The world continues to wait for a conversation between Bret Weinstein and myself on the topic of group selection. It began with Weinstein’s conversation with Richard Dawkins, where he tried and failed to get the great master to admit that religions might be adaptations rather than mind viruses. That led me to look up what Weinstein has written on the subject. There wasn’t much , but it was enough for me to tweet that he was invoking group selection. Weinstein didn’t see it that way, so we both started to tweet about having our own conversation. It would be punchy but respectful, in keeping with the ethos of the so-called Intellectual Dark Web.
Alas, the event has yet to materialize so the world continues to wait. Fortunately, a lengthy conversation between Weinstein and Joseph Walker  provides all the information needed to justify my original hunch. Bret Weinstein invokes group selection in every way except using the words.
If this phrase sounds familiar to some readers, it is because I have written a similar critique of Richard Wrangham , based on his new book The Goodness Paradox. As an old guy (I turn 70 in July), I even critiqued Richard Alexander, Weinstein’s mentor and PhD advisor, way back in 1999 !
Lest you think that I’m trying to nuke the entire establishment of evolutionary thinkers who traffic in ideas such as selfish genes, kin selection, reciprocal altruism, indirect reciprocity, and extended phenotypes, let me be clear about the nature of my complaint. Imagine that you are fluent in two languages—say, English and Spanish–and you enter into conversation with someone who speaks English fluently and Spanish hardly at all. To your surprise, this person tells you that English is a superior language and that much of what is said in Spanish is flat out wrong. What an incredible boor!
That is my complaint against the establishment of evolutionary thinkers who reject group selection in their own minds when they can barely speak its language. Why can’t they become bilingual, for heaven’s sake! I can speak and appreciate what is said in their language–why can’t they do the same in mine?
Actually, the peer-reviewed literature has become more bilingual over the decades, I am happy to report. Here is what two highly respected researchers, Jonathan Birch and Samir Okasha, write in a 2014 article titled “Kin Selection and Its Critics” [5, p. 28].
In earlier debates, biologists tended to regard kin and multilevel selection as rival empirical hypotheses, but many contemporary biologists regard them as ultimately equivalent, on the grounds that gene frequency change can be correctly computed using either approach. Although dissenters from this equivalence claim can be found, the majority of social evolutionists appear to endorse it.
A 2014 survey of anthropologists from PhD granting departments found that the majority accepted group selection as an important force in human cultural evolution and understood the concept of equivalence described in the passage quoted above . Unfortunately, the majority of social evolutionists contributing to the peer-review literature aren’t very active on the internet. In addition, there are generational effects. The adage “science progresses funeral by funeral” doesn’t hold for everyone, but it does hold for some, who will go to their graves proclaiming that English is superior to Spanish and that things said in Spanish are just plain wrong.
Weinstein laments Dawkins’ inflexibility on the subject of religion, but he himself is a seed that has fallen close to the tree of his mentors, Robert Trivers and Richard Alexander, in his portrayal of group selection. At least he is fluent in the language that he knows, so his interview with Walker provides all of the background needed on the core concepts of selfish genes, kin selection, reciprocal altruism, indirect reciprocity, and extended phenotypes, which became established during the second half of the 20th century, along with the stock argument for why group selection doesn’t work. Let’s take these in turn.
What all the core concepts share in common is the appearance of selfishness, or self-interest if you prefer. Everything that evolves boils down to selfish genes. Inclusive fitness is about maximizing the copies of your genes in the bodies of others in addition to yourself. Reciprocity is about helping others in expectation of return benefits to yourself. Extended phenotypes are about your genes reaching beyond your body, such as a beaver dam as the phenotype of a beaver. Indirect reciprocity notes that the return benefits from helping others need not be restricted to the individual that you helped. In all cases, the explanation ends up being all about you.
The stock argument against group selection is that in any group containing both altruistic and selfish individuals, it is inevitable that the latter will replace the former. Hence group selection won’t work. This is what Dawkins says about group selection in The Selfish Gene and what Weinstein says in his interview with Walker—over four decades later—hasn’t changed a bit. There’s an enduring meme for you!
I have always marvelled at how this argument against group selection could ever have been taken seriously when it doesn’t even acknowledge the solution offered by group selection theory. The reason that altruism can evolve, despite declining in frequency in each and every group also containing selfish individuals, is because groups with a higher frequency of altruists contribute more to the total gene pool than groups with a lower frequency of altruists. What evolves in the total population reflects a balance between the opposing forces of selection among individuals within groups (favoring selfishness) and selection among groups in a multi-group population (favoring altruism). You can’t dismiss the evolution of altruism by focusing only on the negative within-group component!
Allow me to illustrate this point with one of the most influential models of group selection—the haystack model of John Maynard Smith . The advantage of a model is that it makes everything precise. Maynard Smith fancifully imagined a population of mice that lives in haystacks. At a single genetic locus, one allele codes for aggressiveness (a form of selfishness) and its alternate codes for docility (a form of altruism). Each haystack is colonized by a single female fertilized at random by a single male—therefore a sample of four alleles drawn at random from the total population. The population in each haystack grows for a number of generations without any migration between haystacks. Then all of the mice disperse, mate randomly, and colonize a new set of haystacks to repeat the cycle. During the period of time spent in isolation, the selfish gene completely replaces the altruistic gene in all haystacks that were colonized by both alleles. In haystacks that were colonized only by the altruistic allele, however, the mouse population grows larger and contributes more to the total gene pool than the selfish populations.
Unlike the stock dismissal of group selection, the haystack model includes both within-group selection (favoring selfishness) and between-group selection (favoring altruism) in accounting for what evolves in the total population. Given the assumptions of the model, within-group selection is by far the strongest force, so altruism goes extinct not only within each haystack, but in the total population. When the haystack model was published in Nature magazine in 1964, it was regarded as a drop-dead argument against group selection as a whole.
But wait! Looking back, the assumptions of the model are highly unrealistic. In particular, the assumption that altruistic genes are completely replaced by selfish genes within each haystack before dispersal occurs makes within-group selection as strong as it can possibly be. In fairness to Maynard Smith, 1964 was before the advent of desktop computers and many of his assumptions were made to simplify the math. But what if we were to revisit the model with more realistic assumptions? For example, what if we use the same benefit and cost terms that Hamilton used in his model of inclusive fitness? In other words, in every haystack containing both types, altruists deliver benefits (b) to others in their group at a cost (c) to themselves, while selfish individuals accept the benefits without paying any costs. Let’s also alter the number of generations spent within each haystack (e.g., 5,10,15) as a variable of the model. What happens then?
This is the question that I asked in a 1987 article published in the journal Evolution . What happens is that the altruistic gene declines in frequency in each group colonized by both types, but it doesn’t entirely go extinct. Take the extreme case of a single altruistic mutation in a population of selfish genes. The mutant allele finds itself in a haystack with three selfish alleles, or an initial frequency of 25%. Depending upon the values that we assign to the b’s and c’s, its frequency might decline to 19% after 5 generations, 14% after 10 generations, and 9% after 15 generations. That’s the bad news. The good news is that the single haystack containing the altruists might grow 20%, 45%, and 80% larger than the entirely selfish haystacks during the same period. Thanks to this fitness difference at the group level, the frequency of the altruistic gene can increase in the total population, despite decreasing within the haystack. A thorough exploration of the parameter space, made possible by the advent of desktop computing, showed that altruism can robustly evolve in the haystack model, given assumptions that are more reasonable than Maynard Smith’s original model.
This was only one of many computer simulation models, laboratory studies, and field studies demonstrating that between-group selection cannot be dismissed as invariably weak. Instead, the balance between levels of selection must be evaluated on a case-by-case basis. One of the most general and elegant theoretical formulations of multilevel selection is the Price equation, which convinced Hamilton that his theory of inclusive fitness had been invoking group selection all along . None of this is reflected in Weinstein’s perpetuation of the stock argument against group selection, which only takes note of within-group selection.
Something else that Weinstein gets wrong in his interview with Walker is that group selection models are confined to interactions among non-relatives. I wish I could say otherwise, but this simply reflects ignorance of the literature, including everything that W.D. Hamilton wrote from 1975 onward. If Weinstein was bilingual, he would realize that the coefficient of relationship (r) in Hamilton’s rule translates into an index of genetic variation among groups in a group selection model. When r=0, individuals are randomly distributed into the groups. When r=1, group members are genetically identical and all of the variation is between groups. Throughout the entire range of r values, what evolves in the total population reflects a balance between levels of selection. In all of the classic group selection models, groups are colonized by small numbers of individuals, which are therefore related to each other compared to the total population. In the case of the haystack model, the mice within each haystack start out as full siblings, but this did not prevent Maynard Smith from calling it a group selection model and contrasting it with Hamilton’s model of kin selection. Years were required to reveal that not only is between-group selection a potent evolutionary force, but that it is implicitly invoked by all of the theories that were developed as alternatives.
Perhaps the best way to make this point is for me to present my fantasy version of a conversation with Weinstein.
David Sloan Wilson (DSW): Bret, do you agree with me that evolutionary models are based on relative fitness? It doesn’t matter how well an individual survives and reproduces in absolute terms, only that it does so better than others in its vicinity—right?
Bret Weinstein (BW): Of course!
DSW: Then doesn’t it strike you as curious that all of the models framed in terms of self-interest don’t calculate relative fitness?
BW: What do you mean?
DSW: They all assume that individuals or genes evolve to maximize their absolute fitness. For example, an individual that follows Hamilton’s rule will make more copies of its genes than if it doesn’t follow Hamilton’s rule. The reciprocal altruist gets a larger net benefit than if she doesn’t reciprocate. They all take the form of “what’s my payoff if I behave this way, compared to my payoff if I behave that way”. That’s the maximization of an individual’s (or gene’s) absolute fitness, not its relative fitness compared to others in its vicinity.
BW: But the models show that these behaviors evolve in the total population, compared to those who don’t maximize their absolute fitness. Isn’t that relative fitness?
DSW: Yes, but it’s relative fitness all things considered. By “in the vicinity,” I mean within the groups where the social interactions are taking place. Let’s take the concept of extended phenotypes, for example, which notes that a beaver dam can be regarded as part of the phenotype of a beaver.
BW: Yes—that’s one of Dawkins’ concepts that I like, even though I disagree with him about religion.
DSW: Ok, but now let’s look more closely at a beaver pond. Does it contain only one beaver?
BW: No, typically it contains more than one.
DSW: What if the beavers in a single pond vary in the amount of work they put into building a dam? Which have the highest relative fitness?
DSW: Spit it out, Bret! The free-riding beavers have the highest relative fitness! Calling the dam an extended phenotype of beavers doesn’t alter this fact!
BW: But what if the beavers are related?
DSW: That doesn’t matter either! In every pond containing both types, the free-riders will have the relative fitness advantage. Genetic relatedness is important because it clusters the dam-builders and free-riders into different ponds more than if they were distributed at random. So-called kin selection increases variation among groups, therefore the importance of between-group selection compared to within-group selection. Partner choice among genetically unrelated individuals would do the same thing. It’s variation among groups that’s important, not genealogical relatedness per se. Hamilton got that in 1975. Where have you been all these years?
BW: Hey! We of the Intellectual Dark Web like to keep the conversation respectful!
DSW: So do I, but I am cordially telling you that there are academic standards to maintain. Our reading public deserves to know that we are both experts in the topic that we are discussing. But, to lighten up on you a bit, I know that there is a weird dynamic in which some experts aren’t bilingual when it comes to theories of social evolution. They only think in terms of the individualistic perspective, only read that literature, and therefore dismiss group selection and can’t see it in their own models when it is front of their faces. Here’s an example from Richard Wrangham’s new book The Goodness Paradox, which is just like the beaver example. Instead of beavers in their dams, we have chimps in their communities. The counterpart of dam-building beavers is male chimps that kill or harass members of neighboring territories. The cost of doing this might not be large, but it is still an individual cost, while the benefit of expanding the territory over a period of years goes to the whole community. This is by Wrangham’s own account, but he can’t see the role of between-group selection any more than you can. The same goes for your mentor, Richard Alexander, but you should have the flexibility to learn what has become the consensus in the peer-review literature.
BW: Well, gosh, David! Thanks for enlightening me! I’m really beginning to see it your way now!
Ok, that last line was total fantasy, but this is the conversation that I will attempt to have with Weinstein, if it ever materializes, and I don’t see how he can evade the conclusion. His stock dismissal of group selection—selection against altruists within groups—takes place in every model of social evolution framed in terms of self-interest and can be seen merely by comparing the relative fitness of individuals in the groups where the social interactions are taking place.
Much of the conversation between Walker and Weinstein is framed in terms of the study of religion. For Weinstein, it is a no-brainer that religions are adaptations, not mind viruses. They are also “memeplexes”—whole complexes of cultural traits that evolve as a package—not atomistic memes. Weinstein wishes that Dawkins could be flexible enough to get beyond mind viruses. I wish that Weinstein could be flexible enough to realize that possibly—just possibly—the evolution of religions as adaptive memeplexes might require a process of selection among alternative memeplexes and cannot be explained entirely in terms of selection among alternative memes within memeplexes. That should also be a no-brainer.
I also hope, respectfully, that in future conversations with myself or anyone else on the subject, Weinstein displays a grasp of the peer-reviewed literature on religion from an evolutionary perspective. As far as I can tell, he remains mostly within the bubble of the New Atheist community, which for the most part does not conduct serious research on the subject. When was the last time that Richard Dawkins, Daniel Dennett, or Sam Harris contributed to the peer-review literature? How familiar is Weinstein with the likes of Candice Alcorta, Scott Atran, Robert Bellah, Pascal Boyer, Joseph Bulbulia, Michele Geland, Joseph Henrich, Dominic Johnson, Ara Norenzayan, Richard Sosis, Peter Turchin, and Harvey Whitehouse in addition to my own scholarly work? Go here [10,11] for recent overviews.
To pick an example central to New Atheist concerns, consider the concept of the resurrection associated with Christianity [what follows is distilled from a chapter of my book The Neighborhood Project titled “The Natural History of the Afterlife”]. When did it arise in human history and what caused its spread, compared to alternative beliefs? Historical scholarship is sufficiently detailed to answer this question . In the Hebrew Bible, death is mentioned about a thousand times. In most cases, people just die and nothing is said about an afterlife of any kind. Judaism is centered around establishing the nation of Israel on earth, not in an afterlife.
In about seventy cases, an afterlife is mentioned in the Hebrew Bible, but it is not the heaven associated with Christianity. Instead it is Sheol, a gloomy place similar to Hades in Greek mythology, where everyone goes regardless of whether they have been good or bad.
Sheol is mentioned in a specific context—when people face the prospect of dying without having achieved much of anything during their time on earth. Their gloomy thoughts about the afterlife mirror their gloomy thoughts about their lives. So much for the concept of heaven as an individual-level adaptation for allaying anxiety about death!
The concept of the resurrection associated with Christianity doesn’t unambiguously appear until the Book of Daniel. Scholarship of the period is so detailed that the Book of Daniel can be dated to the time of the Seleucid persecution of 167-164 B.C.E., which makes it one of the latest texts of the Hebrew Bible. The challenge to Judaism at that time was assimilation: many Jews were attracted to Hellenic culture. According to the Book of Daniel, eternal life would be granted to Jews who maintained their traditional ways and all other Jews would suffer everlasting abhorrence. Belief in the resurrection was arguably a key factor that enabled the Maccabees, the traditionalist faction, to resist the Hellenist monarch Antiochus in a victory that is still celebrated in the festival of Hanukkah. Here is how two eminent scholars of the period put it .
The Jewish expectation of a resurrection of the dead is always and inextricably associated with the restoration of the people of Israel; it is not, in the first instance, focused on individual destiny. The question it answer is not the familiar, self-interested on “Will I have life after death?” but rather a more profound and encompassing one, ‘Will God honor his promises to his people?’
In other words, the concept of the resurrection originated as a cultural mutation that increased solidarity in the context of between-group competition within Judaism. The belief did not increase the fitness of individuals, compared to other individuals in the same group who did not believe. It increased the fitness of the groups that believed, compared to the groups that didn’t believe.
The religion that formed around Jesus was initially a twig on a branch of this cultural tree that sprouted several centuries previously. The idea of a Messiah who would die and return to signal the end of days was by then thoroughly familiar. The followers of Jesus thought that he was the Messiah, that he had indeed risen three days after his crucifixion, and that the end time had begun. The Jesus sect would probably have remained a footnote in religious history were it not for a key difference that had nothing to do with its otherwise standard (for its time) afterlife beliefs: Gentiles could now become the chosen people. A belief system that previously had been restricted to one ethnic group could now spread throughout the world. The same dynamic would result in the origin and spread of Islam centuries later.
But wait! There’s more! Have you ever wondered why the four gospels of the New Testament are so different from each other, not to speak of the gospels that weren’t included in the New Testament? Was it simply because of faulty memory and the passage of time, which in evolutionary terms would be a form of cultural drift? Not according to the distinguished scholar of religion Elaine Pagels. In The Origin of Satan and other books, she interprets each gospel as a sacred story that became locally adapted to a particular Christian community in the context of its specific challenges. The gospels that made it into the New Testament were the ones that did the best job of creating strong communities.
There is much, much more, but perhaps I have said enough to make my general point: Human history provides a fossil record of cultural evolution so detailed that it puts the biological fossil record to shame. It isn’t necessary to speculate about whether religions are adaptations or whether particular beliefs and practices evolved by within-group vs. between-group selection. There is ample evidence to decide the facts of the matter on a case-by-case basis. The situation is similar to the study of natural history during Darwin’s day—mountains of reliable information about plants and animals, which were waiting to be organized by his theory of natural selection. When we examine religious beliefs and practices on a case-by-case basis, we can find group-level adaptations, individual-level adaptations, mind viruses that spread at the expense of both individuals and groups, byproducts, mismatches, and cases of cultural drift. If you could say only one thing about religious beliefs and practices, however, it would be that they are impressively adapted to create strong communities of believers. Emile Durkheim—the son of a rabbi—got it largely correct when he defined religion as “a unified system of beliefs and practices relative to sacred things…which unite into one single moral community called a Church, all those who adhere to them.”
The term “memeplex” is newspeak for what Durkheim meant. That makes Weinstein right compared to Dawkins, but if Weinstein thinks that he can explain religious memeplexes entirely on the basis of competition among memes within single communities, as opposed to competition among communities in a multi-group population, he’s on a fool’s errand.
The original cover of Richard Dawkins’ 1982 book The Extended Phenotype features an optical illusion called the Necker Cube, which causes the mind to toggle back and forth between two configurations. That’s how Dawkins sees the gene’s eye view of evolution compared to the more conventional view of individual organisms as adaptive units. Both are correct, but they yield different insights.
I like that metaphor and I’m happy to credit the gene’s eye view with insights, as long as it’s not used as an argument against group selection. One way to restate my point about different languages is to say that there is a third configuration—the group’s eye view—that needs to be added. In this configuration, behaviors that are “for the good of the group” tend to be selectively disadvantageous within groups and therefore require a process of between-group selection to evolve. If we define selfishness in “increasing relative fitness within groups” and altruism as “increasing the fitness of one’s group, relative to other groups”, then the evolution of altruism can be explained at face value without describing it as a form of selfishness.
Of course, if altruism evolves by group selection, then altruistic individuals are more fit than selfish individuals in the total population, all things considered. The same will be true for the genes that code for the individual behaviors. This is how altruism can be made to appear as a form of selfishness, whenever it evolves by group selection. To use that as an argument against group selection is a fallacy, pure and simple. It is easy enough to see this in retrospect, but consider these two passages from Weinstein’s mentor, Richard Alexander:
In 1966 [George C.] Williams published a book criticizing what he called “some current evolutionary thought,” in which he set out the general argument against “group selection” and chastised biologists for invoking selection uncritically at whatever level seemed convenient. William’s book was the first truly general argument that selection is hardly ever effective on anything but the heritable genetic units of “genetic replicators” (Dawkins 1978) contained in the genotypes of individuals. (Alexander 1978: 36)
I suspect that nearly all humans believe it is a normal part of the functioning of every human individual now and then to assist someone else in the realization of that person’s own interests to the actual net expense of those of the altruist. What this “greatest intellectual revolution of the century” tells us is that, despite our intuitions, there is not a shred of evidence to support this view of beneficence, and a great deal of convincing theory suggests that any such view will eventually be judged false. (Alexander 1987: 3).
In the first passage, Alexander erroneously treats the gene’s eye view of evolution as an argument against group selection. He can’t see the third configuration. In the second passage, he denies the possibility that the evolution of beneficence can be explained at face value, without needing to re-describe it as a form of selfishness.
This is the legacy that Weinstein must find the flexibility to go beyond on the topic of group selection. That said, I admire him and what he stands for in most other respects.