Evolution for Everyone

Dear Richard,

I do not agree with the cynical adage “science progresses–funeral by funeral”, but I fear that it might be true in your case for the subject of group selection.

In your response to the recent article in Nature by Martin Nowak, Corina Tarnita, and E.O. Wilson, which critiques inclusive fitness theory, you say this about group selection:

Edward Wilson was misunderstanding kin selection as far back as Sociobiology, where he treated it as a subset of group selection … Kin selection is not a subset of group selection, it is a logical consequence of gene selection. And gene selection is (everything that Nowak et al ought to mean by) ‘standard natural selection’ theory: has been ever since the neo-Darwinian synthesis of the 1930s.

I do not agree with the Nowak et al. article in every respect and will articulate some of my disagreements in subsequent posts. For the moment, I want to stress how alone you are in your statement about group selection. Your view is essentially pre-1975, a date that is notable not only for the publication of Sociobiology but also a paper by W.D. Hamilton, one of your heroes, who correctly saw the relationship between kin selection and group selection thanks to the work of George Price. Ever since, knowledgeable theoretical biologists have known that inclusive fitness theory includes the logic of multilevel selection, which means that altruism is selectively disadvantageous within kin groups and evolves only by virtue of groups with more altruists contributing more to the gene pool than groups with fewer altruists. The significance of relatedness is that it clusters the genes coding for altruistic and selfish behaviors into different groups.

Even the contemporary theoretical biologists most critical of multilevel selection, such as Stuart West and Andy Gardner, acknowledge what you still deny. In an earlier feature on group selection published in Nature, Andy Gardner is quoted as saying “Everyone agrees that group selection occurs”–everyone except you, that is.

You demonstrate more ignorance about group selection when you contrast it with gene selection. You correctly say that gene selection is standard natural selection theory. Essentially, it is a popularization of the concept of average effects in population genetics theory, which averages the fitness of alternative genes across all contexts to calculate what evolves in the total population. For that reason, it is an elementary mistake to regard gene selection as an alternative to group selection. Whenever a gene evolves in the total population on the strength of group selection, despite being selectively disadvantageous within groups, it has the highest average effect compared to the genes that it replaced. Please consult the installment of my “Truth and Reconciliation for Group Selection” series titled “Naïve Gene Selectionism” for a refresher course. While you’re at it, check out the installment titled “Dawkins Protests–Too Much“.

The Nowak et al. article includes several critiques of inclusive fitness theory that need to be distinguished from each other. One issue is whether inclusive fitness theory is truly equivalent to explicit models of evolution in multi-group populations, or whether it makes so many simplifying assumptions that it restricts itself to a small region of the parameter space. A second issue is whether benefiting collateral kin is required for the evolution of eusociality and other forms of prosociality. A third issue is whether inclusive fitness theory, as understood by the average evolutionary biologist and the general public, bears any resemblance to inclusive fitness theory as understood by the cognoscenti.

In an e-mail discussion among theorists on both sides of the debate that I organized several years ago, I conducted a citation analysis comparing Hamilton’s original 1964 papers on group selection with his 1975 reformulation in the light of the Price equation. These papers are cited in a ratio of approximately 15:1 with no tendency for citations of the 1975 paper to increase in frequency over the decades. In other words, most evolutionary biologists still have a pre-1975 conception of inclusive fitness theory, no matter how much it has been elaborated by the cognoscenti. This degree of illiteracy about foundational issues is an embarrassment for the field of evolutionary biology. The Nowak article is a wake-up call for the average evolutionary biologist and the general public to reconsider the conventional wisdom about inclusive fitness theory, in addition to the debates that will take place among the cognoscenti.

Richard, if you wish to join the debates among the cognoscenti, you will need to abandon the priestly way that you make pronouncements, expecting what you say to be taken on faith, and rejoin the ranks of scientists who hold each other accountable for what they say.


  1. #1 Julian Garcia
    September 4, 2010

    My take on the fierce reactions, not just by Dawkins, but basically by most of the people on the kin selection camp (notable exception Alan Grafen), is that they fail to criticize the actual assumptions of the model in the paper. They should take on the challenge and do the maths to counter the arguments presented by Nowak, Tarnita and WIlson. That would be more insightful than just embarking on verbal theorizing that is in the end bound to be shallow and susceptible to end up diluted in semantic unsubstantial issues.

    If the model is wrong, they should come up with reasons why, for instance, the assumptions do not hold in the real world. If they claim that the worker-centered approach is a better one, or an equivalent one, they should do so by coming up with a model with similar assumptions in which the traditional approach yields equivalent, or even better (simpler?) predictions. None of the critiques has even dare to delineate such an approach.

    The debate seems to be completely ignoring the paper, and rather bringing an old discussion without a common ground. For instance, part A in the supplemental information has been completely dismissed in the reactions. That is a key issue, because it shows (formally!) the shortcomings of inclusive fitness, which, granted, is not equivalent to kin selection, but is definitely the tool of choice to do the accounting when thinking of kinship as the leading force of selection.

    All in all, I’d like to see a healthy formalization of the debate. The key issue here is a mathematical model, and that should be what the debate is about. All this reactions dismissing the actual model are sad, and do not help science.

  2. #2 Charles Hixson
    September 4, 2010

    My take is that theoretically the gene selection theory is correct, but that the math is too horrendous to do, and probably involves some knowledge we don’t have with sufficient precision. (E.g., how to you model epigenetic modifications?)

    I feel that the group selection approach is theoretically more manageable, and also gives answers in good (though not perfect) agreement with what a gene oriented theory would provide. For an analogy, consider Newtonian and Einsteinian physics. Newtonian is much easier to use, and is good enough for most purposes. Einsteinian provides slightly more accurate results, and occasionally the difference is significant. The difference is that we don’t really have a full mathematical workup of genetic evolution, and probably can’t at the current stage of our knowledge. But it’s still the “right” theory. Just not a practical one.

  3. #3 JP
    September 5, 2010

    Julian (or others in the know), what is Alan saying about the matter? I’ve not seen comments from him yet.

  4. #4 Emil Karlsson
    September 5, 2010

    I am not aware of any group selectionist who has successfully dismantled the three main arguments against group selection:

    – “For the good of the group” behavior is often not an ESS and will not be selected for.
    – Groups are not sufficiently isolated.
    – Differential reproduction and extinction of group occurs too slow for group selection to be a significant mechanism.

    These criticisms are made routinely in textbooks on behavioral ecology (eg. Krebs & Davies) for decades. Or is the biological establishment putting “lies in the textbooks” as Hovind so frequently claimed?

    Also, Dr. Wilson is making the appeal to novelty fallacy.

  5. #5 Doazic
    September 5, 2010

    This was one of the big surprises after taking a Evo course having read Dawkins. There are experiments that show group selection works, it’s accepted, blah blah blah. It doesnt seem fair for him to promote “Evolution” while actually only promoting his view of it and not the consensus.

  6. #6 Julian Garcia
    September 5, 2010

    JP, I am not aware of any response by Grafen. I take it as a good sign that he has not rushed to say something. I expect his reply to be formal and to the point. A workshop is being organized in Amsterdam, where he and other mathematical biologists from both sides are invited. I am sure this paper will be talked about a lot in that meeting.

  7. #7 OTE
    September 5, 2010

    Emil, in response to your post

    I suggest you read up a bit more on the literature as these 3 critiques of GS have been thoroughly dismantled long ago, or you could just read David’s Truth and Reconciliation blogs which highlight the history of the literature. Please do not take offense but that statement is typical of misconceptions about GS that perpetuate the debate.

    for example.

    1. “For the good of the group” behavior is often not an ESS and will not be selected for.

    This statement makes no sense as of course it is selected for, how else do you explain altruism in the first place? Please note that reciprocal altruism, kin altruism and other strategies are indeed “for the good of the group”, b/c in either case, the altruistic strategy does not locally outcompete his selfish partner. It is by virtue of GS (differential fitness of homogeneous pairs)that altruism evolves and becomes an ESS.

    2. – Groups are not sufficiently isolated.

    This was never a condition for GS. This was naively attributed to GS. What IS required is variation amongst groups. This can easily be achieved when considering assortment or the conditional movement of individuals which is ubiquitous and occurs in organisms as simple as bacteria. For example, if you don’t like your partner, walk away and find another. This will increase variation amongst groups and strengthen GS. There has been theoretical (see Athena Akitpis as well as works by John Pepper) and empirical work (See Omar Eldakar) to illustrate these points. Not to mention this was thoroughly discussed in previous Truth and Reconciliation blogs.

    3. – Differential reproduction and extinction of group occurs too slow for group selection to be a significant mechanism.”

    Groups do not have to go extinct for GS to occur. All that is relevant is that group membership influences the fitness of individuals beyond what is predicted by individual phenotype alone.

    I simply make these points to dismantle the misconceptions of GS so we can get out of this haze and evaluate the theory accordingly.

  8. #8 Tyler DiPietro
    September 5, 2010

    I’m wondering how one even demarcates between the effects of kin-selection and the effects of group selection. Strict kin-selection, as far as I know, is predicated on close genetic relatedness. Group selection, on the other hand, would tend to promote traits within groups and lead to greater genetic relatedness. The difference between them seems to be in the direction of the causal arrow, but it would seem to me to be difficult to identify a prediction unique to either one.

  9. #9 Ben Allen
    September 7, 2010

    (Full disclosure: I am a postdoc in Martin Nowak’s lab. However, the views I present are my own.)

    @Tyler- That’s kind of the point. There are no predictions that are exclusive to inclusive fitness theory. Inclusive fitness theory is not a model. It is a mathematical formalism in which the effects of frequency dependence are re-organized to show benefits given to kin. This formalism can help analyze certain models, but it makes no predictions of its own.

    Moreover, this re-organization of terms only works under certain assumptions, such as weak selection and additive fitness effects.

    Group selection, on the other hand, can be modeled. So can spatial selection, tag-based altruism, kin discrimination, and reciprocal altruism. These models make specific predictions that can be tested. My view is that the altruism we see in nature is due to a mixture of these (and possibly other) mechanisms.

  10. #10 Anthony Cole
    September 20, 2010

    I’m listening to your conversation with Philip Adams tonight on Australian radio (abc.net.au/rn/latenightlive/stories/2010/3016959.htm), and I was interested to hear you say that the mechanism that allows altruism to flourish within a group is marginalisation of non-altruistic people.

    You are probably aware of this but, just in case, Roy Baumeister and Nathan DeWall seem to have found that rejected persons suffer a drop in self regulation and control of attention. This, in my view, should make the marginalised less able to repair social breaches; so, not only do altruistic people marginalise non-altruistic people, but the marginalisation is hardened by processes within the marginalised brain.

  11. #11 Graves
    September 24, 2010

    Sorry, I am very late to this conversation.

    #7 says:
    1. “For the good of the group” behavior is often not an ESS and will not be selected for.

    This statement makes no sense as of course it is selected for, how else do you explain altruism in the first place?

    This is agonizingly wrong. “For the good of the group” behavior is NOT selected for–that is at the heart of this issue. There are NOT instances where individual entities sacrifice “for the good of the group” where those entities don’t indirectly benefit (e.g. via benefits to kin). If selection favored behaviors acting “for the good of the group” a la Wynne-Edwards, we would not see so much conflict and cheating just below the surface. Selection is not favoring “the good of the group” but something more akin (pardon the pun) to “the good of the group, so long as the costs aren’t too great to myself and the group I am helping is sufficiently related.” It is easy to see that the results of selection favoring the former would be different than the latter. What do we see in nature? Adaptations reflecting a long history of the latter.

  12. #12 John Kwok
    September 28, 2010

    Dawkins is speaking tomorrow night at Cooper Union in Manhattan (New York, NY) at approximately 6:30 PM in the Great Hall of the original building (where Abraham Lincoln and Barack Obama spoke). Wonder if anyone would raise this very point with him.

  13. #13 Emil Karlsson
    September 28, 2010

    OTE, your supposed “dismantle of the misconceptions” leaves the arguments against group selection largely unscathed.

    1. “For the good of the group” behavior is often not an ESS and will not be selected for.

    This can be illustrated with an example found in mamy introductory texts on behavioral ecology: male lions who have recently taken over a group will kill the previous infants to get the female into ovulation faster. This behavior is clearly not for the good of the group (infants die), yet it is an ESS.

    Any “for the good of the group” behavior will always be exploited by “for the good of me” individuals provided there is no “for the good of the gene” behavior.

    Kin selection evolved because it was good for the gene, not for the group. In fact, kin selection can be bad for the group since it can foster outgroup hostility.

    2. Groups are not sufficiently isolated.

    If groups are not sufficiently isolated, there is nothing that stops “selfish” individuals from entering an exploiting others who are supposedly practicing “don’t exploit resources to much by giving birth to more offspring”.

    This is the analogous criteria for standard Darwinian selection. In general, if populations are not sufficiently isolated, no speciation will occur.

    Bacteria in close vicinity as a result of asexual replication are usually very closely related, so those results may be explained with standard gene selectionism.

    3. Differential reproduction and extinction of group occurs too slow for group selection to be a significant mechanism.

    Are you really trying to say that evolution does not require differential reproduction? Is there even a testable mechanism for the reproduction of groups? What is the material basis for heredity for “groups”?

  14. #14 Neil Jordan
    September 28, 2010

    I’m trying to understand why this discussion is so fixated on kin selection and reciprocal altruism, and bereft of any mention of competitive altruism and particularly the handicap principle. Now, I may not be a PhD and as such may be lay compared to some of the minds here, so I would appreciate an explanation about that. To me it seems that the handicap principle amply explains the propagation of supposedly fitness-reducing properties. Sex selection always has a disproportionate effect over the other tangibles, as is demonstrated by the diversity of attraction mechanisms, and the only way to establish a reliable “fitness” signal is to behaviorally reduce it.

    #7 – this also answers your incredulous rhetorical question, “how else do you explain altruism in the first place?”. A behavior doesn’t have to be for the good of the group, though that certainly helps. It just needs to be harmful to oneself in order to be a strong signal.

    If you think I’m missing something, go ahead and correct me.

  15. #15 CherryBomb
    September 28, 2010

    Of course, the underlying assumption is that altruism is adaptive in SOME way, we just have to figure out how to model it mathematically. If it is not adaptive, but is simply the consequence of other adaptations, the whole thing collapses.

  16. #16 Gus
    September 29, 2010

    No wonder that Professor Dawkins hasn’t bothered to reply to this pathetic taunt. I’m pretty sure he has more important things to do than debate with a pseudo-pedantic Templeton Foundation accomodationist whose greatest “scientific” achievement has been hijacking an old and forgotten bit of evolutionary theory and repackaging it as a revolutionary insight. Oh yeah, “new” group selection. It comes in various flavours: faith, ignorance and stupidity. No wonder only equally frustrated and pretentious scientific wanna-be’s like PhDx3 Massimo Pigliucci pay attention to this kind of intellectual masturbation. Some people devote (and risk) their lives to promoting knowledge of evolution, while duplicitous free-riders obscure it with scholastic hair-splitting. Repulsive.

  17. #17 Perplexed in Peoria
    October 1, 2010

    I think that D. S. Wilson would be far more successful in convincing the skeptics that group selection works if he would stop insisting on calling this mechanism “group selection”. The difficulty is that people think they already know what a “group” is: It consists of many individuals, it is (mostly) reproductively closed, and it persists for a much longer time than an individual’s lifetime. They have in mind a notion of a human tribe or village. And with this picture of a “group” in their mind, they are totally correct in saying that “group selection” just doesn’t work.

    But Dr. Wilson perversely insists on using the word “group” for collections that persist for periods of time shorter than an individual’s lifetime, collections that emphatically are not reproductively closed, and collections that contain only a handful of individuals (perhaps as few as two). He points out that group selection does work, but fails to make clear that it only works for the kinds of temporary groups that he has in mind. And so the miscommunication continues.

    Truth and reconciliation requires some give on both sides. If Dr. Wilson would either stop calling this stuff “group selection” or just admit that his group-selection models work best with short-lived, reproductively open “groups” such as hunting parties and pair-bonded mating pairs that cooperate for a single mating season, he might find that he gets a more willing reception of these ideas.

  18. #18 daedalus2u
    October 2, 2010

    Greg Laden had a post on a carrion beetle that carries mites around and the mites protect its offspring from competition.


    This is a group (carrion beetle plus mites) that is not reproductively closed.

    It out competes other groups (carrion beetles without mites), (mites without carrion beetles).

    No group needs to go extinct for the group of carrion beetles plus mites to evolve and become more common. The groups of isolated carrion beetles and isolated mites can be sustained by recruitment of individuals from groups of carrion beetles plus mites. If the group carrion beetles plus mites exhibits 10% greater reproductive success per generation than isolated carrion beetles and isolated mites, then the group is fabulously successful.

    Is it “altruism” that causes the mites to spare the offspring of carrion beetles? Is it “altruism” that causes the carrion beetles to carry the mites around? It certainly isn’t kin selection because the carrion beetles and mites are not related, they are not kin of each other.

  19. #19 bob koepp
    October 3, 2010

    daedalus2u –
    Are there genetic factors that correlate with this arrangement, on either the beetles’ or the mites’ part? If not, there’s quite a bit more that needs to be said to portray this as a case where “out competing” is an istance of natural selection in action. I’m willing to consider alternatives to genetic heritability, but they need to be specified and shown to be “compatible” with the structure of selective processes.

  20. #20 daedalus2u
    October 4, 2010

    I am not sure what your objection is. The behavior and life cycle of these mites is well known.


    Are you suggesting that the behavior of the mites and beetles is not genetic but is somehow taught or learned?

  21. #21 Carmi Turchick
    November 14, 2010

    Emil, you are quite clearly confused about what a group is, and this is a large part of your issue with GS.

    ” “For the good of the group” behavior is often not an ESS and will not be selected for.

    This can be illustrated with an example found in mamy introductory texts on behavioral ecology: male lions who have recently taken over a group will kill the previous infants to get the female into ovulation faster. This behavior is clearly not for the good of the group (infants die), yet it is an ESS.”

    Male lions are not group selected, they compete for mating opportunity at the individual level or as two or three siblings. Killing nursing lions to pass on more of their own genes is simple individual level selection and has nothing whatever to do with GS questions.

    “2. Groups are not sufficiently isolated.

    If groups are not sufficiently isolated, there is nothing that stops “selfish” individuals from entering an exploiting others who are supposedly practicing “don’t exploit resources to much by giving birth to more offspring”.”

    Here you show that you still think of GS as being what it was back in the 1950’s, conflating group with species level selection. A “group” is not a species. Evolution cannot select for organisms that avoid overexploitation of resources, and GS does not postulate otherwise.

    It is very simple. Evolutionary Theory asserts that resources are finite and unequally distributed, that populations will always grow to exceed available resources, and that the resulting competition for fitness limiting resources will result in evolutionary selection. Put simply, competition for resources = evolutionary selection. If this is true, then it is always true. One cannot claim that only some competition for resources will result in selection (over time of course) without asserting that one can disprove the Theory of Evolution.

    Now, some species, such as humans, compete for resources at both the individual level and the group level. Most here reading these posts can readily obtain far more resources than they most likely would be able to if they were born into the group called Zambia, instead of the group called America or Britain or whatever other wealthy nation that they were born in.

    Again, if competition for resources = evolutionary selection, then if competition for resources happens at the group level, at all and for any length of time, we must either call this group level selection or disprove the most central assertion of the Theory of Evolution. The view that opposes GS is therefore not consilient with the Theory of Evolution in the most basic way possible.

    Now, looking at altruism would be a very long post but let me just briefly say, as I have many times many places, that the thing about selfish is that they are selfish. And the thing about altruists is that they need not be naive, entirely inflexible in their altruism, or incapable of retaliation. A group of altruists is a fitness limiting resource for a selfish individual. Selfish should and do therefore attempt to exclude other selfish. The more selfish in the group, the lower the benefits for the selfish. Especially given group level selection, fitness of the exploiting selfish goes down rapidly if the group has too many other selfish and becomes likely to lose wars and be unable to defend its resources, ie, its territory. So we must expect the strategy of selfish to evolve to attempt to exclude other selfish, not overrun and replace the group of altruits. Selfish are selfish, they do not want to share. These efforts against other selfish result in a fitness cost for selfish, reducing their fitness advantage over altruists. Their need to avoid detection further limts their benefits and adds to their costs, because groups of altruists are highly unlikely to stop at excluding selfish from exchanges in the real world.

  22. #22 Edd
    December 2, 2010

    The model itself is beyond my math to deal with, but its assumptions do stand out as surprising. One such puzzling assumption is that workers are just “robots”? How does the model explain conflict between workers and reproductives then (of which there is a good body of empirical evidence)? What, in fact, does the model actually predict at all?

  23. #23 martin
    March 29, 2011

    I believe the only significant selection to talk about is gene selection. Gene selection automatically translates to kin selection, and altruism on the level of individuals is just a consequence of the selfishness of the genes. Both individual selection and group selection are empty words. Sometimes one may see “the good for the individual” or “the good of the group” prevail, but that is only because “the good for the genes” happens to coincide with the good of the genes. This, as far as I understand, is the so-called neo-Darwinian vew.

    So far, nothing original. HOWEVER: there is an issue of complexity. Most models of strategic competition considers a relatively small number of strategies. In such cases, one sees a rapid convergence to an ESS. In practice, however, I would expect there to be a huge number of competing strategies around. Being a dynamicist (=someone who studies chaotic dynamical systems), I would expect to see strange attractors rather than simple fixed points. That is, instead of having a fixed porcentage of individuals using each strategy (like an ESS), there is a constant and seemingly erratic interchange between strategies. Minor difference in circumstances will dramatically change the course of the evolution of these trajectories within certain boundaries (think of the butterfly effect). CONSEQUENTLY, if a species is somehow divided in two groups, they are expected to follow different trajectories (in the simplex of strategy densities). At some subsequent time, one group will “by chance” have a mixture of strategies (perhaps more cooperative) that “happens” to be favorable for the group. This group will then prevail, and it will look like a group selection has taken place, while in fact it is caused by a lack of convergence to an ESS.

  24. Their need to avoid detection further limts their benefits and adds to their costs, because groups of altruists are highly unlikely to stop at excluding selfish from exchanges in the real world.

  25. #25 Tim Tyler
    September 9, 2012

    Perplexed, I don’t see how it is “perverse” to use the word “group” for collections that persist for periods of time shorter than an individual’s lifetime. That is common usage of the word “group”. For instance you can say “see that group of thugs om the street corner”. The term “social group” is standard terminology in the social sciences, and it has no implications of reproductive isolation. The dictionary apparently offers no support for your conception of the term “group”. The idea that “groups” are necessarily reproductively isolated appears to stem from the 1960s attack on group selection – not necessarily a good source of information on the topic.

New comments have been disabled.