"Group selection" is the controversial theory that natural selection operates not just at the individual level but also at the level of the social group. I recently participated in a conversation at Edge.org about the utility of this theory. The conversation included many reasonable people but their opinions were quite polarized: some saw the theory as useless, others as essential. Much of the debate focused on group selection’s relevance to human cooperation, and here I’m going to try and make some progress on that topic.

I’ll focus on the debate’s thorniest question: is a trait that disadvantages an organism compared to others within the organism’s group, but that advantages the organism’s group in competition with other groups, likely to be favored by selection? I’ll argue that although the selection of such traits is a mathematical certainty under a specific set of assumptions, in reality these traits would be pretty lousily designed from an adaptationist standpoint, and are unlikely to characterize human nature.

First, it’s important to understand how an individually-disadvantageous, group-advantageous (IDGA) trait could evolve. Imagine a population of organisms that is subdivided into numerous groups. In one of these groups, individual X has a mutation for an IDGA trait, which causes X to benefit the group at X’s own expense. The group’s mean fitness goes up, so the gene is favored by between-group selection. But X’s own fitness goes down, so the gene is disfavored by within-group selection. Whether or not the gene increases or decreases in frequency in the overall population depends on the relative strengths of between-group versus within-group selection. To point that out is simply to state a mathematical truism; if certain assumptions are met about the effects of the mutation on the fitnesses of X and X’s group members, and the effects of individual versus group selection, then the IDGA trait can evolve. Somewhat extravagantly, this truism has even been incarnated in laboratory studies. For example, in a study of E. coli [1], individual bacteria who incurred costs to bestow antibiotic resistance on their neighbors were shown to be disfavored within groups, favored between groups, and capable of proliferating in the general population.

Could genes for an IDGA trait have been favored in humans in a similar fashion? It’s unlikely, not because of any faulty mathematical assumptions about how the evolution of an IDGA trait could be modelled, but because IDGA traits are of relatively crude adaptive design. IDGA traits cater to between-group selection pressures, but fail to exploit within-group pressures; they have no ability to harness what they could (in principle) use as a valuable resource for propelling themselves into the next generation. The more powerful within-group selection is, the more egregious is their failure to utilize it. And virtually everyone agrees that within-group selection has been a formidable force in human evolution.

IDGA traits pay to shower benefits on “free riders” who take these benefits without sharing the costs; that’s why IDGA traits are disadvantageous within groups. Therefore, individual selection would favor any mutation that made an IDGA trait slightly more strategic, and less indiscriminate, about who its beneficiaries were. If the IDGA carrier could do anything to increase the probability that its beneficiaries would be other IDGA carriers, and not just randomly-selected individuals, this would improve IDGA genes’ ability to exploit the power of individual selection. That’s true because the first rule of individually-adaptive cooperation is cooperate with other cooperators (and its corollary is avoid free riders); the more an altruistic gene can assort positively with other altruistic genes, the better it will replicate. This is the principle underlying Hamiltonian genic self-favoritism (e.g. kin selection or Dawkins’ “green beard” scenario), and its mathematics are no less self-evident than those discussed above regarding IDGA genes.

E. coli cells that produce antibiotic resistance are passive altruists; they favor whoever happens to be their neighbor and exercise no partner choice. Sometimes this might be the best that evolution can do, and when between-group selection is strong it could be better than nothing. But evolution has done much better than this with humans. People are highly strategic cooperators. They aren’t just dropped into groups at random, to indiscriminately bestow benefits on accidental neighbors, like E. coli in a petri dish. They create groups, they form collective actions, and they join coalitions. They are highly strategic in their choice of cooperative partners and in their decisions about how much to cooperate. In forming a group, for example, cooperators preferentially select other cooperators as partners and flee free riders; and after forming a group, they closely monitor partners' contribution levels and stop cooperating in response to (and/or punish) free riding. In a 2006 paper [2], I review the evidence for these kinds of highly strategic cooperative behaviors in humans, and suggest that it supports the view that individuals are exquisitely designed not only to harvest the benefits of cooperation in groups, but also to minimize the costs of being exploited by free riders.


  1. Chuang J. S., Rivoire O., Leibler S. (2009). Simpson's paradox in a synthetic microbial system. Science 323: 272-275.
  2. Price M. E. (2006). Monitoring, reputation and “greenbeard” reciprocity in a Shuar work team. Journal of Organizational Behavior 27: 201-219.

Copyright Michael E. Price 2012. All rights reserved. 

You are reading

From Darwin to Eternity

Did Religion Evolve to Benefit the Weak or the Powerful?

The paradox of who benefits most from religion

Surprised These Rockers Are Still Alive?

Why don’t we celebrate stars more while they’re still with us?

Star Wars, Christmas, and Constraints on Secular Influence

What's so special about religion (and Star Wars and Christmas)?