Group selection, again. Yay.
I was amused to see that David Sloan Wilson took a weird poke at Dawkins, got thrashed by Jerry Coyne, and didn’t like it. In fact, I was going to leave this as a link post, but while searching for a link to Coyne’s piece (Wilson can’t seem to figure out how to embed links to anything but his own blog in his posts) I came across a post by a blogger who calls him/herself “The Verbose Stoic”. This piece is problematic on several points, but discussing this is going to take some space so I’ll do it here instead of a comment on Verbose Stoic’s blog; from here on, I’m going to refer to him/her as VS.
VS starts off by questioning “examples”:
What I want to talk about is the objections that Coyne raises against Wilson’s group selection theory:
Dawkins’s argument against the efficacy of group selection was that this form of selection is usually unsuccessful because groups are vulnerable to subversion from within by those selfish replicators. That is, “cheating” replicators that are “good” for individuals but bad for the group as a whole will tend to propagate themselves. Yes, altruism may help groups propagate, but altruistic groups are susceptible to invasion by cheaters unless the “altruism” is based on kin selection or individual selection via reciprocity.
That’s the main one, but he goes on to fill in more later:
Dawkins’s (and my) beef with group selection as a way to evolve traits that are bad for individuals but good for groups is that this form of selection is inefficient, subject to subversion within groups, and, especially, that there’s virtually no evidence that this form of selection has been important in nature.
Let me deal with the two minor ones before getting back to the main event. Starting with the last one, we can see that it’s a bad argument, because what Coyne is doing here is saying that one of the reasons to reject the examples Wilson’s giving of cases where group selection has been important in nature is … that you haven’t found examples of cases where it has been important in nature. Except, perhaps, for the specific cases Wilson is citing. You can’t in any way reasonably claim that the fact that you haven’t found examples of it yet means that you can dismiss this proposed example.
Look, Wilson isn’t citing any specific cases of group selection occurring in nature, mostly because there aren’t any. When I say that, I mean that Wilson hasn’t been able to demonstrate that a trait arose because of group selection and not kin selection or natural selection or any other evolutionary force. Wilson’s argument is that (1) group selection (a.k.a. “new” group selection or multi-level selection) is something different than any other variety of selection, and (2) that it is responsible for the evolution of traits such as altruism. But (1) group selection reduces mathematically to inclusive fitness (follow the links in my previous post), and so (2) is trivially true. Sure, it arose by “group selection”, but that’s an empty statement. Wilson’s ‘proposed example’ is a theoretical model that was dealt with when he proposed it nearly 40 years ago (Wilson, 1975), and though it’s been refuted dozens of times since, he keeps holding on to it and insisting that he’s already won. I’ll quote at length from West et al. (2007) to drive home the point:
It has since been shown that kin selection and new group selection are just different ways of conceptualizing the same evolutionary process. They are mathematically identical, and hence are both valid (Hamilton, 1975; Grafen, 1984; Wade, 1985; Frank, 1986a, 1998; Taylor, 1990; Queller, 1992; Bourke & Franks, 1995; Gardneret al., 2007). New group selection models show that cooperation is favoured when the response to between group selection outweighs the response to within-group selection, but it is straightforward to recover Hamilton’s rule from this. Both approaches tell us that increasing the group beneﬁts and reducing the individual cost favours cooperation. Similarly, group selection tells us that cooperation is favoured if we increase the proportion of genetic variance that is between-group as opposed to within-group, but that is exactly equivalent to saying that the kin selection coefﬁcient of relatedness is increased (Frank, 1995a). In all cases where both methods have been used to look at the same problem, they give identical results (Frank, 1986a; Bourke & Franks, 1995; Wenseleers et al., 2004; Gardner et al.,2007).
VS also isn’t happy about “efficiency”:
The first one is also a pretty bad argument when you look at evolution. The argument is that Wilson’s proposed solution would be inefficient, but it seems to me that one of the main thrusts of evolution is that it can indeed be — and often is — inefficient but as long as it works, that’s not a problem. When has it become a criteria for evolutionary explanations that it achieve maximal or even reasonable efficiency. To go down that route would risk re-introducing a need for a designer, to ensure that the mechanisms stayed efficient. That can’t be what Coyne wants. But, again, why is efficiency even a factor? Why would you sort evolutionary arguments by efficiency? Being more or less efficient isn’t a hallmark of evolutionary mechanisms, so if two mechanisms are proposed but one is more efficient than the other that says absolutely nothing about which one is more likely to be true.
That leaves us with the main complaint: cheaters. The main issue here is that there is an issue raised against the individual selection explanations of altruism as well, even kin and reciprocal altruism and it is … cheaters. Cheaters will benefit if they can get away with it, and so those individuals will prosper and those who are altruistic will be outstripped, and so altruism is not self-sustaining at the individual level. To get around this, the proponents of evolutionary explanations for altruism end up appealing to cheater detection mechanisms […]
Additionally, it seems to me that group selection can actually get this without having to apply specific cheater detection mechanisms. After all, group selection would imply that the relevant competing entity is the group. Thus, if a group has a significant percentage of people who are altruistic, then it outperforms groups that don’t. Thus, if you have a group where this happens and where too large a percentage of the group are cheaters, then that group will cease to get those benefits and be outcompeted and presumably eventually exterminated by the groups where that does not happen. Thus, group selection here becomes self-sustaining; if you are above or at the magical percentage that means you benefit from being altruistic, you benefit over other groups as long as it stays there, but if it ever drops below that your group may well collapse and your individuals, then, all lose. Note that we would still see cheater detection mechanisms emerge because they are mechanisms that make the group stable and so less likely to fall below that percentage and collapse.
It seems like VS might be on the verge of confusing old and new school group selection, as the bolded statements (my emphasis) suggest. West et al.’s paper has a great figure that shows the difference between the two:
In the text of their article, they point out that “[a]nother way of looking at this is that the new group selection approach looks at the evolution of individual characters in a group structured population, whereas the old group selection approach looks at the evolution of group characters”. VS’s own words make him sound like a disciple of Wynne-Edwards, which would be unfortunate since Wynne-Edwards was decisively crushed by George Williams in the 1960s. But even if he’s just poorly recapitulating Wilson’s models, VS is still wrong on the evolution of altruism. There are a number of possible explanations for altruism, including inclusive fitness, but I don’t want to get into a long conversation on how altruism might have evolved because I would have research and then write an inconveniently long book to do so.
Having said that, Coyne’s use of “cheating” (even in quotations) is a little unfortunate, because it overlaps with the literature on Prisoner’s Dilemma and cooperation. Cooperation and altruism are not the same concept (again, see West et al. for a good breakdown of the different concepts, or any text on social evolution); altruism might be a subset of cooperation, depending on how you define the terms, but usually altruism comes at a cost to the altruist while cooperators do not necessarily pay a cost to cooperate. “Cheating detectors” is more appropriate for a conversation about cooperation than altruism (see also Figure 2 of this paper), which makes the rest of VS’s argument difficult to respond to. I think that Coyne is using ‘cheating’ to refer to individuals who receive the benefit of altruistic acts without paying the price of altruism, but that’s not the same as cheating in models of cooperation. (Honestly, ‘cheating’ isn’t a great word on a lot of grounds, including confusion with other areas such as signalling and an implication of conscious intent where none is necessary).
Returning to the posts that started this digression: my honest belief is that this group selection debate should have been over years ago, but I will still support Wilson’s right to continue trying to make his case. If he’s going to attack people like Dawkins and Coyne, though, he’d better learn to be prepared for them to hit back. And though it’s unlikely that either of them will ever read this post, I’d like to tell them that they’re not alone.
P.S. Can I take this opportunity to point out a further example of Wilson claiming that people agree with him when they don’t? If you read the end of Wilson’s second piece, he says:
For readers who are up for a challenge and want to learn more about the theoretical basis and empirical evidence for group selection from someone other than myself, I recommend Steven A. Frank’s “Natural Selection. III. Selection vs. Transmission and Levels of Selection (Journal of Evolutionary Biology, 2011). For Frank, it goes without saying that natural selection is a multilevel process and that the group level is often a significant evolutionary force.
I’ve actually read that paper. In it, Frank once again points out that kin selection and group selection are the same thing:
The equivalence of r and Hamilton’s formal theory of kin selection establishes the exact equivalence of multilevel group selection and kin selection.
And then, after a long analysis, he compares the usage of the two methods in a section entitled (tellingly): Reasons to favour kin selection over group selection. It contains exactly what the title says. If you can get it and you like technical discussions of evolutionary biology, I urge you to read the paper yourself. If you don’t, then just do me a favour and don’t accept Wilson’s claims about this paper at face value.