Group Selection A Gardner, University of Oxford, Oxford, UK © 2013 Elsevier Inc. All rights reserved. This article is a revision of the previous edition article by MJ Wade, volume 2, pp 894–899, © 2001, Elsevier Inc. Glossary individual’s character on her own fitness) and indirect ’ Adaptation The appearance of design in the living world, (impact of an individual s character on the fitness of her originally attributed to deities and now known to be due genetic relations) components. to natural selection. Levels of selection An approach to group selection Between-group selection The part of natural selection theory that defines group selection as that component that owes either to the impact of the group character upon of natural selection that owes to fitness differences individual fitness (contextual-analysis approach) or to the between groups. differential fitness of groups (levels-of-selection Natural selection The component of evolutionary change approach). in heritable characters that owes to fitness differences Contextual analysis An approach to group selection between individuals (within and between groups, and theory that defines group selection as that component of mediated by their own characters and those of their social natural selection ascribable to the impact of the group groups). character upon individual fitness. Superorganism The idea that a social group can be Group adaptation The appearance of design at the group considered an adapted organism in its own right, which is level, driven by between-group selection, that obtains only valid only in the absence of conflicts of interest within the when within-group selection is negligible. group. Inclusive fitness The quantity that individuals appear Within-group selection The part of natural selection that ’ designed to maximize as a consequence of the action of owes either to the impact of the individual s character natural selection. upon her own fitness (contextual-analysis approach) or to Kin selection theory A theory of social evolution that the differential fitness of individuals within groups partitions natural selection into its direct (impact of an (levels-of-selection approach). The Origin of Group Selection selection, which for many years was considered empirically dis- tinct from the kin selection hypothesis. Interestingly, Darwin As with many of the big ideas in evolutionary biology, the concept himself did not clearly distinguish the two ideas, pointing to of group selection has its origins in the writings of Charles Darwin. blood relationship between group mates as a key factor driving Darwin discussed problems of social evolution in The Origin of the individual’s incentive to act for the good of her group. Species, focusing upon the paradox presented by the sterile castes The idea of group selection has sometimes been attributed to of social insect colonies. Here, the problem is to explain how the the co-discoverer of natural selection, Alfred Russel Wallace. exquisite adaptations exhibited by these individuals (and, indeed, Indeed, Peter Bowler has argued that a striking difference between that are unique to them) could be molded by natural selection, Darwin’s and Wallace’s papers on natural selection, presented at given that these individuals do not produce offspring. His solution the meeting of the Linnean Society in 1858, is that the former was to notice that kin tend to share heritable characters in com- describes natural selection at the level of individuals, while the mon such that, for example, a cattle breeder who slaughters an latter describes natural selection at the level of groups or ‘vari- animal and finds that it has particularly good meat can select for eties’. However, Michael Bulmer has pointed out that Wallace’s this character by preferentially breeding from the close relatives of notion of a variety simply corresponds to a type – such as a the slaughtered individual. Hence, sterile workers can evolve adap- white-flowering variety within a polymorphic plant species – tations, provided these improve the fitness of their fertile family and that Wallace’s conception of the theory of natural selection members. This is the theory of kin selection, which was more fully is not substantively different from Darwin’sinthisrespect. elaborated a century later by W. D. Hamilton. Certainly, the idea that selection at the group level might counter In the Descent of Man, Darwin took an alternative approach to and even overpower selection within groups is due to Darwin. understanding social traits. He noted that those human beha- viors considered moral typically place the individual at a fitness disadvantage relative to her social partners, but benefit the social The Modern Theory of Group Selection group as a whole. Darwin suggested that these behaviors could have been driven by natural selection acting at the group level. Today, we – like Darwin – understand that the kin selection Groups of ancestral humans would have come into conflict over and group selection approaches to social evolution are exactly limiting resources, and Darwin suggested that the resulting equivalent. Both describe the same evolutionary process, that selection for the most cooperative groups would often have is, natural selection, but split up its action in somewhat differ- overpowered the selection operating against such cooperation ent ways. The kin selection approach splits natural selection within groups. This was the launch of the theory of group into the direct effect of a heritable character on the individual’s 362 Brenner’s Encyclopedia of Genetics, 2nd Edition, Volume 3 doi:10.1016/B978-0-12-374984-0.00663-X Group Selection 363 own reproductive success and the indirect effect of the character has been devised that partitions Price’sequationinaslightly on the reproductive success of the individual’s genetic relatives. different way. This is the contextual-analysis approach, and it The group selection approach instead splits natural selection splits up eqn [1] as follows: into its within-group and between-group components. ÀÁÀÁÀÁÀÁ Δ ðÞ¼β ; j ; þ β ; j ; Irrespective of how one chooses to split up the components NSEg vij gij gi cov gij gij vij gi gij cov gi gij of natural selection, they will always add to give the same sum, ½3 and so kin selection and group selection theories yield exactly where β terms are partial regression coefficients obtained by the same predictions about social evolution. fitting the linear multiregression model E(v |g and g )=E(v )+ However, there is some disagreement as to exactly how the ij ij i ij β(v ,g |g )(g – E(g )) + β(v ,g |g )(g – E(g )) + ε to the popula- group selection split should be made. The first – and most ij ij i ij ij ij i ij i ij ij tion (g ,g,v ) data by means of least squares (i.e., minimizing popular – approach to the theory of group selection is that ij i ij the quantity E(ε 2)). The β(v ,g |g ) term describes the effect of pioneered by George Price and subsequently developed by ij ij ij i the individual’s character value on her own fitness, holding Hamilton, which has been termed the ‘levels-of-selection’ fixed any effect of her group’s average character value, and so approach. This takes Price’s general equation of natural selec- the quantity β(vij,gij|gi)cov(gij, gij) has been interpreted as a form tion as a starting point: of within-group selection. Conversely, the β(vij,gi|gij) term ’ ΔNSEðgÞ¼covðv; gÞ½1 describes the effect of the group s average character value on the individual’s fitness, holding fixed any effect of her own Price’s equation expresses the change in the average value of character value, and so the quantity β(vij,gi|gij)cov(gi,gij) has any heritable character between consecutive generations, been interpreted as a form of between-group selection. under the action of natural selection, as being equal to the The contextual-analysis approach avoids the problem of covariance of relative fitness v and character value g over all diagnosing group selection for nonsocial traits. However, it the individuals in the population. Assigning individuals to does diagnose the action of group selection even in scenarios groups (for simplicity, I assume that there is an equal number where there are no fitness differences between groups. For exam- of individuals in every group), and assigning every group a ple, if density-dependent regulation maintains all groups at a unique index i ∈ I and every individual within each group a fixed size, then individuals with defective livers may achieve high unique index j ∈ J, eqn [1] can be partitioned into the fitness, provided that their group mates also have defective livers, between-group and average within-group covariance between but will achieve lower fitness if their group mates primarily have relative fitness and character value: properly functioning livers. Contextual analysis suggests that group selection is acting here and, moreover, that it favors Δ EðgÞ¼cov ðv ; g ÞþE ðcov ðv ; g ÞÞ ½2 NS I i i I J ij ij defective livers, even though they bring no advantage to the where vi = EJ(vij) and gi = EJ(gij) are the average fitness and char- group as a whole. acter value, respectively, among the individuals belonging to As a consequence of these perceived failures of levels- the ith group. of-selection and contextual-analysis approaches to the theory of The first term on the right-hand side (RHS) of eqn [2] is the group selection, some authors - including David Wilson and covariance of group fitness and group character value, taken Edward Wilson - have suggested that group selection defies for- over all the groups in the population, and defines the action of mal analysis altogether. However, a more pragmatic outlook ‘between-group’ selection. The second term on the RHS of eqn recognizes that the alternative approaches address somewhat [2] is the average, taken over all groups in the population, of different problems, and in a potentially useful way.