.

Avoiding of the topic of group selection,as if it never existed in the history of evo-lutionary thought. We could cite dozensof theoretical and empirical articles from

the current literature that describe selec-tion within and among groups withoutmentioning the term group selectionor anything else about the group selec-tion controversy. As one example, the mi-crobial experiment by Kerr et al. (2006)elegantly establishes the plausibility of


19/23


Wynne-Edwards hypothesis and de-scribes the process matter-of-factly interms of selection within and amonggroups, without citing Wynne-Edwardsor the term group selection. This politesilence enables authors such as West etal. (2006) to publish tutorials on socialevolution for microbiologists that por-tray Wynne-Edwardss hypothesis as atheoretical impossibility. This kind ofpluralism is not helpful (D S Wilson2007a). We hope that our article will helpto refocus attention on the problem thathas always been at the center of multi-level selection theory: the fact thatgroup-level adaptations are seldom lo-cally advantageous and, therefore, mustbe favored at a larger scale to evolve. Thefact that all theoretical frameworks re-flect this problem and its (partial) solu-

tion is a major simplification that shouldbe welcomed rather than resisted.

When Rabbi Hillel was asked to explain the

Torah in the time that he could stand on onefoot, he famously replied: Do not do untoothers that which is repugnant to you. Every-thing else is commentary. Darwins originalinsight and the developments reviewed in thisarticle enable us to offer the following one-foot summary of sociobiologys new theoreti-cal foundation: Selfishness beats altruismwithin groups. Altruistic groups beat selfishgroups. Everything else is commentary.

acknowledgments

We thank Mark Borrello, Samuel Bowles, Anne Clark,Michael Doebeli, Jeff Fletcher, Bert Holldobler, Kevin

Foster, Benjamin Kerr, Samir Okasha, David Queller,Francis Ratnieks, Elliott Sober, Tom Wenseleers, andMartin Zwick for helpful discussion.

REFERENCES

Aktipis C A. 2004. Know when to walk away: contin-gent movement and the evolution of cooperation.

Journal of Theoretical Biology231(2):249260.Alcock J. 2005. Animal Behavior: An Evolutionary Ap-

proach. Eighth Edition. Sunderland (MA): Sinauer.Allee W C, Emerson A E, Park O, Park T, Schmidt K

P. 1949. Principles of Animal Ecology. Philadelphia(PA): W. B. Saunders.

Aviles L. 1993. Interdemic selection and the sex ratio:a social spider perspective. American Naturalist142(2):320345.

Aviles L. 2002. Solving the freeloaders paradox: ge-netic associations and frequency-dependent selec-tion in the evolution of cooperation among non-relatives. Proceedings of the National Academy of Sciences of the United States of America99(22):1426814273.

Aviles L, Abbot P, Cutter A D. 2002. Population ecol-ogy, nonlinear dynamics, and social evolution. I.

Associations among nonrelatives. American Natu-ralist159(2):115127.

Aviles L, Fletcher J A, Cutter A C. 2004. The kin com-position of social groups: trading group size fordegree of altruism. American Naturalist164(2):132144.

Bingham P M. 1999. Human uniqueness: a generaltheory. Quarterly Review of Biology74(2):133169.

Boehm C. 1999. Hierarchy in the Forest: The Evolution ofEgalitarian Behavior. Cambridge (MA): HarvardUniversity Press.

Borrello M E. 2005. The rise, fall, and resurrection ofgroup selection. Endeavour29:4347.

Bourke A F G, Franks N R. 1995. Social Evolution inAnts. Princeton (NJ): Princeton University Press.

Bowles S. 2006. Group competition, reproductive lev-eling, and the evolution of human altruism. Science314:15691572.

Boyd R, Richerson P J. 1985. Culture and the Evolution-ary Process. Chicago (IL): University of ChicagoPress.

Boyd R, Richerson P J. 1992. Punishment allows theevolution of cooperation (or anything else) in siz-

able groups. Ethology and Sociobiology 13(3):171195.

Bull J J. 1994. Virulence. Evolution48(5):14231437.Burland T M, Bennett N C, Jarvis J U M, Faulkes C G.

2002. Eusociality in African mole-rats: new insightsfrom patterns of genetic relatedness in the Da-maraland mole-rat. Proceedings of the Royal Society ofLondon Series B 269:10251030.

Burt A, Trivers R. 2006. Genes in Conflict: The Biology ofSelfish Genetic Elements. Cambridge (MA): BelknapPress of Harvard University Press.

Cannon W B. 1932. The Wisdom of the Body. First Edi-tion. New York: W. W. Norton.

Choe J C, Crespi B J, editors. 1997. The Evolution ofSocial Behavior in Insects and Arachnids. Cambridge(UK): Cambridge University Press.

Cronk L. 2005. The application of animal signalingtheory to human phenomena: some thoughts andclarifications. Social Science Information/Informationsur les Sciences Sociales 44:603620.

Darwin C. 1871. The Descent of Man, and Selection inRelation to Sex, Volumes 1 and 2. New York: Apple-ton.


20/23


Dawkins R. 1976. The Selfish Gene. New York: Oxford

University Press.

Dawkins R. 1982. The Extended Phenotype: The Gene as

the Unit of Selection. San Francisco (CA): Freeman.

Dicks L. 2000. All for one! New Scientist167(2246):30.Dugatkin L A. 2002. Cooperation in animals: an evo-

lutionary overview. Biology and Philosophy 17(4):

459476.

Dugatkin L A, Reeve H K. 1994. Behavioral ecology

and levels of selection: dissolving the group selec-

tion controversy. Advances in the Study of Behavior

23:101133.

Fanelli D, Turillazzi S, Boomsma J J. 2004. Genetic

relationships between females, males, and olderbrood in the tropical hover wasp Parischnogastermellyi(Saussure). Insect Social Life 5:3540.

Fletcher J A, Doebeli M. 2006. How altruism evolves:

assortment and synergy. Journal of Evolutionary Bi-

ology19(5):13891393.

Fletcher J A, Zwick M. 2004. Strong altruism canevolve in randomly formed groups. Journal of Theo-

retical Biology228(3):303313.Fletcher J A, Zwick M. 2006. Unifying the theories of

inclusive fitness and reciprocal altruism. American

Naturalist168(2):252262.

Fletcher J A, Zwick M, Doebeli M, Wilson D S. 2006.

Whats wrong with inclusive fitness? Trends in Ecol-

ogy & Evolution21(11):597598.

Foster K R. 2006. Balancing synthesis with pluralism in

sociobiology. Journal of Evolutionary Biology 19(5):

13941396.Foster K R, Parkinson K, Thompson C R L. 2007.What

can microbial genetics teach sociobiology? Trends

in Genetics23(2):7480.

Foster K R, Ratnieks F L W. 2005. A new eusocial ver-tebrate? Trends in Ecology & Evolution 20(7):363

364.

Foster K R, Wenseleers T, Ratnieks F L W. 2006a. Kin

selection is the key to altruism. Trends in Ecology &

Evolution21(2):5760.

Foster K R, Wenseleers T, Ratnieks F L W, Queller D

C. 2006b. There is nothing wrong with inclusive

fitness. Trends in Ecology & Evolution 21(11):599

600.

Frank S A. 1996. Models of parasite virulence. Quar-

terly Review of Biology 71(1):3778.

Gervais M, Wilson D S. 2005. The evolution and func-

tions of laughter and humor: a synthetic approach.

Quarterly Review of Biology80(4):395430.

Gilpin M E. 1975. Group Selection in Predator-PreyCommunities. Princeton (NJ): Princeton University

Press.

GintisH. 2000. Game Theory Evolving: A Problem-Centered

Introduction to Modeling Strategic Interaction. Prince-

ton (NJ): Princeton University Press.

Goodnight C J. 2000. Heritability at the ecosystem

level. Proceedings of the National Academy of Sciences ofthe United States of America97(17):93659366.

Goodnight C J. 2005. Multilevel selection: the evolu-tion of cooperation in non-kin groups. Population

Ecology47(1):312.Goodnight C J, Schwartz J M, and Stevens L. 1992.

Contextual analysis of models of group selection,soft selection, hard selection, and the evolution ofaltruism. American Naturalist140(5):743761.

Goodnight C J, Stevens L. 1997. Experimental studiesof group selection: what do they tell us aboutgroup selection in nature? American Naturalist150(Supplement):S59S79.

Haidt J. 2007. The new synthesis in moral psychology.Science316:9981002.

Hamilton W D. 1963. The evolution of altruistic be-havior. American Naturalist97(896):354356.

Hamilton W D. 1964. The genetical evolution of socialbehavior, I and II. Journal of Theoretical Biology7(1):

152.Hamilton W D. 1975. Innate social aptitudes in man,an approach from evolutionary genetics. Pages133155 in Biosocial Anthropology, edited by R Fox.London: Malaby Press.

Hamilton W D. 1996. Narrow Roads of Gene Land: TheCollected Papers of W. D. Hamilton. Oxford: W. H.Freeman/Spektrum.

Hammerstein P, editor. 2003. Genetic and Cultural Evo-lution of Cooperation. Cambridge (MA): MIT Press.

Heinsohn R, Packer C. 1995. Complex cooperativestrategies in group-territorial African lions. Science269:12601262.

Holldobler B, Wilson E O. 1990. The Ants. Cambridge(MA): Belknap Press of Harvard University Press.

Jablonka E, Lamb M J. 2006. The evolution of infor-

mation in the major transitions. Journal of Theoreti-cal Biology239(2):236246.

Johnson C R, Boerlijst M C. 2002. Selection at the levelof the community: the importance of spatial struc-ture. Trends in Ecology & Evolution17(2):8390.

Keller L, editor. 1999. Levels of Selection in Evolution.Princeton (NJ): Princeton University Press.

Kerr B, Godfrey-Smith P. 2002a. Individualist andmulti-level perspectives on selection in structured

populations. Biology and Philosophy 17(4):477517.Kerr B, Godfrey-Smith P. 2002b. Group fitness and

multi-level selection: replies to commentaries. Bi-ology and Philosophy17(4):539549.

Kerr B, Neuhauser C, Bohannan B J M, Dean A M.2006. Local migration promotes competitive re-

straint in a host-pathogen tragedy of the com-mons. Nature442:7578.

Kobayashi H, Kohshima S. 2001. Unique morphologyof the human eye and its adaptive meaning: com-parative studies on external morphologyof the pri-mate eye. Journal of Human Evolution 40(5):419435.


21/23


Lachmann M, Szamado S, Bergstrom C T. 2001. Cost

and conflict in animal signals and human lan-

guage. Proceedings of the National Academy of Sciences

of the United States of America98(23):1318913194.

Landi M, Queller D C, Turillazzi S, Strassmann J E.2003. Low relatedness and frequent queen turn-

over in the stenogastrine waspEustenogaster fraterna

favor the life insurance over the haplodiploid hy-

pothesis for the origin of eusociality. Insectes Sociaux

50(3):262267.

Lumsden C J, Wilson E O. 1981. Genes, Mind, and Cul-

ture: The Coevolutionary Process. Cambridge (MA):

Harvard University Press.

Macdonald K S, Ros R, Duffy J E. 2006. Biodiversity,host specificity, and dominance by eusocial species

among sponge-dwelling alpheid shrimp on the Be-

liz

Documents

Rethinking Sociobiology