The exquisite corpse: a shifting view of the shifting balance.

In 1925, a game called ‘exquisite corpse’ was invented by a group of surrealists to harness the random power of imagination through collaboration. Working either with words or drawings, an exquisite corpse was begun by one member of the group, who then folded the paper so that his or her contribution was obscured and then passed it to the next member. The game (named after one of their first results – Le cadavre exquis boira le vin nouveau) allowed the random words or images to be combined, followed by selection of the best results for preservation and dissemination (although they de-emphasized the latter). Therefore, the game serves as an acceptable metaphor for the shifting-balance process. After a recent flurry of papers2–7, we might pronounce the strict adaptationist version of the shifting-balance theory (SBT) both ‘exquisite’ and a ‘corpse’, although the authors of these papers would probably only agree with one or the other of these assessments. In 1931, Sewall Wright8 proposed a theory, which would come to be known as his shifting-balance theory. This began with the idea that interactions among loci could result in populations achieving a genotypic state that was locally relatively fit (a ‘peak’), but was not as fit as possible because of intervening unfit genotypes (‘valleys’). Wright9 suggested several mechanisms by which evolution might allow a population to reach higher adaptive peaks, including novel favorable mutations, relaxation of selection, qualitative changes in the environment and the model now known as the shiftingbalance theory. The shifting-balance process, in its strict interpretation, imagined that species were subdivided into many demes weakly connected by migration. Demes might be small enough that genetic drift can sometimes overwhelm the effects of natural selection and take the population to the domain of attraction of a new peak (i.e. the allele frequencies can drift to a point where the deterministic effects of selection would be expected to take the population to a new peak; this is referred to as Phase I). Individual selection could then take that population near to the height of the new peak (Phase II), at which time Wright envisioned intergroup selection would act to pull the whole species towards the new peak (Phase III). This story has served as a significant theory of adaptive evolution for the greater part of the 20th century. Three years ago, a significant challenge to the SBT came from Jerry Coyne, Michael Turelli and Nick Barton2, in the longest paper ever published in Evolution, where they reviewed the evidence against SBT. At the risk of oversimplification, their main arguments were: (1) Phases I and II are possible, but drift is often not necessary for peak shifts; (2) Phase III is unlikely to successfully spread new adaptive peaks to other demes; and (3) there are no known examples of the shifting-balance process going through all the steps envisioned by Wright. Furthermore, none of the partial examples can demonstrate an increase in adaptation as a result of shifting balance. They argue that parsimony would lead one to choose Fisherian mass selection as the simpler, and experimentally confirmed, explanation for the majority of adaptations. Although Coyne et al. allow for the potential occurrence of the SBT, they strongly reject it as an explanation for a significant proportion of adaptation. Coyne et al.2 is also notable for its appendix, in which the distribution of allele frequencies among populations with drift, selection, mutation and migration, one of Wright’s most enduring contributions to population genetics10, was used to address the plausibility of the SBT itself. Subsequently, Michael Wade and Charles Goodnight3 have, in the same forum, published a defense of Wright’s perspective on evolution. Although not a direct response to Coyne et al., Wade and Goodnight argue that evolution is most effective ‘when nature does many small experiments’; that is, when species are subdivided into a metapopulation. This perspective has prompted a debate5,6 over the adaptive importance of epistasis, group selection, population structure and drift. Wade and Goodnight argue that the evolutionary process is strongly affected by epistasis; that group selection is much more effective than additive models would predict; and that population subdivision allows for multiple semi-independent evolutionary trials and for more effective adaptation and speciation. These papers, and the recent round of rebuttals and responses5–7, disagree greatly; however, they largely address different questions. Coyne et al.2,5 are focused on the plausibility of a narrowly defined version of SBT, while Wade and Goodnight3,6 have more broadly considered the role of ‘Wrightian’ modes of evolution without addressing important criticisms of the SBT. To some extent, the two sides are talking past each other and reading them sequentially is a bit like viewing the unfolded picture after drawing an exquisite corpse. We agree with Coyne et al. that showing that each of the components of SBT is plausible is not equivalent to showing that SBT is a major factor in adaptation, but it would be premature to entirely dismiss the possibility of the SBT (indeed, as Coyne et al.2 have not done; see their final paragraph). Intergroup selection can be a weak force, thus making Phase III difficult, but Coyne et al. have not presented a convincing case that it is impossible – in fact, most of the papers cited by Coyne et al.2 have shown the efficacy of Phase III. Goodnight and Wade6 point out that there are many effective demonstrations of the efficacy of group selection; but, Coyne et al.5 argue that these studies are not controlled for individual selection. However, all that is required for SBT to function is that interdemic selection is possible – it is not necessary for interdemic selection to be faster than individual selection. Perhaps more germane is that group selection might not be sufficient to overwhelm the effects of back-migration of old peak genotypes into the newly shifted population11,12. This causes a particular difficulty of Phase III in the island model, where a newly shifted deme has only a weak influence on all other demes, but collectively the other demes can exert strong pressure via migration for that deme to return to the original genotype. (Note, that Wright’s distribution assumes an island model, thus the case explicitly modeled by the appendix1 is certainly not the most probable scenario for the SBT to work. Strangely, given their interpretation, the calculations in the appendix of Coyne et al.2 demonstrate a substantial parameter range under which SBT would be effective.) As others have pointed out4,7,13, models of population structure with isolation by distance are NEWS & COMMENT