One of the key elements of Gowdy & Krall's (G&K's) model of ultrasociality is the idea that group selection plays a central role in its evolution. I would like to expand on their discussion of multilevel selection, and to argue that agriculture, whether by humans or social insects, has likely involved multi-species community selection as well as single- species group selection.

First off, it is important to recognize that group selection and community selection work. This was first demonstrated in Wade's (Reference Wade1977) landmark study on group selection in Tribolium flour beetles. Since that time, there have been numerous studies demonstrating the effectiveness of group selection (reviewed in Goodnight & Stevens Reference Goodnight and Stevens1997), and multilevel selection has become an important tool for animal breeding (e.g., Wade et al. Reference Wade, Bijma, Ellen and Muir2010). There has also been one study showing group selection acting in human populations (Moorad Reference Moorad2013). Finally, there are several studies showing a response community selection in two species communities (Goodnight Reference Goodnight1990a; Reference Goodnight1990b) and in soil and aquatic microbial communities (Swenson et al. Reference Swenson, Arendt and Wilson2000a; Reference Swenson, Wilson and Elias2000b).

One surprising thing about group selection experiments is how effective they are. For example, Muir (Reference Muir1996), using group selection to increase egg production in chickens, observed a 160% increase in egg production in the selected lines. The cause of the effectiveness of group selection is known to be primarily due to indirect genetic effects (IGEs) (Bijma & Wade Reference Bijma and Wade2008). IGEs are defined as phenotypic effects in one individual due to genes in another individual. Muir's chickens provide a good example of how these work. In his experiment, Muir, rather than selecting on the chickens that produced the most eggs, selected on the cages that produced the most eggs. Chickens are famous for having a pecking order; the most aggressive chickens get the majority of the food and lay the most eggs. The more subservient chickens get less food, lay fewer eggs, and get harassed by more dominant chickens, although they generally do survive if they can run away. In cages, however, the subservient chickens cannot run away, and they often get pecked to death. Individual selection will favor the most aggressive chickens, because they lay more eggs, resulting in more antagonistic interactions, and heightened mortality of subordinate chickens.

Muir, however, by selecting on the productivity of the cage favored those groups of chickens in which food was more equally shared, there was less mortality, and overall more eggs were laid. This illustrates the importance of IGEs in multilevel selection. Group and community selection can act on IGEs and bring about adaptations that are qualitatively different from adaptations that can evolve by individual selection. In the chicken example, individual selection can only act on direct genetic effects – thus, it will favor chickens that more aggressively dominate resources, regardless of its effects on other individuals. Group selection, however, will favor groups of chickens that overall lay the most eggs. Selection for greater overall egg production favors less aggressive chickens, which ensures that all individuals have adequate resources to contribute to the total output of eggs. These indirect effects also contribute to the response to community selection. For example, Goodnight (Reference Goodnight1990b) showed that the response to community selection in two species communities of Tribolium castaneum and T. confusum depended on maintaining the genetic structure of the community.

Ultrasociality concerns the evolution of complex societies with a high degree of specialization of society members. Individual selection cannot by itself lead to the evolution of such complex sociality. Group and community selection can act on indirect genetic effects, and the specialization of the members of a society can be considered an example of IGEs, or in the case of humans, indirect cultural effects. Thus, we can argue that group selection may be essential for the evolution of ultrasociality.

Particularly intriguing is the possibility of community selection being important in the evolution of ultrasociality. As G&K point out, Atta (leafcutter ants) and their fungi are closely coevolved, with the ants being dependent on the fungi as a source of food, and the fungi dependent on the ants for food and protection from pathogenic bacteria. Similarly, Meso-American agriculture was dependent on maize as the basis of the inhabitants' diet, and maize is a derived crop that cannot survive without human intervention (Landon Reference Landon2008). Such coevolution may well be the result of selection acting directly on the community (Goodnight Reference Goodnight1990a). For example, in an Atta colony, an infected fungus may lead to death of not only the fungus, but also the ants that depend on it. Those colonies in which the fungus is resistant to bacterial infection, whether due to intrinsic resistance of fungus or the behavior of the ants, will have a higher overall fitness. Similar scenarios can be imagined for humans and their crops.

Thus, G&K are correct to suggest the role of multilevel selection in the evolution of ultrasociality; however, we should also consider the possible role of community selection.