Evolution for Everyone


The newest issue of Science Magazine includes a lovely demonstration of multilevel selection by Omar Tonsi Eldakar, my former graduate student, who is currently at the University of Arizona’s Center for Insect Science.

Readers who have been following my “Truth and Reconciliation for Group Selection” series will be well prepared to appreciate the import of the Science article. Group selection requires variation among groups. Variation among groups is eroded by dispersal. Therefore, group selection can only take place in groups that are highly isolated from each other. That is part of the reasoning the led to the conclusion that group selection can only take place under highly restrictive conditions.

But wait. This argument assumes that dispersal is random. What if dispersal is conditional? What if individuals stay in groups when they are sufficiently cooperative but leave when they become overrun by selfish individuals? In this case, dispersal might increase variation among groups, improving the conditions for group selection. John Pepper and Athena Aktipis (featured in T&R XII) are two theorists who have studied this “walk away” process in agent-based simulation models.

Omar and his co-authors have provided a lovely empirical demonstration of the walk-away process for sexual conflict in water striders. Male striders differ greatly in the aggressiveness of their mating tactics. Some would qualify as psychopathic sexual predators by human standards, while others are perfect gentlemen. In previous research, Omar and his colleagues created groups in which the composition of males varied from all gentlemen to all psychopaths. Females are far more productive in the company of gentlemen, as one might expect, but the psychopaths have the highest mating success within any group containing both types. Thus, the gentlemen count as altruistic and the psychopaths count as selfish in a classic group selection scenario–what I call “the original problem” in the T&R series.

Variation among groups was artificially created in the previous studies but was allowed to emerge spontaneously in the study reported in Science. i-9e71d1960dfd5229968c6fe7cd6e3311-figure1.jpg

Six compartments could be closed off or opened to allow dispersal among compartments. In the closed condition, there was a strong positive relationship between male aggressiveness and mating success, as shown by the top graph. In the open condition, females dispersed in response to the presence of psychopathic males. The psychopaths could also move, but the end result was a considerable amount of variation among groups, as predicted by the walk away models. Thanks to the clustering of the females around the gentlemen, the relationship between male aggressiveness and mating success becomes hump-shaped with the peak at a low level of aggressiveness, as shown in the bottom graph. The fact that the same individual striders were used in both conditions makes this a particularly good demonstration of the effect of population structure on the fitness consequences of i-b7ebd0dd2b8b9c9aac0bbffb25eeffac-figure2.jpg

social behavior.

Group selection explains how locally disadvantageous traits can nevertheless evolve in the total population by virtue of increasing the productivity of their group. Knowing what evolves in the total population does not provide sufficient information to evaluate whether group selection needs to be invoked. It is the comparison between local fitness differentials and global fitness differentials that provides sufficient information. This study is lovely precisely because it provides the comparison at both the local and global scales and because the variation among groups emerged from the movement of the animals themselves, rather than being imposed by the experimenter.


  1. #1 InfuriatedSciTeacher
    November 6, 2009

    The figures don’t come through, FYI. That may be a result of the network here at work rather than anything that happens elsewhere.
    I was all ready to ask how the laboratory set-up actually mirrored the natural environment for water striders, but you seem to have addressed that disconnect in the last paragraph of the paper. How dissimilar are the dispersal patterns, or haven’t they been examined?

  2. #2 noel
    November 6, 2009

    ISTeach: I am not getting the figs. either, but I am also a teacher on a network. Regarding the question about the natural environment, the situation only has to be a plausible approximation to work as a model. And it doesn’t even have to be a plausibly realistic situation for water striders to show that group selection can occur in general.

    D.S.Wilson: Someone complained about the length of this series? I would like to say that I am enjoying learning about your point of view. I was told by a philosophy instructor a long time ago that group selection was shown to be unworkable by “powerful arguments” of Dawkins. I didn’t exactly take his word for it, but I didn’t look into the issue myself until now. From a philosophy of science perspective, this is really interesting.
    piker and the other’s who disagree with the posts: be more specific in your rebuttals. More substance, less animosity. Sounding insulting without having convincing arguments – or producing arguments that obtusely miss the point – actually illustrates Wilson’s point.

  3. #3 piker
    November 6, 2009

    Seems I’ve been specific enough to get more significant clarifications on terms and theory from Dr. Wilson by my methods than you have by yours. But of course you don’t seem to need any.

    “the situation only has to be a plausible approximation to work as a model. And it doesn’t even have to be a plausibly realistic situation for water striders to show that group selection can occur in general.”

    If that is your take from your philosophy of science perspective, you may need to re-examine the concept of plausibility as you’ve come to understand it.

    To be more specific, that statement doesn’t meet the test of the self-evident truth, because since it doesn’t work on any other level of logic, that would have been its only recourse.

  4. #4 oez
    November 6, 2009

    As far as approximating the natural environment of striders, it sure does! Not only that but sexual conflict studies confining individuals to isolated groups is definitely not close to the natural environment. When individuals are confined to pools, aggression is favored, when they can freely assort this relationship disappears, which resembles the relationship observed between aggression and mating in the field. Nice work.

  5. #5 piker
    November 6, 2009

    What if the use of strategies is conditioned by group dispersal and in turn dispersal is dependent on the conditional nature of the inherent strategies? And if so, can you accurately refer to this relationship as representative of “how locally disadvantageous traits can nevertheless evolve in the total population,” if in fact this perceived evolution is simply one of conditional change, while leaving the inherent elements of the strategic algorithms intact?

  6. #6 Guy
    November 6, 2009

    David: Just FYI from my end the figures did come through.

  7. #7 daedalus2u
    November 6, 2009

    piker, that is a very interesting observation, which parallels a thought I had yesterday in another thread to explain the threatening and abusive behavior of a troll. (comment #30 by me).


    Which comment was before this post by DSW.

  8. #8 oez
    November 6, 2009

    Piker, I’m not sure I quite get what your stating. I hope this gets to the point. These strategies were not conditional but stable as evident in the methods of the paper. Individuals did not alter their strategy over time or across treatment. The paper preceeding this one has more details and may be a more useful comparison. http://www.springerlink.com/content/p41033688v8233x2/

  9. #9 noel
    November 6, 2009

    Uh, piker, all I meant was that an organism under controlled conditions can be a model for the behavior of other organisms (“in general”). But of course it’s better if the experiment doesn’t have to be interpreted that way, as oez asserts.

  10. #10 piker
    November 6, 2009

    The inherent nature of a strategy would involve the consistency of an organism’s initial response to any problem where, as in these examples, the choices are between aggressive and less aggressive approaches. The conditional aspect is where an aggressive individual has the option of tempering that aggression when it appears to be counterproductive – i.e., when the expectations are that its normally aggressive approach will fail.

    We can see where these striders clearly have that option, and that may be the most important thing to be learned here – the range of such options as they apply to each species. Here in particular we discover an insect with options that are not common to insects as a general rule – or so I at least had been led at one time to believe.

    And what was meant by the observation that “locally disadvantageous traits can nevertheless evolve in the total population” if. as oez interprets this, “Individuals did not alter their strategy over time or across treatment.” Clearly here WAS an alteration of strategy according to the change in the group dynamic. My quarrel with this is that the traits cannot be said to have evolved if in fact the changes were simply conditional and the heritable aspects of the strider traits remained the same regardless.

    (I refer to algorithms as I also believe such heritable traits are in effect instinctive algorithms, all coming with some conditional strategies, and those strategies in turn with a variety of optional tactics according to the species, and after that according to phenotypic potential within the species.)

  11. #11 bob koepp
    November 6, 2009

    Correct me if I’m wrong, but I think the idea that “locally disadvantageous traits can nevertheless evolve in the total population” does not imply anything about individuals altering their strategies/behaviors (i.e., traits). What alters is the relative frequency of individuals displaying the traits in question.

  12. #12 piker
    November 6, 2009

    So that’s what was meant by traits nevertheless evolve! Like aggressive individuals evolving their traits by joining motorcycle gangs! Nobody ever ‘splained it like that before.

  13. #13 oez
    November 6, 2009

    If individuals could optimize their strategy, then in the isolated group condition, all males would be extremely aggressive as that is what the selection pressure promotes. this was not the case. When dispersal was allowed, individuals maintained their same level of aggression as they did when in the isolated condition. The only thing that changed was that the advantages to being ‘nice’ finally arose. Even if, and this is a big if, there was a conditional strategy, then this does not change the fact that being less aggressive is locally maladaptive. If an individual can vacillate between ‘altruistic’ and ‘selfish’ strategies it does not change the main point that the advantage of altruism is at the group-level. Unless there is some way that an altruist can beat a selfish individual in a heads up competition. It still remains that a reciprocal altruist cannot defeat a selfish non-reciprocator in a one-on-one, a tit-for-tat loses to selfish individuals in one-on-one and even a cooperative sibling loses to a selfish sibling within a kin group (Even Hamilton stated that inclusive fitness is only advantageous at the group level when he revised the theory). All of these theories require group level interactions to see the benefit, this is not individual level. The advantages of altruism and less aggressive mating males simply do not exist locally and therefore cannot be explained locally.

  14. #14 Stuart
    November 6, 2009

    The advantage of being ‘nice’ isn’t huge, really nice males are not particularly successful in either treatment. In treatment 2, females are given some choice in who they mate with, disadvantaging hyper aggressive males that reduce the females own fitness. In treatment 2 males in the middle of the bell curve of aggressiveness are the most successful (females chose intermediately aggressive males, not nice males), with males on both ends having reduced success… The existence of groups and movement between groups is vital in giving females this choice, but given that hyper aggressive males disadvantage female fitness apparently, giving them a chance to escape them allows them to optimize their own fitness by choosing…. Individual selection is undoubtedly important in creating the pattern we see (females optimizing their own fitness)… There is a multi level element, I’m just not sure how important it is in the evolution of the pattern we see in this particular example…

    p.s i don’t have access to the science article at the moment, so may be misinterpreting things.
    p.p.s enjoying the series…

  15. #15 piker
    November 6, 2009

    oez writes:
    “If individuals could optimize their strategy, then in the isolated group condition, all males would be extremely aggressive as that is what the selection pressure promotes. this was not the case.”
    The premises of that contention are simply not true, since biological strategies are designed to be effective in the environment where the organism is expected to dwell. If they are fixed, it’s virtually always to a flexible degree, or strategies per se would have no survival potential – no potential for adaptation (I’m trying not to use the word fitness here – the subject is confusing enough as it is). All individuals in a tight group choosing the same competitive strategy as optimal would be a disastrous choice all round. And aggression IS a competitive strategy. What’s missing in a lot of these models is the extent to which the groups evolve rather than, or in addition to, the individuals from aggressive competition as the norm to non-aggressive cooperation.
    (And in my view, the selfish/altruistic dichotomy comes from a misconception of what is most likely a competitive/cooperative one – aggression being heritable, selfishness as cause rather than consequence likely not being.)

  16. #16 oez
    November 7, 2009

    piker, please explain the local benefit to less aggression? How does a non-aggressive outcompete a high aggressive in a local interaction?

  17. #17 David Sloan Wilson
    November 7, 2009

    With respect to phenotypic plasticity, in part it’s an empirical question. Omar and Mike Dlugos (his collaborator) measured aggressiveness in the same individuals repeatedly over time and also under different conditions, such as food level, day length, and presence and absence of fish predators. Individual differences were remarkably stable except in the presence of predators, when everyone quieted down. It didn’t have to turn out that way–they could have been more behaviorally flexible, but they weren’t. The most aggressive striders did not change their ways in the open condition, even though their mating success had gone way down.

    When we examine different types of phenotypic plasticity in theoretical models, we find that the basic logic of multilevel selection remains (e.g., the tit-for-tat strategy, which still never beats it social partner), as I and others have stated in previous comments. Remember that behavioral flexibility was part of the exercise in Athena Aktipis’s class exercise. Whether a fixed strategy or a part of a flexible strategy, selfishness has the local advantage.

  18. #18 Omar Tonsi Eldakar
    November 7, 2009

    Just reading over the comments it seems that some are caught on minor issues and losing sight of where the benefits of aggression and non-aggression are. As David notes, although flexibility was not part of our study, strategies like TFT are at a within group disadvantage when paired with a selfish individual but at a between group advantage when paired with another TFT (compared to a selfish-selfish pair). In the end, the fate of TFT lays with the net effect of these interactions. Same applies to less or nonaggressive males and conversely for hyperaggressive males. The quadratic relationship is the result of the within and between group forces facing the continuum of male strategies.

  19. #19 piker
    November 7, 2009

    “Except in the case of predators, when everyone quieted down.”
    That’s the key. Aggression was tempered by the condition. The striders had a conditional strategy available. in other words. They also quieted down cooperatively, not throwing one of their number to the wolves, so to speak. (Which in some species is however a tactic within the otherwise cooperative strategy.)

    And again, theoretical models are just that, theoretical. Tit for tat strategy, which may favor an aggressive purpose, is ironically, a cooperative response to competitive tactics. It involves mutually understood rules and consequences, procedures varying according to individual species and individual difference. Tit for tat doesn’t prevent a dominance hierarchy to develop, for example, but it does seem to prevent most species from eating each other in the process
    Accordingly, there’s no tit for tat between predator and prey, although depending on the particular prey or predator, there are of course mutually understood capabilities and related probabilities of consequence.

    As to empirical questions, the ultimate answers come from the most sophisticated analysis of the most realistic scenarios. Manipulation by experimenters always corrupts to some degree the reality coming out. Misunderstanding the fullest nature or role of the traits you have presumed you should be testing for is corruption going in.

    And what appears to be selfishness has the local advantage because in part the tit for tat strategy balances out that way – the balance of power is the other heuristic that applies here. Try to consider also that there’s a power shift from individual to group when groups become the competitors in the game, and the individuals in each group must cooperate to enhance its power.

  20. #20 piker
    November 7, 2009

    Relevant quote from review of Complexity: A Guided Tour, by Melanie Mitchell, a book which they say explores the nascent science of complexity – the “sheer range of settings” in which complex systems operate.

    “All theoretical models are wrong, but some are useful.”

    Haven’t read the book but noted a reference to it elsewhere on scienceblogs.

  21. #21 piker
    November 7, 2009

    I hadn’t seen Omar’s comment yet when I posted my response to Dr. Wlson’s. But it prompted me to look up a related study of theirs done on selfish punishment. A review of which at Economist’s View included the following:
    “Wilson readily acknowledges this limitation of the selfish punishment model…, “there’s nothing telling us that that mix is an optimal mix,” he explains. The answer to that problem, he says, is competition not between individuals in a group but between groups. That is because whereas selfishness beats altruism within groups, altruistic groups are more likely to survive…”

    Evidently because, as I noted above, it’s the competition between groups that changes the dynamic within each group. Sort of a “duh” moment, I suppose, but it does seem to place the emphasis on the arenas being determinant of the strategies rather than the strategies being more the determinants of their arenas.

  22. #22 mollyrogers
    November 8, 2009

    The coinage “all models are wrong, but some are useful” is from the British industrial statistician George E. P. Box (of Box-Jenkins fame).


  23. #23 piker
    November 8, 2009

    Coinage? Box should have included “theoretical” if given credit for the phrase, since otherwise the meaning is not quite to the point here.

  24. #24 mollyrogers
    November 8, 2009

    I’m glad you raised that point. What do you see is the significance of the qualifier “theoretical”? In other words, “theoretical models” as opposed to what other kind of models?

  25. #25 piker
    November 8, 2009

    Architectural models, design models, etc. Which will always be wrong to some extent, but are not necessarily predictive or intended to be. Designating the category gives us more of a reason to find the general statement relevant to the problem at hand. The inference to be drawn is that these particular models are wrong, not just because they are models, but because they are theoretical to boot.

  26. #26 mollyrogers
    November 10, 2009

    I looked at Melanie Mitchell’s book online, and she quotes directly from Box: “All models are wrong, but some are useful.”

    The version you cite (“All theoretical models are wrong, but some are useful”) is apparently from a review of the book by Bryce Christensen, an author who writes on Christian conservative themes.

    We don’t know what his reasons were for making that change. The distinction is meaningful to you, at least.

  27. #27 piker
    November 10, 2009

    I said at the outset that the quote was from the review and that I hadn’t read the book. But hey, taking your inference a step further, if a Christian conservative was involved, that would make either version suspect.

  28. #28 mollyrogers
    November 10, 2009

    Box’s original statement is profound, as Mitchell recognized. You quoted a third hand source that did not reflect the original, and when that was pointed out to you, you spun a rationalization for why Christensen’s version was in fact superior.

    I did not say Christensen’s version was “suspect”, but it is arguable he might find it meaningful to make a distinction between (human) “theory” and (divine) “truth” that most scientists or philosophers of science would not make.

    I do think your quick argument for the superiority of the version you quoted demonstrates your primary concern is defending a position you have identified yourself with, rather than openly considering the evidence.

    One could say it is individually advantageous to be facile at cauistry — and you are very good at it! 🙂

  29. #29 mollyrogers
    November 10, 2009

    [make that “casuistry”]

  30. #30 piker
    November 10, 2009

    My comment was about coinage. Box didn’t coin the statement made by the reviewer. I pointed out the fact that it came from the review, and hadn’t read the book. You came up with your snarky nitpick, which for all you knew was a pick against the book’s author for lack of attribution. I said I liked the newer version better, and said why. You spent a lot of time trying to prove that I shouldn’t have liked it better. Then you find a religious wedge that you could use to make your point, because otherwise all you still had was your opinion that the one version was profound and therefor the other wasn’t.

    I defended my opinion with illustrative evidence. You had no rational objection to the logic or examples at the time. But it seems you hate to look like a loser, so you look for some other way to perfect your snark.
    Now you feel you can claim I was just being defensive because, irony of ironies, you have come up with a better defense. Except your best evidence is an inference that the original version is more profound because Box had to be more rational than some Christian conservative.
    Talk about casuistry! Your picture is next to the dictionary definition. They moved it over from the snark section.

    (Aside to DSW: I hereby snitch on mollyrogers for questioning my moral integrity. Although her lack of ability to make substantive commentary about the subject at hand could be a mitigating factor.)

  31. #31 mollyrogers
    November 10, 2009



  32. #32 piker
    November 11, 2009


  33. #33 Alan Kestner
    January 24, 2010

    I am studying group selection in human society and have come up with the concept of the “managed-group”. This is a group structure which combines both selfishness and altruism. The selfishness is exhibited by one individual who creates and controls the group. He promotes altruism in the group (although not practicing it himself) so that they work harmoniously together. This one-sided state of affairs between him and the group is accepted because:
    1) the group needs a gatekeeper who makes sure no-one is cheating.
    2) the cheating that now goes on is limited to one individual – the leader
    2) the rewards of group working are so great that both sides can still benefit.
    3) evolution works on individuals to give them the tools to create these “managed-groups” (e.g. language, which of course makes no sense if only selected for at the individual level)

    Competition between these managed-groups then pushes their “evolutionary-like” adaptation further so that those that are better adapted for their particular function flourish at the expense of those that aren’t (e.g. competition between companies within a market sector).

    Can you tell me if any research has been undertaken on this concept? And are there any animal examples which show the same “managed-group” structure?

New comments have been disabled.