A question about group selection

by John Quiggin on January 29, 2016

I’m doing some work on evolutionary models of game theory and need to understand the debate about group selection. It seems pretty clear that the great majority of evolutionary biologists reject the idea of group selection, but I haven’t found an adequate (to me) explanation of why they do so. A crucial problem for me is that the literature seems, without exception as far as I can see, to conflate group selection with co-operation and altruism. But the problem of group selection arises in non-cooperative settings, provided they are not zero-sum.

To illustrate the problem I’m struggling with, suppose that two previously isolated species meet as a result of some change. In one species (peacocks), competition between males for mates takes the form of elaborate, and energetically costly, displays. In the other species (penguins) males compete by providing food to their mates. In all other respects (diet, predators and so on) the two are similar. In particular, they are competing for the same food resources. It seems obvious to me that the penguins, with their more efficient social arrangements, are going to outbreed the peacocks and eventually drive them to extinction.

It seems to me there are only two possibilities here
(a) My reasoning is wrong, and we can’t judge which species, if either, will dominate; or
(b) Even though it involves one group being selected over another, this isn’t what is meant by group selection

I’d really appreciate some help on this. I’m happy to have thoughts from anyone, but I’d most like to hear from actual experts with contact details (mine are on the sidebar, or obtainable through Google).



Farah Mendlesohn 01.29.16 at 6:24 am

You need to add in to your equation:
–what predators are there? (Peacocks also use their tails to make themselves look big and scary)
–how hard food is to find, in an environment where food is abundant females don’t need hunters (see rising divorce rates in prosperous societies)
–it doesn’t actually matter who outbreeds whom if they aren’t competing for the same resources.


Hugo Mercier 01.29.16 at 6:36 am

Hi John,

The issue with your reasoning, I believe, is that the mechanism you describe cannot explain adaptations, which is what we want natural selection to do (at whatever level it operates). Even if penguins were more likely to outlive, as a species, the peacocks, that wouldn’t explain why it is adaptive for them to give food to their mates.
At best, your mechanism could explain part of the distributions of adaptations, but not why they are there in the first place. (Or, in the especially interesting case of cooperation, how they evolved in the first place in any given species).

These papers, by one of the foremost expert in the field, might be useful:

West et al 2007 Evolutionary Explanations for Cooperation
West et al 2011 Sixteen common misconceptions about the evolution of cooperation in humans

I hope that helps.


John Quiggin 01.29.16 at 6:48 am

Farah, I’ve clarified the post to make it clear that the two species are competing for the same food resources. On the idea of tails as defensive against predators, I’m assuming that the display has costs that exceed its benefits in terms of individual survival but are offset by a greater chance to reproduce.

Hugo, my dissatisfaction with West’s papers was the motivation for posting here. In particular, the titles indicate the conflation of group selection with co-operation, which I find problematic.

To respond directly to your point, if the penguins survive and the peacocks go extinct, then we will observe mate-feeding and not display. Isn’t that an explanation of the adaptation in question, just as it would be if selection operated at the level of the gene, individual or kin-group?


Hugo Mercier 01.29.16 at 6:56 am

No it’s not, because it doesn’t explain how the trait appeared in the first place, which is what NS at the gene level does.

And I thought West’s papers are rather clear conceptually — I’m not sure you’ll find anything better out there anyway I’m afraid.


Rakesh Bhandari 01.29.16 at 6:57 am

I am guessing…By making the two groups different species in your example, you eliminate the possibility of gene exchange between the groups, which would greatly reduce the effectiveness group selection.


Rakesh Bhandari 01.29.16 at 6:58 am

small typo
I am guessing…By making the two groups different species in your example, you eliminate the possibility of gene exchange between the groups, which would greatly reduce the effectiveness of group selection.


John Quiggin 01.29.16 at 7:10 am

@4 it doesn’t explain how the trait appeared in the first place, which is what NS at the gene level does.”

The theory of natural selection was well developed before anyone had heard of genes.

@5/@6 Correct


Heather 01.29.16 at 7:15 am

Have you tried David Sloan Wilson? I loved his book and summarised a lot of the points here: https://changingweatherblog.wordpress.com/2015/06/19/can-evolution-explain-human-nature/ Groups adapt to their environments which makes your example difficult – a sudden change would favour the group that adapted fastest to the environment. They also succeed in the niches they are adapted to, Sloan-Wilson tells a great story about a particular fundamentalist religion being on the rise in some developing countries and adapting as fast as possible to beat its own demise in the US. Plus he is trying to tackle the world of economics with his evonomics.com so you guys should talk :-)


Tom Davies 01.29.16 at 7:16 am

I am not a biologist, but I would say that the selection you are describing is not happening at the ‘group’ level: each individual peacock has lower inclusive fitness than any individual penguin — no group level phenotypic features (eg cooperation) come into it. The fact that there are two distinct groups involved, rather than a frequency of an allele across a single population is a red herring.


Janet 01.29.16 at 7:58 am

The adaptations you’re talking about occurred before the peacocks and the penguins met. The traits of each species developed in response to specific selective pressures that have nothing to do with the current situation as you frame it. Competition between penguins and peacocks isn’t driving natural selection in your example; neither species is changing. If there was variation in either population that allowed it to exploit another food resource, then you might see natural selection, but that would be selection within the species.


Hugo Mercier 01.29.16 at 8:01 am

John: Yes, but nowadays that’s how it’s understood. I’m not making a historical point, just stating how things are done in evolutionary biology now, as far as I can tell. In any case, the point remains: NS (at the gene, or, as a good approximation, the individual level) can explain adaptations, the mechanism you describe can’t. (Also, to clarify things, the mechanism you refer to isn’t group selection, it’s species selection, or clade selection. Group selection, in theory at least, could explain adaptations.)

Heather: DS Wilson has an undoubtedly interesting take on things, but it’s good to keep in mind that his views are still very much in a minority in biology.


Cyril Hédoin 01.29.16 at 8:13 am

Consider this basic model: 2 strategies “peacock” (K) and “penguin” (N) and a utility function such that u(K,K)=a and u(N,N)=b with b>a. Assume a standard replicator dynamics. If the population is perfectly segmented (like interacts only with like) penguins will prevail. Is this group selection? Yes and no. Actually selection operates at the individual level but the structure of the population is decisive. This is not group selection in the sense of Wynne Edwards though. Samir Okasha has a book and several papers on this. See also Sober and Wilson’s 1998 book.


Dr. Hilarius 01.29.16 at 8:22 am

I’m many years away from this topic but think you are confused by terminology. Group selection refers to intra-specific selection not competition between different species. The question is how altruistic behaviors can persist in populations and not be extinguished by selfish actors. That is, how can individuals act to lower their own reproductive fitness but still pass along sufficient genes for altruism to persist.

Why do you find the conflation of group selection with altruism a problem? That is how biologists view the issue. That is pretty much the entire issue for group selection.

Your example doesn’t involve group selection. It’s a question of inter-specific competition for resources. There is a vast literature on that topic which is totally distinct from group selection. When I left academic biology in the mid 80s, the inadequacy of existing models of competition was just beginning to be understood. One huge problem is that the environment and resources are not constants in any dimension.

If you re-stated your problem to involve individual differences within a single species, type one forages this way while type two forages some other way, it still would not involve group selection. It would then be a straightforward issue of natural selection for optimal foraging (optimal foraging theory is another vast area of research).


PlutoniumKun 01.29.16 at 8:29 am

I’m not a specialist (although I do have a background in ecology), but years back I was fascinated by the depth of dislike between what might loosely be described as the Dawkinites and the Gouldites when for most readers they were 99% in agreement. The core seemed to be the fury the Dawkinites felt at what they seemed to believe was sloppy and loose thinking by the Gouldites when they talked about things like ‘spandrels’ and group selection.

Andrew Brown wrote a book on the topic and made no bones about the fact that he saw it as an ideological battle, not a scientific one. Dawkinites (who, whether by coincidence or not, tend to be free market enthusiasts as well, see for example Matt Ridley), argued that the Gould view was quasi religious in its view – Brown took the opposite view, seeing the Dawkin side with their selfish gene as expressing an almost theological belief in a pure, simple explanation. A bit like Austrian economists now that I think of it.

I don’t follow the arguments now, but it does seem that most of the protagonists simply got tired of it. As I’m not a specialist I can’t say for certain, but I get the impression that most specialists now don’t accept the simplicity of Dawkins views, while at the same time many of Goulds theories (especially about punctuated equilibrium) are not widely accepted. In short, they’ve moved on.

As for the penguin/peacock scenario, one argument as to why the peacocks may succeed better would be that the penguin strategy is too easy to game – there is quite a lot of evidence from this from bird studies. Put simply, female penguins will learn that the best individual strategy is to recruit a willing male to feed her, while sneakily getting jiggy with the sexy alpha male who hangs around while the keen and willing nerdy penguin is off finding more food. The female gets the best of all worlds – nice guy male looking after her, good genes from the smart sexy one. But in the longer term, this encourages the genes of the lazy sneaky males to dominate, so in the end there are too many lazy and sneaky (if rather handsome) males bred. The peacock strategy, while superficially wasteful, may work because its harder for females to game the system.


bad Jim 01.29.16 at 8:33 am

Each individual is a member of a group. If the members of one group individually survive and reproduce at a higher rate than members of another group competing for the same resources, that group will grow and the other diminish. This is plain old natural selection, operating at the individual level and manifesting itself in the aggregate.

I apologize if I’ve mistaken the question. My understanding is that the idea of group selection arose to explain some instances of social behavior which appear to contravene the individualistic logic of selection.


Ze K 01.29.16 at 8:41 am

I suspect it’s much more complicated than this.


Phil 01.29.16 at 9:16 am

I know nothing about any of this, so I’m just going to bring Abraham Maslow out from behind this sign. The peacocks, as you describe them, would drive themselves to extinction – in reality, surely the individual organism would tend to get survival (for long enough to spread its genes) nailed down first, and then build whatever mating practices were called for by its niche on top of that. Similarly with the penguins – you seem to assume that “males compete to supply food” would burn up the resources, either through waste or through females getting hellaciously fat, but either of those side-effects would be directly counter-productive for the survival of the penguins themselves.

So no, it wouldn’t happen like that. (Waits to be corrected by actual ecologists.)


ajay 01.29.16 at 9:18 am

the selection you are describing is not happening at the ‘group’ level: each individual peacock has lower inclusive fitness than any individual penguin — no group level phenotypic features (eg cooperation) come into it.

Yes, this is correct. (And it’s an unnecessarily complex scenario too because you’re bringing in sexual selection pressures). For this to be a case of group selection, you’d have to somehow come up with a scenario where the behavioural difference was disadvantageous at the individual level, but advantageous at the group level, and that led one group to survive over the other. Like, say, non-reciprocal altruistic behaviour between unrelated individuals.


ajay 01.29.16 at 9:24 am

The theory of natural selection was well developed before anyone had heard of genes.

Well, it was and it wasn’t. This is why neo-Darwinism was such a big deal – it gave a mechanism for how advantageous traits might arise and spread without being diluted out of existence. Blending inheritance was the model Darwin worked with and he was never quite happy with it.
The example Darwin used (and boy has it aged badly) was: imagine a white sailor shipwrecked on an island inhabited by thousands of black natives. Now obviously the white man will be brighter and more industrious and more ingenious than all the blacks, so he’ll probably end up doing well with like 3 or 4 wives and lots of kids. And all his mulatto (I think I mean mulatto) sons and daughters will also be much brighter than the natives (though not quite as bright as dad) and so will do well too. But you iterate that through several generations and the result will be that the population has equilibrated and everyone on the island is still pretty much haplessly black. The advantageous mutation of whiteness has just been diluted away to nothing. So how are advantageous mutations ever going to spread without being diluted away?

The answer is of course that traits are inherited as particulates, not blended. Gene theory gave a demonstrable mechanism of how that might happen.


bad Jim 01.29.16 at 9:35 am

Just to clarify: reproduction is generally distinguished from social behavior. Birds are difficult to characterize. Their sexual activity ranges from opportunistic rape to lifelong monogamy, and the amount of care they provide for their young and its division between the sexes is similarly diverse, but the egg, hatchling and fledgling are the responsibility of at most two birds of the same species, not the flock, and the survival of their offspring depends almost entirely on their own competence.


ZM 01.29.16 at 9:35 am

I think if the animals were direct competitors for food they would not share habitat. For instance our predator birds here are eagles and ravens (crows) and the eagles mostly stay in the high areas and the ravens stay in the low areas. then you have other birds like cockatoos and rosellas and magpies and less common birds like kookaburras and owls. And lots of others.

The eagle and raven don’t share habitat as much, they are the totem animals, with the little bat who is supposed to help people find their way in the dark. But the other birds are more or less common. I don’t think peacocks and penguins would bother each other much, or drive each other to extinction. Peacocks are more tropical and penguins are more temperate birds. And you see from the example of our eagles and crows, that when the birds are sort of direct competors for food they don’t share the same habitat as a matter of course, although they might fly through.

We have a nice example of co-operation between plant and animal and human species here, the Manna Gum, or Wurun is the indigenous name.

The Manna Gum sometimes gets attacked on its leaves by a little grub called the djeri, and the Manna Gum leaves release a sweet sticky gum which attracts ants, and the ants eat the djeri grub. The sweet gum crystallises into a little sweet white marshmallow-y bits that then drop on the ground, where if it’s clean people can pick it up and eat it as a bushfood treat. it was called Manna by the Europeans who must have thought it was like the Manna from heaven in the bible.

This interesting mostly co-operative interaction between the tree, the grub, the ants, and people is what gives the Melbourne Indigeous people their name – the Wurundjeri people.


PGD 01.29.16 at 9:42 am

Have you looked up the debates that happened (and are happening) around the EO Wilson, Nowak, and Tarnita paper in Nature on eusociality in 2010 (http://www.nature.com/nature/journal/v466/n7310/full/nature09205.html), the various responses (http://www.nature.com/nature/journal/v471/n7339/full/nature09831.html) and the attempts at a synthesis (http://www.bu.edu/writingprogram/journal/past-issues/issue-6/hammond/)? I think that’s a useful debate for getting a sense of the issues


Jeff Johnson 01.29.16 at 9:44 am

I think (b) is correct. There is much confusion about the meaning of group selection. Clearly group membership confers benefits for social animals.

But what is the unit of selection, the group or the individual? Groups do not replicate, individuals do. Natural selection is selecting the genes in replicating individuals that represent adaptations leading to success. The mechanisms for success may be speed, senses, flight, etc. including the ability to engage in cooperative group behavior.

Still groups do not replicate and groups are not selected as aggregates. Groups are a mechanism whereby individuals succeed and are selected.


david 01.29.16 at 9:51 am

the fact that this isn’t group selection has been pointed out already

it should be added that biparental care is not always the most fitness-improving – one could, e.g., instead look for resistance to parasites, where conspicuous display will be more helpful. The conditions which favour the “peafowl strategy” are not quite the same as those which favour the “penguin strategy”, so it’s questionable under what conditions these could coincide

on the nitpicking theme, some peafowl species are serially monogamous a la penguins


Anarcho 01.29.16 at 10:06 am

Yes, group selection is often confused with mutual aid (aka, co-operation) and co-operation is often confused with altruism. There is a lot of confusion — usually driven by political needs (especially the Rand influenced propertarian right).

Group selection tended to suggest that an individual member of a species would sacrifice itself for the benefit of the group. This is at odds with Kropotkin’s mutual aid which argued that co-operation between individuals forms a group and this bolsters their chances of survival. In other words, co-operation (mutual aid) gives an evolutionary advantage over competition (mutual struggle). Needless to say, Kropotkin never denied both existed (as the subtitle of Mutual Aid indicates, it is A factor of evolution!).

I discuss Kropotkin’s ideas and how well they have aged (very well!) in my pamphlet Mutual Aid: An Introduction and Evaluation (this is the first edition, the second edition is available via AK Press). I also cover it somewhat in my introduction to Direct Struggle Against Capital: A Peter Kropotkin Anthology.

Also, Stephen Jay Gould’s Kropotkin Was No Crackpot is worth consulting (I discuss it in my introduction).

Ironically, group selection seems to be making a comeback — probably because the gene-perspective does not exclude individual’s sacrificing themselves for the group!

Hope that is of interest…


Peter T 01.29.16 at 10:36 am

First, not an expert.

Couple of points: genes were originally seen as the primal unit, and this view remains strong (eg with Dawkins). But it’s now clear that genes are more like words in a language than atoms – what they do depends on multiple contexts (so the same gene can do different things, different genes the same thing, genes in cooperation do things differently than when alone and so on. And then there’s the burgeoning field of epigenetics – feedback between genes and environment – which mean that the same genome can do different things depending on environmental cues. So where is the continuous information stored? In the gene, the genome, or the genome and the environment? If the latter, then the environment includes the group to whatever degree that confers advantage.

Given this, and the common requirement for some minimum level of population greater than two necessary to sustain a species’ existence (and therefore the existence of the individual members of the species), there are clearly some effects at higher levels than the individual.

If you ask “what is the minimum size of a human population” the answer seems to be around 1000 (the smallest size a language can sustain itself), which in itself points to strong selection for large group size. Through what mechanisms this is maintained remains to be fully understood.


John Quiggin 01.29.16 at 10:36 am

@8 That’s getting close to the way I perceive the terms to be used, but I don’t think it can be sustained. Suppose we add in a third population in which competition for mates takes the form of combat between males. Isn’t that a group level phenotypic feature?


ajay 01.29.16 at 10:46 am

27: I think you’re getting yourself on a bit of a red herring with this sexual selection thing. Just posit two species of deer, Leapers and Creepers. Leaper deer have a genetic trait that means they spend one hour a day pointlessly jumping up and down. Creepers don’t. The two species are otherwise identical and coterminous in habitat.
Now, do Creepers gradually outcompete Leapers and drive them to extinction? Of course they do. Leapers are wasting time and energy; Creepers aren’t.
But that is definitely not an example of group selection. It’s an example of interspecific competition that is perfectly explicable in selfish-replicator terms. Leaping (or Creeping) is a “group-level feature” in the sense that all members of the group have it, but that doesn’t mean that it is a product of group selection any more than legs are a product of group selection.


Doctor Science 01.29.16 at 11:30 am

My undergrad training in evolutionary biology was in the mid-70s, and we got a *lot* of stuff on why group selection was (mostly) bunk. The subtext was always race (see, e.g. Darlington).

My memory of the debate is flavored by my biases, I guess, because what I remember is the math. Basically, group selection has to be *really strong* to overcome individual selection — it falls off as the power of 2 of the genetic distance. (not sure if that makes sense to you …) If the group selection debate these days is about altruism, that’s because all the other characteristics which, before the 70s, were explained via group selection turned out to be better explained by individual selection and/or inclusive fitness.

As others have said, the examples you gave are not of group selection, but of competing reproductive strategies (at best). Are these the kind of examples that you’re used to seeing? If so, avoid those sources, for they are bunk.


Walt 01.29.16 at 12:33 pm

I read a bunch about this after the EO Wilson paper that PGD linked to, and I came to a similar conclusion as Plutonium Kun: that the extent of condemnation is driven by academic politics more than anything straightforwardly scientific. It may not even map to real-world politics — academic politics can be intense for entirely internal reasons.


BenK 01.29.16 at 12:44 pm

Group selection is a tricky topic, and it relates in a sense to Hamiltonian fitness. The basic rule, which is being expressed by Doctor Science, is that group selection has to be really strong; and when it is so strong, it is hard for the species to be direct competitors as you describe in your thought experiment (such thought experiments tend to be misleading because they are not grounded in a real example, and so while something ‘could’ happen, this does not reflect the ‘practical importance’). Further, there are forces at work that typically weaken group selection, like cheating. If the forces of cohesion are strong enough to overcome issues like cheating, then you start to have a ‘major transition’ as it is called – the question of individuality is broached, and you may have ‘eusociality’ with reproductive specialization, or multicellularity, etc. Simply by recognizing that individuality exists on a continuum turns the argument on its head and defuses it. This is the same play that defuses Dawkins’ selfish gene dogmatism – levels of selection. The levels of selection are all at play all the time; degrees of linkage are continuous.
I think you’ll find that there is some truth to the arguments made here that science has been driven by ideology and personality; and the comments here reflect echoes and continuations of that. Caution is advised.


david 01.29.16 at 12:55 pm

John @27 Suppose we add in a third population in which competition for mates takes the form of combat between males. Isn’t that a group level phenotypic feature?

Is that disadvantageous at an individual level? (i.e., is that in itself an example of group selection?) Probably not. Where there are losers, there are winners.

Is that group likely to lose to the “penguins” (as an example of penguins exhibiting group selection?) First – why would it? The conditions under which such intraspecific competition is favoured are when biparental care is irrelevant (perhaps because food is abundant). Male contributions in terms of sustained parental effort is irrelevant; male contributions in terms of disease resistance or such traits is far more appealing (again, as an individual rather than group trait). Under such conditions, the penguins are going to get their heads flushed in the toilet. Conversely, under the conditions more typical of the Antarctic circle, individual males who fail to invest in the care of their young will have poor fitness. The incidence of intraspecific competition is not an arbitrary event.

Second, suppose we do have Antarctic conditions, so that the “penguins” do have greater fitness. Why is that a group trait? Each individual penguin will still improved fitness compared to non-penguins.


Ronan(rf) 01.29.16 at 1:01 pm

JQ, I read this a while ago, but from more of an ignorant starting point, while trying to understand the topic. I found it useful enough, though it might not get into enough detail



Bill Benzon 01.29.16 at 1:10 pm

First, I’m not an expert on group selection, but it’s something I’ve thought about with some care and to the limits of my capacity. As a result I’m basically agnostic – my capacity doesn’t allow me to go deep enough into the technical literature – but I do think there’s conceptual confusion sloshing around in these debates.

Count me among those who think the example you give is just plain-old interspecies competition, which is not what the group selection debate is about. In my underestanding, Doctor Science (#29) and BenK (#31) are pointing out the central issue.

I’m copying out a passage from an essay review of a book by DS Wilson on religion that I did some years ago (Rock Art in Darwin’s Cathedral). I hope it helps.

* * * *

What we observe, in animals as well as humans, is that individuals often do things that benefit the group but that cost them individually. As an example Wilson considers a hypothetical group of birds where individuals will emit calls warning the group of predators but also attracting the predators attention to them as individuals. These individuals are likely to attract the attention of these predators and so are more likely to be killed than others in the group and less likely to reproduce. Conversely, individuals who never emit warning calls nonetheless benefit from the calls of those who emit warnings and are thus more likely to reproduce. How can the genes favoring warning behavior reliably survive from one generation to another if signal-prone individuals are less likely to reproduce than “free-loading” individuals?

Darwin’s solution to this problem was to imagine that we have many groups and that those groups are in competition with one another. Groups containing altruistic individuals are more likely to survive than groups without them. Thus, we now seem to have a way of explaining how altruism can be inherited from one generation to another.

This solution, known as group selectionism, requires that the group be a real entity, rather than just some arbitrary collection of individuals. And that is where the problem is, figuring out the conditions in which this is possible. The problem was readily solved in the special cases of kin selection and reciprocal altruism, but a general case has been more elusive.

The problem is that groups must be relatively isolated breeding entities, for that is the only way to keep genetic variance between groups large enough for selective forces to work on whole groups rather than directly on individuals (Boyd and Richerson 1985; Leigh 1999).


Z 01.29.16 at 1:21 pm

Because I agree with all the commenters that your example is emphatically not group selection (just like Darwin correctly recognized that the evolutionary success of social insects like ants is not a prima facie example of group selection, contrary to appearances), let me suggest a pure example of group selection (due to Darwin himself, I believe).

Consider the sex ratio within a species (assumed to use sexual reproduction involving males and females). Because reproduction is (typically) more costly for females (almost by definition, usually), the optimal sex ratio for the species (that is to say the proportion male/female ensuring the maximal reproduction rate, or more technically the ratio maximizing the spread of the allele encoding for it) is usually not 50/50, and among mammals it can be very skewed. However, for any individual, it is a reproductive advantage to belong the less numerous group. Hence, if there exists groups or species with skewed sex ratio, then this is good evidence of group selection. Now, Darwin observes, even among elephants whose gestation plus maternal feeding period commonly exceeds 5 years the sex ratio at birth is an almost perfect 50/50. So, he concludes, group selection plays a negligible role compared to individual selection.

Inverting the logic, if we do find a species whose sex ratio is skewed in the direction predicted by group selection, then this is a prima facie good candidate for group selection. The eusocial spider Anelosimus eximius is often considered an interesting example in that respect: its very skewed sex-ration (even at birth, or in that case in egg sacs) is often interpreted as the stabilizing effect between group selection (skewed sex ratio allow larger colonies, and large colonies of spiders ensure higher rate of survival) and individual selection (in large colonies, many individual do not reproduce).


JoB 01.29.16 at 1:28 pm

Not an expert (in anything).

From a biological point of view, I think whatever stupidities Dawkins is uttering nowadays his contribution on gene selection stands. The notion ‘gene’ is problematic in itself but it is clear enough to exclude group selection (as Darwin already did).

I would however be surprised if JQ was writing on evolutionary models in biology. I see no reason why group selection could not be made to work in a game theory setting. Imagine a game between the group of humans and the group of whales. Clearly the strengths of some individual members in these groups are not very relevant to the outcome. Actually, there’s only one thing which makes the difference: humans act consciously as a group, the whales have little to do with it (at a biological evolutionary scale they might adapt by being either less attractive or more costly to track down but the speed of their game has nothing on the speed of human knowledge development). A more interesting game would happen if (say: Greenpeace vs. whale-killing industry) two groups of humans would compete. Then in the end one group would conceivably win (in the end: the argument) because they adopted the more successful social strategy.

Even in this case (at least let’s hope so) there’s a purely non-group element: which group has the best argument. The flat earth society might not be extinct fully just yet (important group dynamics hold pockets of it together) but I don’t see a social strategy in which they will thrive again.


Z 01.29.16 at 1:31 pm

Oh, and about the implicit question ” but I haven’t found an adequate (to me) explanation of why they [reject group selection].” Biologist tend to reject group selection not because of pure logic but because its putative effects have proven to be elusive. Of course, some evolutionary biologists (Gould, Eldredge, Vrba, Lloyd…), while not disputing the fact that individual selection is predominant, think selection at a higher level does play a significant role. Personally, and as an outsider, I am rather convinced by the high-speciation rate advantage argument within clades in case of mass extinction (selection at the level of the species).


bianca steele 01.29.16 at 2:17 pm

The association of group selection with political issues goes back well before Gould and Dawkins, as the mention of Kropotkin illustrates–as will any anthology of Darwinian-influenced thought. Another example is Hofstadter’s book on Social Darwinism, from the sixties or seventies.


temp 01.29.16 at 3:13 pm

There are two different group selection debates that are related but quite distinct. The first debate is about definition and conceptualization. Everyone agrees that selection sometimes favors individuals reducing their own reproductive output to increase the reproductive output of their relatives. People disagree about what this should be called. Some say “group selection”, some say “kin selection”, some say “gene selection”, some say it depends on the circumstances (some have more complicated views).

The second debate is about, once you decide on exactly what constitutes “group selection”, how common a phenomena is it in nature. The frustrating thing about this debate is that people often decide to engage in it before settling the first debate. This leads to each side talking past each other and nothing getting resolved.


ne 01.29.16 at 3:52 pm

(Not a biologist)

Good question, this was interesting to think about. Here are some stylized facts for why group selection is difficult:

1. Populations of organisms, with genes of different frequencies, get selected* by their environment so that their phenotypes roughly ‘track’ the environment (climb the local hills in design space).
*Technically they all make copies and the least-successful copies are destroyed in the process of survival and reproduction, leading to a change in population frequencies over time.

2. Each organism (genetic pattern) is ‘trying’ to make the most copies it can. It competes for various resources in the environment, including food (against members of its own species, and rival species) and its own body (against predators).
Those patterns that make less copies gradually disappear from the population over time.

3. There are usually gains from cooperation— if two or more organisms work together and coordinate, they can get bigger payoffs than either of them could alone. However, there are almost always bigger payouts from defection-exploitation, and so cooperation needs some way to ‘pay its way’ or change the payoff matrix in order to make the larger payoffs reliable. Cooperation can arise from:
-Shared genetic patterns being copied across the pool of neighboring organisms (kin selection)
-Stable cooperation from iterated games (reciprocal altruism)
-Simple individual benefits from group coordination (mutualism)
-Incentive-alignment (costly punishment)

4. “Group selection” is about individual organisms that make population-level cooperative sacrifices, but *their genetic pattern is not paid back in fitness by one of the compensating mechanisms in #3*. Their pattern decays from the population, but the population expands to outcompete neighboring populations at a speed great enough that the individual altruist decay is dampened.
Classic examples would be a prey sacrificing itself so that the non-kin group can survive; or members of an overpopulated species sacrificing themselves so that the non-kin group doesn’t starve.
These patterns are not group-level stable because patterns that did not self-sacrifice would survive and copy in the long term. So to be stable, the groups that contain that have to beat other groups to a compensating degree.
Groups split into new groups much slower than organisms split into new organisms, so it’s hard to get the numbers to work out.

5. Cooperation and altruism are the focus of group selection debates because ‘increasing the total social fitness of a group, at the expense of the pattern’s individual fitness’, is the only puzzle that group selection could hypothetically solve, as far as I can see. What would be another example?

6. It may be clarifying to distinguish three different things:
A. Microfoundations altruism— kin selection, reciprocal altruism, mutualism, altruism-with-punishment-for-defection. These all make sense and are sustainable in a fitness economy— they pay their way in the fitness of individual patterns.
B. Group selection altruism— cooperative behavior that costs the individual more than it benefits (including in nearby kin patterns), but has such high returns for the group that groups with these individuals destroy and replace neighboring groups, ~even though the hypercooperative individuals are constantly decaying out of the population at the margin~.
C. Competition between different species that do not interbreed. This is just a separate topic, since there are no “genetic patterns sustaining themselves through survival of the groups they’re in” puzzles.
The penguin/peacock question seems like a case of C, and so even though it involves “groups competing” it’s not what biologists mean by “group selection”.

7. Note that group selection involves variants of genes within a population, with the self-sacrificing cooperative genes finding a way to remain competitive with non-self-sacrificing genes via group expansion.
When the two groups have separate gene pools, as in penguins and peacocks, this is not a question of group selection, just two different species competing for resources.
The similarly comes because both engage in cooperative behaviors, and both are sort of competing at a group level. But note that both the cooperative behaviors, sexually selected tails and food giving, are of Type A rather than Type B. There’s no group selection puzzle here.


mbw 01.29.16 at 4:11 pm

As others have said, your example is simply not on topic because your groups are different species. Here’s the issue- and it’s a quantitative, not qualitative one. Take some trait that’s advantageous to the group if spread throughout (say courage in battle) but disadvantageous to an individual. There are two competing processes in a group where the trait is common. The group tends to out-reproduce others. Within the group, however, the non-trait gene out-reproduces the trait gene. Which wins? Does the gene become more common?
The anti-group-selection people point out that it’s almost impossible to find parameters under which such a trait wins in this simple scenario. One tweak, however, is that if another trait emerges (sexual selection for trait-holders, punishment of non-trait holders) the individual level-selection can now align with the group-level selection. So there are ways that group selection can occur that are less implausible than suggested by the simplest models.


Theophylact 01.29.16 at 4:11 pm

Hey, wouldn’t it be nice if you could get an actual evolutionary biologist — an expert, that is — to weigh in here, instead of a lot of “I am not an expert” whoevers? Not condemning their contributions, but I’d like to see something better than guesswork. Maybe send PZ Myers an email? Or ask someone at href=”http://ncse.com/?gclid=CPObkJG0z8oCFcwYHwodY-kIHw”>NCSE?


Theophylact 01.29.16 at 4:12 pm

Damn. Preview, PLEASE!


mbw 01.29.16 at 4:17 pm

I forgot to add: The reason that this whole question makes sense only within species involves the question of what happens to a pure-trait group. You might think that it will just benefit, since there is no within-group gain of the non-trait. In any realistic within-species picture, however, there’s substantial genetic exchange with out-groups, so there’s a always a chance of the non-trait gaining within the group.
This also suggests a plausible mechanism for the evolution of auxiliary features stabilizing the trait. Say that the group has evolved anti-out-group behavior (as many do) for unrelated reasons. If the non-trait serves as a marker for out-groupness, then the group is pre-adapted to punish the non-trait. Then all that’s required to get a stable evolution of the group trait going is one little group where it happens to be prevalent by chance.


marcel proust 01.29.16 at 4:31 pm

When I have generic genetics/evolution questions, I often drop them into an open thread at the blog Gene Expression (about 1 every weekend). Whatever you think about RKhan’s politics, he is very knowledgeable on these issues (also widely read in history) and seems willing to tutor the untutored or at least point them to appropriate sources. IIRC, TN Coates approached him several years ago with questions about genetics and race and blogged about that at The Atlantic (also here).


BenK 01.29.16 at 4:43 pm

To calm Theophylact #42, I am a working bacterial eco-evo biologist who has published evolutionary game theory, non-transitive selection,… but, the degree of specialization is serious. I am focused on bacteria. As a result, I find myself caught a tad off guard when I need to think about diploid organisms, for example. Bacteria have more social behaviors than one might think, but many of the common features and questions that evolutionary biologists think about in plants and animals are rather off my radar screen.


lw 01.29.16 at 4:50 pm

Group selection is am empirical fact: http://www.ncbi.nlm.nih.gov/pubmed/8786932

I think it’s a pretty rare effect in actual populations in nature. Myself, I prefer to avoid speculative discussions of human behavior as a phenotype. There’s a discussion of the (interesting) chicken paper here:


mbw 01.29.16 at 4:54 pm

As some here have suggested, and others seem to have muddled a bit, group selection is not a synonym for general higher-level selection. Species-level selection must occur but it can’t do the sort of evolution of complex functions that individual level selection can. A few million species for a few million species quasi-generations is just pitiful compared to the numbers of individuals and generations needed to evolve a complex trait like say an eye.


Donald Johnson 01.29.16 at 4:59 pm

I’m not an expert either, but occasionally read a little of this stuff. However, I’m pretty rusty even on what little I know (or knew).

You might want to do some googling on the Price Equation, though not everyone thinks it is a conceptually useful thing to play with. But it is supposed to incorporate in one equation both individual selection and group selection. If you want a math-laden book by someone who is a fan of the Price Equation (sounds funny to describe someone as a fan of an equation, but anyway), then read Sean Rice’s book. Actually, since I’ve only skimmed parts of it, I can’t say that it will clarify things. But maybe next time I am feeling ambitious I will tackle it myself. It has a chapter on Price.



Donald Johnson 01.29.16 at 5:03 pm

First piece that appeared when I googled Price Equation had what purports to be a gentle introduction to the subject.



Donald Johnson 01.29.16 at 5:11 pm

Price himself was an interesting guy, who ended up committing suicide.



Marshall Peace 01.29.16 at 5:38 pm

It seems to me that the question is whether “groups” are actual entities, which the “selfish gene” folks vociferously deny. Are social facts real?

Consider Texas circa 1800: your decendents’ future is totally circumscribed by whether you are an Anglo, a Mexican, or an Apache. It is true that you can ascribe the (evolutionary) benefit of being in a group to individuals … that is, reify the group in terms of individual membership … but it’s simpler to say that the Anglos owned the future and the Apaches did not, for reasons related to the respective cultural resources.

Clearly this retail displacement or genocide of group-on-group behavior has been a big factor in the development of human society as we know it, as it depends on and reinforces social life. That the question even arises must be similar to the discussion about whether typical mammals have emotions … the idea threatens the unique privilege of civilization-as-we-know-it.


Glen Tomkins 01.29.16 at 5:58 pm

I think the problem is partly semantic. The evolutionary biologists do seem to accept kin selection. Social behaviors that get your relatives better survival tend to pass on the genes you share with those relatives — to the extent the behaviors are genetically based. If husband penguins foraging for the wife penguins improves group survival, sure, insofar as that behavior is genetically determined or even facilitated, the biologists would accept that as what they insist on calling kin selection. Your example is not exactly spot on, in that the father penguin gets his genes passed on directly by foraging for the penguin fostering his offspring. Change your example to a group of penguins in which third cousins forage for nesting mothers, and you’re there, at kin selection.

This kin selection translates to behaviors within human societies only insofar as genes that promote the survival of society, but tend to keep the individual from passing on his or her genes, are shared in common.

You really, really don’t want to go there. The obvious reason is that we’re now talking about race.

For one thing, the imagined human races are a construct that have no support in biology. We have no reason to believe that, to pick an example at random, there is an Aryan race that could possibly owe its rise to domination of a large chunk of our planet to the superiority of its social interactions as determined by the genes this imagined race passes on to later generations of Aryans.

Human societies as we observe them do not at all seem genetically based. Even agricultural, and more so urban societies, are quite cosmopolitan in respect to intermarriage. We tend to project back to some imagined stage of human societies at which they supposedly were large extended family groups roaming the landscape hunting and gathering. But what we observe among peoples who today still hunter-gather in tribal groups, is a strong tendency to seek intermarriage with other groups, rather than preserve the purity of blood lines. If a society doesn’t do that, inbreed to preserve a shared genome, it can’t have kin selection. They could have the world’s best individual behaviors for preserving their society, and those behaviors will absolutely not even a little bit, be passed on in the genes. There will be no genetic, biologic, group selection, without kin selection.

Your society can only pass on behaviors genetically, through biologic evolution, to the extent that it is racially pure. And, since we human beings have never done this to ourselves, practiced the same inbreeding of desirable traits and culling of undesirable traits that we’ve known for thousands of years — from animal husbandry — is how you isolate a breed genetically, we don’t actually have any actual human races on which to practice purity.

One conclusion you can infer from the absence of actual human races, is that we’ve found that there is a high survival value for societies that don’t go in for racial purity. Racial purity is this idea that keeps coming back despite what has to be a terrible group selection bias against it, because societies must end up doing themselves horrible damage if they succumb to the temptation — presumably derived from our experience with animal husbandry — to think of its members in racial terms. I’ve already gone Godwin by talking about Aryans, so yes, the Third Reich is the obvious example.

It is perfectly reasonable to think of human societies in evolutionary terms. We do pass things on from generation to generation. But we do so socially, not genetically. We have learned to avoid any attempt to pass traits on biologically, genetically, because of the awful social consequences. Also, socially desirable traits are presumably much more complex that what we find desirable in, say, a draft horse. You can’t breed for it if you can’t identify it.

Don’t give in to Physics Envy! Don’t be a self-hating social scientist! Just because you can’t ground social group selection in biology does not make it any less real. Accept that the two spheres, the biological and the social, are perhaps often analogous, and have some points of peripheral contact, but are nevertheless quite distinct.


Chris Stephens 01.29.16 at 6:01 pm

Hi John,

I think the answer to your question is (b). One of the best attempts at disentangling the issues surrounding group selection is Samir Okasha’s book, Evolution and the Levels of Selection. Incidentally, he is a colleague of Chris Bertram’s.

I can’t really tell what you’re after, but ch. 7 of his book on species selection might be of help.

Biologists usually think of group selection as an issue WITHIN a given species – think of Darwin’s original example of human morality (groups of humans with individuals willing to die for their fellow group members can out-compete groups that don’t have such individuals). In Sober and Wilson’s framework (Unto Others), they allow that group selection can be at work even if it is not a case of altruism – it is just that the altruism cases are the ones that seem to require group selection.

at any rate, I’d look at Okasha if you haven’t already-



A H 01.29.16 at 6:07 pm

The problem is basically one of micro foundations. Groups are not coherent units of selection because they neither replicate nor face selection as a group.

The study linked by lw at 46 is the exception that proves the rule. In that study they artificially created an environment where groups fit the requirements to be a coherent unit of selection.

Some of the most interesting things going on in evolutionary theory now are what happens at the edges between different definitions of units. How did eukaroytes evolve from groups of different non-eukaroytes? How did multicelluarity rise?


Bill Benzon 01.29.16 at 6:29 pm

The link that Ronan(rf) gave up there in #33 is worth looking at. It contains a piece by Steven Pinker arguing against group selection and is followed by replies from a number of thinkers, pro and con, with final remarks by Pinker. The respondents include a number of prominent advocates of some version of group selection including DS Wilson, Pete Richardson, Rob Boyd, Herb Gintis, Jonathan Haidt, and of course those against, including Dawkins. I particularly recommend the comments by David Queller (biologist) and Joseph Henrich (anthropologist).


david 01.29.16 at 6:50 pm

Bill Benzon @34 – How can the genes favoring warning behavior reliably survive from one generation to another if signal-prone individuals are less likely to reproduce than “free-loading” individuals?

Groups of birds are closely related; inclusive fitness prevails.

Z @35 – Inverting the logic, if we do find a species whose sex ratio is skewed in the direction predicted by group selection, then this is a prima facie good candidate for group selection.

But, again, kin selection is quite sufficient to explain eusociality.


BenK 01.29.16 at 7:16 pm

Just because the genetics of races came up, this preprint is apropos.


Marshall Peace 01.29.16 at 7:38 pm

A classic discussion of the problems for group selection in genetic or more broadly Darwinian evolution is George C. Williams (1974) Adaptation and Natural Selection. Good stuff, AND it’s important to realize that we aren’t doing Darwinian genetic evolution all day every day.

EG, “race” would be a valid concept in sociological evolution just because it is socially constructed. Likewise class, nationality, ethnicity, fashion choice, etc etc etc.


engels 01.29.16 at 7:51 pm

the extent of condemnation is driven by academic politics more than anything straightforwardly scientific

That was my gut feeling from reading Dawkins (many moons ago).


protoplasm 01.29.16 at 8:46 pm

If Peter T at #26 has anything like the spirit of the controversy in his sights, it makes me ask: why the insistence on foundationalisms, what is the attraction of a sort of “primal” individualism, and why the resistance to contextualisms? (I wonder, for example in economics, what we lose by discarding the idea that macro admits of micro foundations. Would macro being sui generis, with respect to the individual firm, cause insurmountable analytical difficulties?) On pain of attributing sloppy thinking and ignorance of the is–ought distinction to scientists, I’m assuming that debates about individualism in, say, biology or economics are wholly unrelated to what passes under the name of “individualism” in ethics and politics.

Any particular analysis will have its individual, but it seems to me we lose nothing by acceding to the idea that it’s the analytical frame that helps set the stage for what will count as an individual; nature hasn’t cut herself at the joints, we have, and it’s not a problem that the knives we choose and our purposes in so cutting contribute to our final pictures of her. What seems odd to me is that it is the insistence on foundations and a unique individualism, rather than contextualism or a many-levels understanding, that are most akin to theology.

Furthermore, and probably I’m going off into the weeds now, is there really no sense to be made of groups replicating? Or of environments replicating?

Finally, if lw at #46 and A H at #54 have it, then group selection isn’t conceptually incoherent or biologically impossible, it’s merely rare. Was the Dawkins/Gould debate about impossibility or frequency/preponderance?


Jack Morava 01.29.16 at 9:11 pm

Just a shout-out to say that George Price’s work seems significant. I don’t know much
about this, or about the history of the subject, but it seems to me that it’s only starting
to get the attention it deserves.


BenK 01.29.16 at 9:24 pm

Scale matters in ecology and evolution.
Most, if not all, results in these fields are intractable without simplifying assumptions; frequently of the nature that the factor or the change in the factor are set to zero. Setting the right things to zero is a huge issue; and differing assumptions lead to entirely distinct intellectual traditions, methods, questions, choices, and so on. It isn’t mere hand-waving to ask whether this or that force is significant at a particular scale. If sexual reproduction is ‘vanishingly rare’ in a group of organisms, then bifurcating or bush-like patterns of heredity form. If mutation is very rare, then parsimony can be used to reconstruct heredity. If populations are infinite, then drift goes to zero, and selection is the only pressure of significance. And so on.


Glen Tomkins 01.29.16 at 9:36 pm

What’s wrong with the example chosen

I gave a long-winded response earlier, so perhaps a more direct answer to the specific question posed would be more user-friendly.

The example as written doesn’t even involve kin selection, much less group selection. The male penguin who feeds his mate thereby increases the survival chances of his own offspring. Society isn’t really involved. The same survival value to feeding one’s better half would exist for survivalist penguins who had gone off the grid, as for penguins with thousands of Facebook friends. (

Okay, a society made up of individual penguin families who all did better because their dad’s didn’t neglect their better halves, would do better than a society of deadbeat dad penguins. Welcome to the libertarian world-view. It’s not a perspective that helps understand anything more complicated than direct, biological, evolutionary advantage. This is the analysis that leads to using evolution to support Social Darwinism.)

The biologists would accept this as an example of just kin selection, if three conditions are met. One is that the altruistic penguins have to gather food for anyone in the group, not just the mothers of their own offspring (if this condition is not met, see last para, we’re talking about direct selection). Two is that the socially useful behavior has to be genetically based. Three is that the group has to share the genetic basis of this socially useful behavior.

We don’t know exactly how the behavior in the example works, but it is inherently unlikely that the specific behavior is hard-wired in the genes, so condition two is not met.

Condition three is not met if condition two is not met. That said, yes, peacocks and penguins are clearly genetically isolated. they’re more than different races, they’re different species, incapable of inter-breeding. If these two behaviors of theirs were both hard-wired in the genes, sure, the survival advantage for the penguin society would be an example of kin selection. While peacock display behavior is perhaps hard-wired, feeding other penguins is almost certainly not, so condition three cannot be met.

All that said, even if this were an example of kin selection, kin selection is it, in terms of biological evolution. The group can only pass on any advantageous trait biologically insofar as it is a kinship, with common genes. What is this group selection of which you speak? What room is there for it in the heaven and earth imagined by the philosophy of the biological sciences?

Why not just admit that group selection is a matter of social rather than biological evolution? All sorts of altruistic behaviors will help one society do better than another, but these behaviors, and their passing on to the next generation, have nothing to do with genetics. Why the impulse to ground an observation that has excellent direct observational evidence in biological evolution? Enough already with the Physics Envy. who cares if the biologists don’t know what you’re talking about, and/or think it’s all stuff and nonsense. Physicists think the same thing about them.


Gary Othic 01.29.16 at 9:59 pm

Blast, missed out on a debate I could have actually contributed too (my PhD work covers some of this ground) -_-

Most of what needs to be said is already here, but one thing to be careful with is in distinguishing between the unit of selection and the unit of replication. They are not always the same thing (hence D S Wilson’s concept of multi-level selection).

His article, with Elliot Sober, “reintroducing group selection to the human behavioral sciences” (apologies can’t provide the link at the moment) is good on this, and has comments from others. Herbert Gintis has a book on it as well, though the title escapes me.


pete 01.29.16 at 10:14 pm

I believe Dr. Hilarius in #10 has the best guidance here. I will add the following.
First, the original notion of group selection was put forth as an answer to why altruism appears to exist in some species. So group selection and altruism are linked in the history of evolutionary biology. It is also essential that one frame the altruism/group selection problem in terms of evolving populations of one species. Otherwise one is more appropriately looking at various population ecology models (interspecific competition, predator-prey, etc.). The original version of group selection maintained that there could exist genes for individual behavior that benefits the group (the population) and hence provide fitness benefits that maintain an altruistic population.

To explain the problem with this, it is useful to define altuism. Altruism in evolutionary biology occurs when an individual provides a fitness benefit to another individual(s) in a population while incurring a fitness cost as a result; the fitness cost (reduced relative reproductive success) is essential in defining biological altruism. From this, the dilemma is immediately clear. A gene for altruism would not survive in the face of selfish genes; altruistic genes would be at a relative fitness disadvantage in the the population and individuals with those genotypes and phenotypes would decline in frequency.

To explain altruism, the most common approach is Hamilton’s rule which expresses the costs-benefits of altruism as: C = rB. r = coefficient of relatedness, B is benefit from the altruistic behavior, C is the cost. So rB > C is required for altruism to succeed as an evolutionary strategy. This approach is also referred to as kin selection. This model is the workhorse for analysts working in the areas of cooperation and altruism in evolutionary biology. It also explains the phenomena without resort to group selection.

David Sloan Wilson has other ideas. He and some others (including the esteemed EO Wilson, no relation) maintain that a type of group selection can be saved by using what they refer to as multilevel selection theory. This maintains that the force of selection can act on units besides individuals(genes) within a population. Roughly, it maintains that if the strength of selection on different, separate populations is high enough relative to the strength of selection within a population, then it may be that group selection can occur. For one discussion of this, see Samir Okasha, Multi-level selection, Price’s Equation, and Causality. It puts the theory in terms of the Price equation, which is interesting. You might also look at an old paper by Richard Lewontin (The Units of Selection). Note, though, that this is a distinctly minority position within this field. I believe (may be wrong here) that most analysts would say the likelihood of the type of group selection referred to by the two Wilsons etc. is very unlikely as an empirical matter. Morever, it is maintained that once this small empirical likelihood is granted, multilevel selection theory doesn’t added anything beyond what the kin selection framework already provides.

Personally, when I read this literature, I got the impression that these two approaches looked at the same phenomenon from two different perspectives. The first from the gene-level, the second from the nature of the forces of selection and what biological units are selected and when. Each is interesting. I also got the impression the group selection analysts made claims in criticism of kin selection analysis that were probably off-base.


mbw 01.29.16 at 10:20 pm

@BenK- Thanks for linking to a fascinating article and for your other astute scientific comments.

With regard to many of the comments by others here, it’s a little stunning how quickly they slip from some glancing recognition of an actual scientific question into ideological fighting mode. I’m all for ideological fights, but sometimes a question is just a question.


Bill Benzon 01.29.16 at 10:31 pm

“Personally, when I read this literature, I got the impression that these two approaches looked at the same phenomenon from two different perspectives.”

Both Henrich and Queller (among others) speak to this in the Pinker symposium Ronan(rf) linked to in #33.


pete 01.29.16 at 10:49 pm

Here is a link to a paper I remember that I thought was useful in clearing away a lot of confusion regarding cooperation, altruism, etc.



steven johnson 01.30.16 at 12:00 am

Although Pinker is neither a neuroscientist nor a population geneticist, his piece has been highly praised as a useful corrective to the misapprehensions of group selectionists. So it is useful to highlight some remarks from it.

“But other extensions are so poetical that they shed no light on the phenomenon and only obscure the real power of natural selection.”

Both the handful of group selectionists and their opponents share a commitment to genetic determinism of human behavioral traits. Some like Coyne and Pinker have agree with the likes of Nicholas Wade that “races” are biological reality, though of course they vigorously deny this implies racism. The rest of us should be skeptical of all parties. These are the kind of people who believe in evolutionary psychology. But in addition to being genetic determinists, the larger part of them oppose group selection because they are panselectionists, that is, every trait is deemed to be adaptive. They may not have Paley’s God, but Evolution has its Natural Theology. That too they will vigorously deny when caught at it, but in practice I’ve never found them even considering limits to the power of natural selection.

“Except in the theoretically possible but empirically unlikely circumstance in which groups bud off new groups faster than their members have babies, any genetic tendency to risk life and limb that results in a net decrease in individual inclusive fitness will be relentlessly selected against. A new mutation with this effect would not come to predominate in the population, and even if it did, it would be driven out by any immigrant or mutant that favored itself at the expense of the group.”

I do not know how these people know that natural selection is not just relentless but effective. But it is explicit here that it is the unfailing power of natural selection that prevents the emergence of individually deleterious traits and its spread to predominate in a group. I see two basic problems here.

Less importantly, kin selection, even when it is more discreetly renamed inclusive fitness, makes no sense as a display of the power of natural selection to weed out negative traits, even assuming genetic determinism. It makes much more sense to say that an episodically or contingently or randomly negative trait cannot be weeded out of a population because the genes prerequisite for the trait are still carried by the kin. But this is definitional/conceptual, and I suppose confusionism in such matters is acceptable.

More importantly, the implication that groups cannot replicate faster than the members have babies is more or less nuts when you’re talking about people. Human groups fission and fuse with bewildering speed, changing the genetic compositions of the groups. Personally I should think such nonsense is like a small turd in a large punch bowl: It’s quite enough to ruin it all.

“Compared to the way people treat nonrelatives, they are far more likely to feed their relatives, nurture them, do them favors, live near them, take risks to protect them, avoid hurting them, back away from fights with them, donate organs to them, and leave them inheritances.” And…Women are from Venus, Men are from Mars. Incidentally, it’s not so clear, except to those who insist on the power of natural selection, that the variety in human beings, on a genetic level, is comparable to the variety in other longer extant species. But as one readily sees, you could exchange “nonrelatives” for “strangers” and “neighbors” for “relatives,” and this statement would be just as true…because it is just as vacuous.

At any rate, mating is usually carried out with non-relatives. It is not clear how natural selection can produce behaviors that favor kin rather than strangers, except when they might turn out to be mates (and maybe offspring who will become kin) without some kind of prophetic power. It’s rather like people spontaneously, without reward, trying to help strangers’ children because of inclusive fitness, as opposed to adopted children. But enough, Pinker is very slick and it takes much more effort to explain the tricks than for him to do them.

The remarkable thing is, though Pinker has a background in linguistics, he has nothing to say about language, possibly because this is a group trait. Personally, being neither a genetic determinist nor a panselectionist, I don’t think group selection is as determinative of human culture as sexual selection is in male and female personality according to evolutionary psychology. But as an explanation for the emergence of something like language, something you might call human nature rather than culture (way of life,) group selection so far as I know is a reasonable candidate. At this point, it seems to me likely it is the only significant genetically determined behavior.


Jeremy Fox 01.30.16 at 12:07 am

Actual evolutionary biologist/ecologist here, who has published papers on the Price equation and its application to problems in ecology.

Hard to give a short answer–this is a big and technical literature, and it continues to be divided into opposing camps, who are mostly arguing about non-empirical matters. The relative merits of different ways of defining terms, for instance. Also different ways of interpreting the same, or closely related, abstract mathematical frameworks such as the Price equation. As a commenter above noted, the dustup over Martin Nowak’s recent Nature paper with Tarnita and Wilson is the latest flare-up of some deep-seated conceptual disagreements.

Philosopher of biology Samir Okasha’s work is one fairly accessible and fair-minded entry point into this literature. His book Evolution and the Levels of Selection is very good, though I don’t agree with all of it. If you don’t want to bite that off, his 2004 paper in Evolution is a nice summary of what in my view as one of the key insights of the book: that we have conflicting intuitions about what “group selection” is. His 2015 paper with Jonathan Birch in BioScience is an excellent teasing apart of what’s at stake in the dispute between Nowak et al. and their many critics.

Another good source on the Price equation (which crops up a lot in discussions of group selection) is evolutionary biologist Steven Frank. He’s probably the world’s leading expert. Have a look at some of the papers in his recent “natural selection” series in the Journal of Evolutionary Biology: http://stevefrank.org/natSel.html


mbw 01.30.16 at 12:22 am

On Wilson, Nowak, Tarnita: That whole paper and dust-up was a sad event. One point of the paper was that the original bar-napkin version of kin selection was an approximation only valid in a smallish range of parameter space, and that typical teaching didn’t emphasize that. It would be as if we taught that all fluids are ideal gases. Good point.
The rest of the paper was a philosophical disaster, in which they basically claimed that if you wrote down a tautological expression “exact fitness=exact fitness” you’d made an important scientific claim.


Doctor Science 01.30.16 at 1:05 am

To go back to John Quiggin’s original question:

It seems to me there are only two possibilities here
(a) My reasoning is wrong, and we can’t judge which species, if either, will dominate; or
(b) Even though it involves one group being selected over another, this isn’t what is meant by group selection.

I would say the answer is “both”. Your reasoning is wrong, AND this isn’t what is meant by group selection.

Whether a peacock (no male contribution to offspring) or penguin (much male contribution to single offspring at a time) strategy is selected will depend on all kinds of things you haven’t stated, especially how much of the danger the offspring experience is random, versus how much can be averted by parental investment. Your assumption that the penguin strategy will always be better is not correct.

But also, as we’ve been saying, what you describe isn’t group selection.


Michael 01.30.16 at 3:30 am

Hi John,

I’ve been trying to get my head around it too, and it seems to me that there are a few different issue being conflated, which leads to people talking past each other. A thought, which I hope is helpful:

I think that the emphasis on co-operation is partially because this is needed for a group to exist in any useful sense. ‘Sociability’ might be a better (broader, less loaded) term. So I agree with some others here that the comparison between penguins and peacocks seems like competition between two strategies, not two groups in the relevant sense – even though both strategies in your example do involve reciprocal behaviour within ‘groups’. Group selection is about groups as entities (*or something like that* – again, people seem to talk past each other a bit). My go-to example is a case of a species which has two cultures – some groups are stag hunters, some are hare hunters, and the groups are competing as teams for other resources like water etc. There are so many ways that the situation (imaginably) can go from here – stag hunters learn to cooperate in other ways on the basis of their acquire stag hunting skills, or periodically become too successful for their own good and wipe out their prey, or the two groups diverge from each other toward speciation… But perhaps the key question is, whether the groups themselves can survive in a substantial enough sense for us to see them as entities with a genealogy.


Peter T 01.30.16 at 5:18 am

Just to add, evolutionary theory looks to explain whether and under what conditions group selection might operate genetically. Surely game theory should look to different models? Ibn Khaldun’s notion of “asabiyah” (group solidarity), its formation and dissolution, and Lanchester’s explorations on the advantages of numbers in combat (combat power rises as the square of the number of people, so long as they cooperate) would seem more relevant.


Bruce Wilder 01.30.16 at 6:40 am

Your example behaviors revolve around sex, but you have not seemed to think it necessary to think about how sexual reproduction affects the problem of variation in descent. No individual is simply reproducing herself from one generation to the next.

There is a gene pool, mixing in the gene pool, and so on. One of the advantages of sexual reproduction is, paradoxically, that it resists the elimination of variations that presently confer no advantage. The gene pool, with mixing, is being reproduced from one generation to the next, but even environmental conditions that last several generations will not necessarily eliminate variation with no current fit to those environmental conditions. The potential remains, perhaps hiding as a recessive trait.

The sexually reproducing species has the advantage that it retains this latent fitness, and can manifest it as adaptability in response to sudden shocks. A time of strong selection in relation to one particular challenge from the environment is not enough to impose a rigid homogeneity on species potential.

This is group selection in the sense that variation in the gene pool is enabling the species to survive environmental shocks. But, it is variation in the pool, not a particular shared and defining characteristic shared among representatives of the species, that enables survival.


Z 01.30.16 at 8:50 am

@david “But, again, kin selection is quite sufficient to explain eusociality.”

As Darwin pointed out, individual selection is quite sufficient to explain eusociality. However, individual selection has a hard time explaining skewed sex-ratio in egg sacs and IIRC kin selection has been ruled out as well for Anelosimus eximius.


bad Jim 01.30.16 at 9:35 am

It ought to be noted that, in the anthropocene era, group selection may have become a dominant influence when humans hunted various species into extinction and marginalized others through competition and habitat alteration. Where are the charismatic megafauna of yesteryear?


JimV 01.30.16 at 4:19 pm

It’s probably been said (and better said), but the penguins and peacocks in the scenario are not competing as groups (of breeding pairs), but as individuals (individual breeding pairs). Put one breeding pair of penguins and one breeding pair of peacocks in a bio-dome with finite sustainable resources and you get the same situation.

My uniformed guess is that altruistic groups and selfish sociopaths can exist in a kind of equilibrium in a population, like parasites and hosts. (Jack “What idiot is loyal to a company?” Welch would have liked to lay off everybody but himself but wouldn’t have survived himself if he did.)


roger 01.30.16 at 10:20 pm

Until someone can propose a falsifying experiment, this whole discussion is about something other then science

this suggests a test: altruistic acts (a bee stinging an invader, and dying, to protect the queen) are determined by the genetic relatedness of the altruistic person


John Quiggin 01.30.16 at 10:43 pm

Thanks to everyone, and particularly to Jeremy Fox. The discussion in terms of the Price equation fits neatly with the game-theoretic problem that got me started. In the standard presentation, everything is divided by w, the average or standardized fitness. Roughly speaking, w corresponds to the average payoff for players in a game.

In (non-zero-sum) games, changing the set of strategies will, in general, change the average payoff. Thinking about evolutionary interpretations of games, the question occurred to me “how should we interpret the change in average payoff associated with a different strategy set?”. In terms of the Price equation, that means, I think “how should we interpret a change in w?”.

The interpretation that occurred to me, as stated in the OP, is that a group or species would do better, both absolutely and relative to other groups, if its w were high, than if it were low. (That obviously gets into dynamical problems, which I’m still trying to get on top of).


BenK 01.30.16 at 11:13 pm

Relying on average/normalize/global fitness is a grave problem with many models. Particularly in game theoretic treatments, it is undesirable. Dealing with finite, particularly structured, populations requires different approaches. Moving to graphs is one extreme (a la Lieberman, Hauert, and Nowak 2005; but also Adami, or Kerr, in many publications). Anyway, many relevant studies about cooperation, cheating, non-transitive fitness, and so on have progressed beyond global fitness/average payoff.


SamChevre 01.31.16 at 1:08 am

I’m the opposite of what you asked for–I’m an amateur in a related area–but your #81, I think, is missing a key point.

Group selection, in many cases, is about the fact that the payoff regime switches; you need some form of regime-switching model for w. In that case, I would expect the traditional regime-switching result to hold: the maximum achievable value of w across regimes will be lower than the achievable value given a a regime. Basically, some amount of strategic behaviour is useless given the regime, but is critical to a different regime.


Gavin Kostick 01.31.16 at 9:49 am

Very interesting discussion thanks.

At the risk of making an idiot of myself.

Two groups of creatures are identical in every way except in one all individuals are entirely selfish and only look after themselves while the others are altruistic and look out for each other.

A catastrophe comes and survival can only be achieved for a proportion of the group if they co-operate and some individuals accept they risk of, and/or will definitely loose their lives, for the survival of the remainder of the group.

The selfish individuals of the first group can’t do that and they all die. The altruistic members of the second group can do that and so some survive.


James Wimberley 02.01.16 at 12:45 am

Gavin Kosgick: one classic example ( possibly from Jared Diamond) is care for old women in Pacific island tribes exposed to hurricane disasters at long intervals. The old women, who don’t pull their weight in normal times, are the stores of knowledge about emergency foods you have to resort to after the hurricane. It’s very plausible that tribes that just let the grannies die don’t get through the hurricanes. This is cultural selection at work, not genetic.


Kiwanda 02.01.16 at 1:37 am

I don’t quite see how the framework of the Price equation would yield an explanation, as opposed to a way to concisely express the consequences of an explanation. (Or maybe I’m expecting too much: the same might be said for Newton’s laws of motion.)

Also: whether or not it sits under the heading of the biology topic of “group selection”, the setup in the OP still raises a question that I haven’t noticed a direct answer for: why wouldn’t the “useful” male penguin reproduction strategy always outperform the”wasteful” peacock strategy, and drive it out? That is, why do we observe peacocks at all, in the sense of “male birds using expensive displays to attract females”?

Along similar lines, males do not produce children, and so the resources consumed by them are not going into the next generation, which would arise faster without them. So: why aren’t males tiny? Maybe the situation of herd animals amounts to this: one male becomes “herd bull”, and fathers most of the herd offspring, while the losing males wander off and, without herd protection, die. So the resources consumed by males are less than those devoted to females, which is advantageous for the species.


Dr. Hilarius 02.01.16 at 4:33 am

James @ 85: It’s also likely to be genetic selection. In most species, individuals die at about the same time as they become post-reproductive. There are a number of species that exhibit prolonged post-reproductive survival. Post-reproductive survival, like longevity in general, has a genetic basis. Killing grandma when she becomes a burden is cultural. The fact that grandma is capable of living for several generations longer than her own fertility is not cultural. For a recent review see: http://www.sciencedirect.com/science/article/pii/S0169534715001044


SamChevre 02.01.16 at 1:21 pm

Kiwanda @ 86

why wouldn’t the “useful” male penguin reproduction strategy always outperform the”wasteful” peacock strategy, and drive it out.

Because food isn’t always the constraint. If selection pressure comes from parasites, but food is abundant, the “look how healthy I am” strategy of peacocks is better.

The question group selection tries to answer (as I understand it) is how is it that you see both strategies in the same species, when one is clearly better for several generations at a time.


Alex SL 02.01.16 at 9:05 pm

Way too late to the party, but still my two cents: I am a biologist although not an expert on the topic. I feel that Dr Hilarious, BenK, Bill Benzon, and ne have given very good replies.

1. Group selection is invoked as an explanation for how altruistic behaviour can become fixed in a species despite harming any individual who exibits it. It is thus not really seperable from altruism.

2. Group selection is consequently also within-species, because, again, it is about how altristic behaviour can become fixed in a species. It is not about between-species competition.

3. To quote ne, “groups split into new groups much slower than organisms split into new organisms, so it’s hard to get the numbers to work out”. Or in other words, precisely because it is within-species, occasional gene flow between groups should far outweigh the slow effects of selection between groups. As far as I understand, most people reject group selection because it is mathematically virtually impossible…

4. … and because they think they have better explanations for the rise of altruism. Kin selection, costly signaling, the stability of “help everybody at first but stop helping and individual if they don’t reciprocate” strategies, etc.

To those who say the issue is merely semantic or what is wrong with the example chosen: Words have meanings. If somebody asks why physicists reject the possibility of a perpetuum mobile when that is just what they call their car, you would probably also consider them to be a bit off-topic.


BenK 02.02.16 at 11:46 pm

I’m going to throw in one more article in the possibility that it will be helpful:


Cooperation is encouraged by spatial structure (localism, anyone?)…


Adam 02.04.16 at 3:40 pm

@Alex 89

I’m not sure the difference between “group selection” and some of the examples you’ve listed in point 4.

Imagine that you have a population that organizes into hierarchical groupings. For example, individuals are part of colonies that in turn are part of tribes.

Selection pressures exist at every level of this hierarchy. On one time scale, individuals compete with other individuals. One another time scale, colonies compete with other colonies. One yet another time scale, tribes compete with each other. Suppose the consequence to “losing” at each level is the complete failure of individuals in the losing group to have offspring.

It seems like behaviors increasing the chance of “winning” at each of these levels would be favored. Whether a behavior that increased the chance of “winning” at the colony level but decreased the chance of winning at an individual level would be favored seems almost insurmountably fact-dependent. How much does the behavior actually decrease the individual fitness? How closely related are the individuals within the colonies, and how closely are the individuals in one colony related to the individuals in another? How often do colonies and tribes compete?

It certainly seems that for a range of parameters “group selection” would be favored. Whether that range of great extent in a mathematical sense is somewhat irrelevant: if the overall benefits of operating within that range of parameters is great then many individuals, groups, species, etc. will operate within it.

Another possibility appears to be when a collection of behaviors establishes a stable state with a far superior outcome to surrounding unstable states. Then it seems that if you run the system for a sufficiently long time, the population will end up evincing that collection of behaviors.

As another example, consider a solution of bacteria that require nutrients and air to replicate. The nutrients are available in a solution, which contains only some air, and supports a lower population density. At the surface of the solution there is an optimal amount of nutrients and air, supporting a far higher population density.

Suppose that the bacteria can individually produce stuff that collectively allows them to form a mat, and float on the surface of the solution. Obviously, once you have such a mat, selective pressures favor individuals that cheat and fail to produce the stuff. But if too many individuals cheat, then the colony sinks and they all die. So behaviors that punish such cheating are also favored – even if they reduce the fitness of the individual that punishes. The mat, considered as a whole, must have a certain proportion of individuals that punish cheaters, or else all the individuals in the mat will die. Random diversification will (eventually) give rise to a gene or set of genes that enforces such a distribution of behaviors. Elimination of the weakest will (eventually) ensure that the solution is covered in a mat of bacteria that individually sacrifice some fitness to ensure that they collectively don’t sink and die.

Is are these examples of group selection? Are they what you mentioned in as point 4?

I appreciate any insight you can provide…


dsquared 02.04.16 at 7:39 pm

Just wondered – how do things like slime molds, where the group is basically the only thing that has the sorts of properties that could be the object of selection, fit into this?

by the way

As far as I understand, most people reject group selection because it is mathematically virtually impossible…

“Mathematically virtually impossible” means “impossible within a model”, so the word “mathematically” is adding less here than you might think.


Alex SL 02.04.16 at 9:47 pm


As was pointed out above, groups split and die on a much, much longer time scale than individuals. So if GS is meant to explain the spread of altruism-coding genes, it won’t work, because the moment there is a bit of gene flow between two groups a selfish-coding gene will sweep through the group long before the group has replaced all other groups in the species. And this is also why your case of “then all individuals in the mat will die” is not an argument for GS but exactly the kind of outcome that must logically result if GS is supposed to be the way in which altruist genes spread across a species; instead it is the problem for GS right there!

The only solution is to demand that there is no gene flow between groups – in which case they aren’t groups in the same biological species any more but instead separate biological species. So not GS.

(By the way, all of this presupposes that at least one group ever had the altruism-gene fixed across all individuals so that it can even start competing with groups that haven’t. But that should be near-impossible in the first place given the underlying assumptions of the whole GS idea, i.e. that altruism-genes are disadvantageous in inter-individual competition and need GS as a mechanism to become fixed.)

As for why those things under #4 aren’t GS: Because GS is defined as differential survivial and selection at the group level; that is why it is called group selection. The things under #4 are about differential survival and selection at the individual (tit-for-tat) and gene level (kin).


zawy 02.05.16 at 3:36 pm

The Hadza hunter-gatherer groups, noe of the very few remaining, split up and rejoin every few weeks. The individuals select which group they want to be in, and you can leave anytime you want, if you find someone willing to take you in. You can imagine people fighting to be with people who have an altruistic and successful reputation, and since there is a limit, the altruistic are more likely to find each other and have more opportunity to succeed and breed.

I do not think the “problem” or “question” has been well formed. Is gene selection occurring at the genetic, individual, kin, group, species, or planetary level? The genes are only a memory of what chemical interactions exploit energy resources the fastest, emitting as much entropy as possible to the universe. Genes do not exert forces of their own accord, but exist as a result of what the environment forces. In the genes only appear selfish because of what the environment forces with potentials, I question if there are any other forces from the genetic all the way to the planetary level that can or should be viewed as either selfish or altruistic. This is assigning thought as a force as if it is independent of potential and kinetic energy sources which are the only cause of all forces. Thought is a physical process subject to dynamics like everything else, not something mystical. The environment is following the physical potential gradients that release heat and thereby entropy to the universe. The moon might be the pre-existing order that makes Earth different, giving rise to tidal flows where some of the earliest life is known to have lived, and giving rise to a constantly-churning mantel that gives rise to ore concentrations and veins that modern economies depend on. The odd tilt of the earth’s axis, also probably a result of the collision that created the moon gives rises to seasons and wind complexities which are another source of decreased entropy compared to Mars et al. So not only the Sun’s energy, but the moon’s lower-entropy may be giving rise to life (it’s mere existence is like a gas constantly having more molecules forced to one side of a box which is lower entropy).

This is why everything above appears vacuous. There is not any terminology being used in the discussion that I can grab on to and falsify or subject to observation. It’s like trying to argue about the old and discredited mind-body problem. The mind is a result the body. Selection is the result of the flow of energy, not a force in and of itself.


zawy 02.05.16 at 5:35 pm

Sorry about the above typos.

Genes, groups, altruism, and selfishness are results to be observed, not causes to which forces can be assigned.

The Sun and moon seem something like a father and mother: one gives the energy and the other replenishes the order, disrupting the path towards a stability that would have the highest disorder. If the moon ceased to exist and the axis was straight, things might more quickly devolve to a more stable and boring state.

Our economic machine is on a path towards more machines and less biology. Wikipedia says the species extinction rate is 1,000 to 10,000 times the background rate. Muscle, brains, and photosynthesis are being rapidly replaced in their relevance to the economic machine due to oxygen being removed from metal and metalloid ores, creating far stronger bonds which means lower entropy. The selection of which bonds are the strongest is limited compared to the possibilities, so they appear as copies and pseudo-copies. DNA crystals are pretty strong. Copies are lower entropy, but copies may be a side effect of the increasing density of the mass on Earth (stronger bonds result in a lower volume of the mass constituents, which also means lower entropy … see the specific entropy of the elements involved, or look up how the entropy of solids is calculated).

Since the use of rocks, people who utilized the stronger bonds were “economically” more viable than those who did not. We normally view this as simply stronger shells and teeth. Now there is steel, concrete, carbon fiber, and nanotubes available for structures. Refined metals and silicon are replacing the ores that had oxygen attached and they are far and away better at not only structural strength, but in moving matter (electrical motors are about 200x more economically efficient than muscle), thinking about how to move matter more efficiently (CPU’s based on silicon need to move and compute with electrons whereas brains depend on ions and molecules weight 50,000 to 1,000,000 times more), and acquiring the energy to do it (silicon solar cells are 20x more efficient per surface area than photosynthesis).

So maybe the moon is keeping things unstable so that matter on Earth can become more dense. The Earth is an open thermodynamic system, emitting more entropy via 17 photons of low energy and in random directions for each photon from the Sun’s direction. Isolated systems having increasing entropy, but open systems like the Earth have been known since the 1960’s to be capable of increasing order (lower and lower entropy). If this is the case, what is the limit to the order that can be extracted from the moon’s disturbances? It may be that the 75% missing dark matter in the universe is the result of Dyson Sphere’s of “life” capturing all the light from enclosed Stars, turning the excess energy into local matter. The Universe on an expanding basis BTW is not increasing in entropy. This has been known even before the COBE observations. See the highly respected 1970’s book “The first 3 minutes” by Weinberg. This means on a local volume basis, entropy must be decreasing. Gravitational systems (mass concentrations) seem to be emitting the entropy needed for space to expand. Isolated systems with increasing entropy seem to be merely an engineering ideal, not to be used on a cosmological basis. I mention this because there is often the argument that life could not decrease entropy because it can’t possibly be violating the “entropy always increases” law. Entropy is always emitted, not exactly “increasing” on an Earth-wide basis, and definitely not (by theory and observation) on a Universal basis. You can measure it here on Earth by following the commodities and it seems to me to be decreasing as the excess entropy is emitted.

Comments on this entry are closed.