Thanks to Peter Turchin and Michael Hochberg for creating and managing the Social Evolution Forum, which has become an excellent arena for high-level discussion. Thanks also to my colleagues who took the time to write commentaries and to readers who responded with their comments. In addition to this general reply, I have also provided comments in the ‘response’ section of each commentary.
Equivalence check: Superficially, it appears that two commentators (Pagel and Sanderson) agree with my assessment about cooperation but not about group selection, while the others (Hewson, Zimmerman, Hochberg and Whitehouse, Bulbulia, Greenhill, and Gray) are more accepting of group selection. In fact, Pagel and Sanderson are just as accepting of group selection when terminological issues are resolved. In a previous essay published in the Social Evolution Forum, (Wilson 2012), I focused on the concept of equivalence, whereby scientific frameworks are different but worthy of coexistence. Equivalence requires an ability to understand and translate between frameworks, similar to translating between languages. Scientists who insist on employing only one framework run the risk of committing errors comparable to “I don’t speak Russian; therefore everything stated in Russian is wrong.”
This kind of error is on display in the comments by Pagel and Sanderson. The ‘language’ of multilevel selection is easy to speak. It involves identifying where fitness differences exist in a multi-tier hierarchy of units. I clearly indicate my use of the multilevel framework in my target essay, as in this passage concerning the evolution of any given genetic trait: “…did it evolve by virtue of increasing the fitness of genes relative to other genes within the same organism, individuals relative to other individuals within groups, or groups relative to other groups in the total population?”
Pagel and Sanderson are either unable or unwilling to speak this language. Consider this passage by Pagel: “In the simplest case, imagine that you inhabit a group of two and that by helping each other you can achieve more than twice as much as the two of you working alone. Now ramp this scenario up to a larger group. …The returns from cooperation means that what looks like altruism, is really a form of enlightened self-interest.”
Pagel is not speaking the language of multilevel selection theory. If he did, he would see that there are no fitness differences among cooperators within single groups in his example. The fitness differences exist at the group level—groups of cooperators contribute more to the total gene pool than groups of non-cooperators or solitary individuals. If cooperation is cost-free, then the trait is neutral with respect to within-group selection and evolves by between-group selection given any variation among groups. If cooperation involves any private cost, then cooperators are less fit than non-cooperators within groups, which between-group selection must overcome for cooperation to evolve in the total population.
Or consider this statement by Sanderson: “Of course there are cultural traits that may benefit the group as a whole. But all this can mean is that these traits benefit all of the individual members of a group”. Sanderson is not speaking the language of multilevel selection. If he did, he would see that there are no fitness differences among individuals within groups. Natural selection requires fitness differences, and fitness differences at the group level are required for the cooperative trait to evolve in the total population in his own example.
I do not insist that my colleagues speak the language of multilevel selection in their own work. I appreciate that terms such as “self-interest” and “altruism” can be defined in numerous ways; e.g., in terms of absolute fitness rather than relative fitness within groups. But before I brand their results as wrong, I do an equivalence check by translating their examples into my preferred framework. This is what Pagel and Sanderson fail to do in their commentaries.
I invite readers to do an equivalence check for all of the examples in all of the comments on my target essay. By my reckoning, not only do Pagel and Sanderson agree with me on the importance of cooperation, but they also agree with me and the other authors on the importance of between-group selection, as defined within the framework of multilevel selection theory. As Peter Turchin comments on Pagel’s use of the phrase “groups as collective survival vehicles” in his book Wired for Culture, “That’s group selection!”
On groupishness: Group selection does not favor altruism or cooperation per se. It favors any proximate mechanism that causes groups to differentially contribute to the gene (or meme) pool of the total population. Some of the mechanisms appear altruistic in the conventional sense of the word and require substantial individual cost to benefit the group. Other mechanisms might appear selfish in the conventional sense of the word, and need not require much or even any self-sacrifice. Some forms of competition among individuals within groups also benefit the group and will be favored, rather than suppressed by group-level selection. Complex social interactions often result in multiple stable equilibria, which are internally stable by definition. Selection among equilibria results in well-adapted groups that are also internally stable, unlike the internal instability characteristic of altruism. I have emphasized these points in other publications but not sufficiently in my target essay, especially when I stated “it all revolves around cooperation.” I therefore agree with Peter Turchin in his response to Pagel that ‘groupishness,’ not ‘cooperation,’ is the best general term for describing the products of group selection.
These points are part of the consensus exhibited at the Ernst Strungmann forum and are appreciated by some of the commentators but not others. The only evidence that Pagel and Sanderson will accept for group selection is the kind of self-sacrificial altruism found in social insect colonies. All other mechanisms that benefit groups are interpreted as “enlightened self-interest” and are assumed to be explicable without invoking group selection. The only way to evaluate this claim is to perform an equivalence check on a trait-by-trait basis. Please see my response to their individual comments for more.
On the rewards and punishments that support groupish behavior: Human social groups bristle with mechanisms that punish bad behavior and reward good behavior. When these mechanisms are in place, behaving prosocially becomes individually advantageous and behaving antisocially becomes just plain stupid. Nevertheless, we still need to explain how the rewarding and punishing mechanisms evolve. Causing others to promote the common good is usually itself a common good that requires time, energy, and risk on the part of the rewarding and/or punishing individual, which economists term a second-order public good.
These points are part of the consensus exhibited at the Ernst Strungmann forum and are appreciated by some of the commentators but not others. Consider the example of reputation discussed by Pagel. Good deeds increase one’s reputation and bad deeds decrease one’s reputation, so individuals who care only about their reputation will perform good deeds. But what are the traits required for others to bestow a high or low reputation upon a given person? Do these traits evolve based on relative fitness advantages within groups, or do they require between-group selection?
On distinguishing between psychological and behavioral definitions of altruism: The distinction between proximate and ultimate causation in evolutionary theory requires separate definitions of altruism based on psychological motives (proximate) and phenotypic consequences (ultimate). If we imagine a 2×2 table with psychological altruism and selfishness as rows and behavioral altruism and selfishness as columns, all four combinations are possible. This should be common knowledge among all evolutionists, so it is discouraging that some of the commentators appear to conflate psychological and behavioral definitions. Consider the following passage by Sanderson: “The point about conquest is especially important because it is about one group defeating another. On the surface that looks like group selection. But is it? I would say no. It isn’t a matter of a whole society against another, but of the most powerful members of a society conquering others to gain the spoils. Do they want to share the spoils with the rest of society? Not really.” Sanderson is comparing apples and oranges when he uses psychological motives to argue against the fitness differences of traits within and among groups. This example is problematic in other respects that I address in my response to Sanderson’s commentary.
On consensus: Sanderson questions the significance of the consensus exhibited at the Ernst Strungmann forum, even given that it took place. I agree with Sanderson that it is silly for anyone to claim that a given position is true because X people endorse it, no matter what the value of X. Nevertheless, science is not a frictionless pendulum of ideas. Results are established that are durable enough to be called facts, and these tend to become widely accepted. The earth is round and very old. Continents drift. To these we can add the fact that many traits evolve on the basis of the differential fitness of groups, despite being selectively disadvantageous within groups, as these terms are defined in multilevel selection theory. What’s new is that large numbers of scientists are accepting this fact at face value and do not regard equivalent descriptions as a denial of the fact.
On using consensus and equivalence to guide empirical research: The commentaries by Hochberg and Whitehouse and Bulbulia, Greenhill, and Gray are especially welcome because they show how the consensus exhibited at the Ernst Strungmann forum can be used to guide empirical research. I wholeheartedly agree about the importance of history as a fossil record of cultural evolution, which is often so rich that it puts the biological fossil record to shame. Narrative histories often provide sufficient detail to evaluate how new cultural forms originate and spread in competition with alternative forms. An important book in this regard is Robert Bellah’s (2011) Religion in Human Evolution: From the Paleolithic to the Axial Age (an audio interview that I recently conducted with Bellah is available here) Quantification, including the phylogenetic methods described by Bulbulia, Greenhill, and Gray, adds additional power. Please see my response to their individual commentaries for more. The most important general point is that the entire theoretical and empirical toolkit that is used to study biological diversity can be used to study cultural diversity. We can look forward to the same kind of integration for the study of culture during the 21st century that took place for the biological sciences during the 20th century (and continuing).
Bella, R. (2011). Religion in Human Evolution: From the Paleolithic to the Axial Age. Cambridge, MA. Harvard University Press.
Wilson, D. S. (2012). Clash of Paradigms: Why proponents of multilevel selection theory and inclusive fitness theory sometimes (but not always) misunderstand each other. Social Evolution Forum, July 13.
Pagel was not speaking the language of multilevel selection theory, because he was talking about the Allee effect – a very different kettle of fish.
The Allee effect is also responsible for ‘groupishness’ and cooperative social behaviour.
To properly understand human social behaviour you have to be aware of kin/group selection, reciprocity, the factors underlying the Allee effect, the importance of cultural inheritance, and several other factors. This isn’t a “one-theory-explains-everything” problem.
I would generally agree that neither Pagel nor Sanderson showed much sign of understanding group selection. Because modern forms of group selection turned out not to explain anything that wasn’t already modeled by kin selection, there’s not very much incentive for people to learn about it. It seems clear that much of the modern enthusiasm for group selection was in the hope of it proving to have some explanatory power independent of kin selection. Those hopes didn’t pan out – and those who said “that’s just kin selection” were right all along.
Mainstream biology already did the kin selection vs group selection thing a few decades back. The disadvantages of the “group selection” perspective are fairly well known. The human sciences involving culture were slow to accept evolutionary theory, slow to accept symbiology, and slow to accept kin selection. For some reason, it seems hardest for humans to apply basic biological concepts to themselves. Looking at the current degree to acceptance of cultural symbiology and cultural kin selection, the human sciences are around where organic biology was in the 1960s. That suggests that some changes are coming.
I want to comment first on the commentary by Bulbulia et al. on my religion work. Wilson is right that empiricism will resolve this, and I think the history of science shows the more we can rigorously quantify predictions and results the more rapid progress we can make. It’s important to note that my article on religious violence is as much a social network analysis as it is a phylogenetic one. An appropriate criticism of applying phylogenetics to culture was that it was unable to deal with non-branching patterns. We are now probably about 70% done with a solution to this problem, thanks to excellent work by many folks including Russel Gray’s and Charlie Nunn’s groups. Infusing social network analysis in to this space is the next logical step, especially as network autoregression modelling is essentially the same as phylogenetic autoregression (PGLS). The first field to apply those types of statistics was actually spatial autocorrelation in the 1980’s. This is happenning in biology too. Reticulate models from population and quantitative genetics are coming pouring back into phylogenetics. Top people in phylogenetics like Joe Felsenstein have commented on this.
On Wilson’s excellent post above and on group selection more generally. I think the comments by Pagel and Sanderson and Wilson’s rebuttal of them show how inclusive fitness and selfish gene talk are fundamentally counterintuitive and incapable of expressing in a reductionist way the mechanism of evolutionary altruism. I greatly respect both Pagel and Sanderson as scholars and have learned tremendously from their work, but in both their comments they confuse the basic ecological payoff structures of their own examples. In Pagel’s example, the behavior he talks about has no individual fitness cost to performing the behavior and is thus classed as a mutualistic interaction. Sanderson’s claim that group selection relies on all the altruistic individuals having higher fitness also ignores the ecological dynamic that Hamilton’s work sought to solve. Of course all the individuals have higher fitness – but that’s because they receive the benefits of other individuals’ behaviors not because they perform the behavior. The performance itself of an evolutionary altruistic behavior by definition always reduces fitness. This makes it fundamentally different from most behaviors (like eating) that the performance of increases one’s fitness directly. That is why we need a different mechanism to explain evolutionary altruism. It’s not Pagel and Sandersons’ faults that they get all this mixed up, it’s because inclusive fitness and selfish gene talk just can’t express the reductionist mechanism of evolutionary altruism. The verbal components of these theories are essentially teleological and that isn’t how the world is from the viewpoint of science.
With all this talk of ‘equivalence’ I think there has been a bit too much giving away the store or abandoning the fort since Pinker’s absurd essay. Agreed, inclusive fitness and selfish gene mathematics can be made equivalent to group or multilevel selection theory, but only with ‘Jesuitically sophisticated’ contortions. Important and empirically true predictions about the evolution of altruism do not flow naturally from the verbal articulation of inclusive fitness and selfish gene theory, and I would argue there is in fact no non-metaphorical way to express evolutionary mechanisms with the verbal components of these theories. It’s astounding they continue to receive such support by eminent scientists like Pagel and Sanderson when we have an easy to articulate, reductionist, and non-metaphorical alternative available (group/multilevel selection). One concrete example – when I wrote the religion paper in Human Nature, I had to push back on reviewers about my work being a “Greenbeard fallacy” and I had to include an entire section about how Greenbeard effects in genetic systems are empirically verified to happen! Most of my colleagues still think greenbeardism is a fallacy – but Andy Gardner (who collaborates with Alan Grafan) – has shown it happens in nature. Why is this the case that most researchers still think greenbeardism is a fallacy when this belief is empirically false? Because any verbal articulation of inclusive fitness theory would intuitively lead one to think greenbeard evolution doesn’t work – but it does. Sounds like a refuted hypothesis to me. Gardner himself is an inclusive fitness proponent who goes though verbal contortions in his work to avoid talking about group effects, and often ends up doing so anyway. Don’t get me wrong, I love Gardner’s science, but the verbal theory he is using is really hampering the advancement of the field by promoting metaphor and misunderstanding even among eminent scholars.
My more complete opinion on why inclusive fitness and selfish gene verbal theory are incoherent can be found on my personal blog.
The basic ideas seem free from contortions to me:
Kin selection: Heritable elements increase in frequency by helping copies of themselves in relatives – as well as by helping copies of themselves in offspring.
Group selection: Heritable elements increase in frequency via differential reproductive success of groups of organisms – as well as by differential reproductive success of individual organisms.
In the footnotes kin selection needs to say what “relatives” are, and group selection needs to say what counts as a “group”.
It isn’t obvious that these ideas are equivalent – but that’s the way it has turned out. I don’t think much of a case can be made that either idea is “contorted”. The main problem I see is that one of these ideas has inspired a considerable mountain of junk science.
Thanks for the response, but it only further supports my argument. Note how you resort to only further metaphor in explaining kin selection but not in explaining group selection.
“Green beard” genes were invented by Hamilton and christened by Dawkins. If we are keeping score, surely the concept would count as a win for kin selection theorists.
Yes, I believe the term was coined by Dawkins in the Extended Phenotype. In it he presents an argument from verbal gene selection theory that greenbeards would never or almost never occur in nature. Now we have multiple examples from invertebrates and vertebrates in which greenbeards do happen. Saying that’s a win for Dawkins and co. is like saying the heliocentric solar system was a win for the Catholic Church because it rejected the idea even before Galileo supported it.
In “The Selfish Gene”, Dawkins discussed a “green beard” gene. He said it was “not very probable” – but that it remained “a theoretical possibility”. No green beard genes were discovered for decades. “Green beard” genes remain rare for the reason that Dawkins gave.
I suppose it is rather unfair to give kin selection theorists credit for most of the major discoveries in the field – when they so massively outnumbered the group selection theorists – who additionally were busy fighting for their preferred theory. The main issue is not the past, but the future.
As I see them, the biggest problems for the future of group selection are threefold: A) That groups of close relatives are clearly the most important ones in terms of adaptations – yet group selection theorists have mostly ignored those and concentrated on much larger groups where the theory doesn’t work, and produces bogus science. B) That theorists rarely use their group “weights” – and so it hasn’t been much of a quantitative predictive theory so far. C) That group selection adds very little to the long-established mainstream biology of kin selection.
At the moment, group selection seems to me to be acting as a substantial negative force, mainly because of point A. If all the group selection enthusiasts abandoned it and used kin selection instead, the quality of the resulting science would probably shoot up. I agree that group selection does have some merits, but also think it needs a kind of health warning – along the lines of: “WARNING: this theory has led to a lot of junk science”. Group selection does currently have dodgy reputation that plays a similar role – and that is probably a good thing. As a result, the group selection “problem” is mostly confined to the social sciences – which has a long history of fumbling biological concepts.
It’s amazing to pay a quick visit this web page and reading the views of all colleagues on the topic of this article, while I am also eager of getting experience.