I'm a biologist, and I never realized that group selection wasn't broadly accepted until I read this article. I'd always assumed it was trivially implied by the theory of natural selection. Natural selection has been observed at every level, from entire organisms all the way down to the subcellular level (transposable elements, mitochondria, and others) I've never doubted that it also operates on populations of organisms.
I've often explained aspects of human society to my non-scientist friends in terms of group selection. (Short version: Humans who formed tribes out-competed the loners.)
On an unrelated note, I don't see how group selection and kin selection are mutually exclusive.
Kin selection: organism X helps organism Y because it improves the prospects for the part of X's DNA that Y also has. The more related X is to something, the more willing X is to sacrifice itself for it. The most extreme example is also the most familiar: each of your cells has DNA identical to each of your other cells (if all is well). So a cell is completely willing to kill itself for the "collective", even at the expense of its own ability to reproduce. A less extreme example is that of a woman declining to pursue kids of her own in order to help tend her sister's, which could be adaptive when, e.g., food is scarce.
Group selection: members of a particular group (e.g., a tribe) are willing to engage in altruistic actions (e.g., suicidal defense of said tribe pre-procreation) even though doing so maladaptive at the level of the individual's DNA. This kind of cooperation is adaptive for the group as a whole, but not in the least for the individual. Groups that encourage this kind of behavior will outcompete groups that don't, so groups are selected for this trait.
IIRC, the primary reason that group selection is thought to be wrong is that it (apparently) can't yield an Evolutionarily Stable Strategy (ESS) -- it will always be better for the persistence of any individual's DNA if that individual refuses to be altruistic.
>IIRC, the primary reason that group selection is thought to be wrong is that it (apparently) can't yield an Evolutionarily Stable Strategy (ESS) -- it will always be better for the persistence of any individual's DNA if that individual refuses to be altruistic.
I don't think that could be right, because the altruistic individuals altruism could include killing/punishing non-altruistic individuals. Shooting deserters comes to mind, as a real-life human example.
"IIRC, the primary reason that group selection is thought to be wrong is that it (apparently) can't yield an Evolutionarily Stable Strategy (ESS) -- it will always be better for the persistence of any individual's DNA if that individual refuses to be altruistic."
This sounds reasonable if you consider only one generation, or only the individual. But if you consider the whole group or species in the long run, the effect you see is exactly what you describe, i.e. groups that encourage this kind of behavior will outcompete groups that don't. So for a species to survive, it is much more important that the group survives than just an individual. If the individual survives, but it can't reproduce, then the species won't survive in the long term.
I think the important thing is to determine what behavior better represents reality, meaning what would explain the survival of current species on earth, not just try to confirm survival of the fittest for an individual.
> IIRC, the primary reason that group selection is thought to be wrong is that it (apparently) can't yield an Evolutionarily Stable Strategy (ESS) -- it will always be better for the persistence of any individual's DNA if that individual refuses to be altruistic.
"always will be better" seems to assume that group viability has no effect on individual survival.
While it's best to be the most selfish in the group, that doesn't imply that it's good enough. You need to also survive, and some groups help with that by having some altruism.
Re exclusivity: The article portrayed it as a debate over which one is the cause. Does group altruism exist because we desire to help kin, or does kin altruism exist *because we desire to help the group?
Group selection sounds more reasonable (by being more easily implementable at a sub-conscious level) to me.
I think what rcthompson was asking is why can't both kin altruism and group altruism exist as primary causes? If there's a genetic (as opposed to memetic) basis, it's conceivable* that there is a single gene that, if mutated, would cause individuals to care about the group, but not direct kin any more than the group at large, or to care about close kin, but not the larger group. Given how much of biology is layers of overlapping, cooperative systems, I would guess this is the most likely idea: that both kin selection and group selection are active, to varying degrees in different populations.
What does it mean to say that natural selection is observed at the level of an organism? I'm not a biologist, but I was under the impression that natural selection on all of the 'levels' of biology all shook out to being natural selection of snippets of DNA.
Natural selection at the level of organisms is just the "regular" form of evolution that we are familiar with. I used the word "organism" simply as a contrast to other levels at which natural selection can act.
Note that although the DNA is the genetic material, a selective pressure such as the threat of predation acts on organisms. Very rarely would selection act directly on the DNA. Rather, the correlation between carrying a snippet of DNA and not getting eaten causes that snippet to become overrepresented in the population over time.
Got it, thank you. I'd love to see a good pointer to worked out examples of natural selection happening at non-organism levels if you have one handy. It sounds pretty fascinating.
Well, one of the best cell-level examples I can think of is somatic hypermutation, the mechanism by which your immune system refines a good antibody into an excellent one. Briefly, your body starts out with the B-cell that produces the good antibody, and randomly mutates the critical regions of the antibody gene, generating a large population of B-cells that each produce slightly different antibodies. The B-cells with the best antibodies (highest affinity for the antigen) are selected and given the "do not die" signal. The rest of the B-cells do not receive this signal and self-destruct. The end result is a population of B-cells that produce higher-affinity antibodies than the original. Repeat while ill.
I may misremember some of the details, but that's the gist. Your body basically generates antibodies via an evolutionary algorithm. Wikipedia has a good explanation. Keep an eye out for the word "selection": http://en.wikipedia.org/wiki/Somatic_hypermutation
Wow. That is nothing short of amazing. Thank you, your comment motivates me to learn more about biology. For one, I have been labouring under the illusion that the brain holds a monopoly in terms of organizing intelligent behaviour. The immune system is clearly another. Got any more examples?
This suggests that it takes a complex organism to fight disease. Once multicell organisms developed do you think an immune system was the first type of organized intelligence or that foreign invaders and parasites served as a catalyst to ever more complex systems?
It doesn't necessarily take a complex organism to fight disease. Pretty much every organism, including unicellular ones such as bacteria, possesses some form of innate immunity against that organism's most common pathogens. Simpler organisms fight disease by simply reproducing faster than the disease can kill them.
As for the brain having some sort of monopoly on "organizing intelligent bahavior", that is a complex issue. I wouldn't say that the adaptive immune system necessarily possesses intelligence. It consists of many small and relatively simple components that interact to produce complex behavior. Realize that during somatic hypermutation, the immune system is not intelligently choosing which mutations to generate. It simply generates a random sample of all possible mutations and then sets up a simple selection process that produces a result that looks intelligent to us.
On the other hand, one could plausibly make a similar argument about the brain's intelligence being the product of the interaction of individual neurons. Many people believe that our brains are no less deterministic than a silicon-based computer. Complexity is complex.
Anyway, to give you another example, cancer cells are subject to greatly increased mutation rates, and mutations that result in faster growth or better recruitment of nutrient-supplying blood vessels (vascularization) will quickly become fixed in the tumor as the cells with that mutation overtake the growth of others. This also provides an example where selection at the cellular level is at conflict with selection at the organismal level. Within the tumor, selection favors faster-growing cancer cells, but these same cells are clearly the most detrimental to the organism as a whole.
The remarkable thing about his new theory, group selection is that to an outsider and a lay person, it sounds simpler and a lot more dynamic then the original kin selection theory. This is no crackpot prediction by someone who enjoys causing a ruckus. It very neatly explains related observations such as adhoc gangs/groups which are as common in the animal kingdom as they are with humans.
I'll put a calendar reminder in 2 years from now to check back on how this one turned out :-)
Thanks for sharing! This is indirectly but wonderfully on topic for this site.
Are adhoc gangs/groups that perform altruism common across genetic lines?
While some animals do exibit adoption, I don't think its nearly as common as altruism towards biological children. And group membership that requires altruism overwhelmingly seems to have genetic tie-in.
In order to have altruistic behavior in a group you need to spread a gene that promotes such behavior within the group. The simplest way for this to happen is for the gene to appear in one member, and then spread to others by inheritance.
For this to happen across, say, two genetic lines, it seems you need the "cross-altruistic" genes to emerge more or less simultaneously in the two groups so they can cooperate. The probability for this to happen is smaller, so it's not surprising that it happens less often than kin-style altruism. But if this does happen (and apparently it does), then it means kin selection is not enough to explain altruism.
It seems to me that Wilson's theory is an elegant generalization of kin selection. I don't see the contradiction between them, although I'm not a biologist so quite likely I'm misunderstanding something.
"In order to have altruistic behavior in a group you need to spread a gene that promotes such behavior..."
This premise seems to be taken as fundamental dogma in the evolutionary theory community, but perhaps should be questioned. Here is my (layperson's) counterexample: a species might have a gene that switches between whether an individual is a drone or a reproducer over the course of their life; a switch like that could thrown early in the development based on ambient hormones or whatever. If the individual is not a reproducer, they don't spread their own genes but they help the hive or whatever.
I think there is an unexamined assumption guiding contemporary evolution, that ALL behavior must be explained by genes ALONE -- in the case above, lifetime behavior is determined by an environmental trigger, even though the set of possible behaviors is determined by genes. This dogma seems, well, silly and wrong.
(Throw in cultural behaviors with outcomes like longer life and more offspring, and the classical system completely falls apart.)
Certainly not all behavior is the direct result of genes; I don't think this is the common viewpoint. For example there are genes that enable learning. This is extremely useful since disseminating behavior by learning is much faster than doing it through evolution. The rapid changes in human behavior in the last few decades (at least) can be explained in this way, and not by natural selection which is too slow.
Genes that encode specific behaviors are just the simplest possible mechanism that works, and they do account for a vast number of phenomena. I think Occam's razor applies: If you suggest a more complicated mechanism, you should have some supporting evidence.
Well, elephant mothers in a group help protect each others' young from predators by creating an impenetrable elephant wall around them. A single elephant mother could not protect her own young by surrounding them in this way, but she can protect her young by joining the group, if she is also willing to help protect others' young as well.
On the other hand, members of a herd are usually genetically related, so it is possible that kin selection can also explain this phenomenon.
>Are adhoc gangs/groups that perform altruism common across genetic lines?
It's common in the sense that symbiotic relationships are common across ecosystems and probably a definitive element of what makes them a "system".
Stephen Gould took the idea further in his idea of "hierarchal evolution". Evolution isn't just happening on the genetic, individual, group, and species levels. It's happening at every level of organization.
Thanks for this link. That helped me understand why group selection may be intuitive but is most likely wrong. To summarize, group selection can work mathematically, but you must carefully choose the parameters of the mathematical model in order for "grep-selective" behavior (i.e. altruism toward strangers) to be selected. (These parameters define things like the cost and benefits of altruism to the "giver" and "receiver".) The slightest variation in these parameters results in a collapse of altruism and reversion to selfishness. So while it is theoretically possible, in practice the incentives just don't line up in a way that selects group-selective behaviors over selfish ones.
My guess is that E. O. Wilson is probably right about ants, David Sloan Wilson is probably not right about group selection being like Daniel Dennett's universal acid, and Steven J. Gould was probably wrong in most of what he promoted.
>Yet, strangely, self-sacrifice exists in the natural world, even though it would seem to put individual organisms at an evolutionary disadvantage: The squirrel that lets out a cry to warn of a nearby predator is necessarily putting itself in danger.
its obvious - just compare 2 species of squirrels, one which demonstrate behavior described above and another which doesn't. Which species would survive over time better? Without making complicated statistical calculations, the Nature has already calculated the answer - we have the squirrel species of the first kind, and not of the second.
It is also obvious why selfishness still exists - because altruistic traits seriously decrease the survivability chances of the individual specimen carrying the traits.
Well no, because selection operates at the level of the individual, not the species. A "defector" mutant squirrel without the instinct to cry out when it sees a predator would have an advantage over the regular squirrel.
Some folks have done those complicated statistical calculations, and I believe the current consensus is that group selection is a pretty minimal effect, except in exceptional circumstances.
On the other hand, I'm not sure this is a great example. As a big scary potential predator I've scared away many a predator in my time, and y'know what? They don't cry out at all! They just run!
Does anyone have experience with screaming squirrels?
>Well no, because selection operates at the level of the individual, not the species. A "defector" mutant squirrel without the instinct to cry out when it sees a predator would have an advantage over the regular squirrel.
Evolution is a statistical macro process aggregating individual selection events. Thus the "defector" mutant squirrel have higher chances to survive, yet the (sub)species consisting of only such "defector" squirrels will lose to the (sub)species where statistically significant share of specimen demonstrate altruistic behavior.
People discussing evolution usually make 2 major mistakes :
- assigning explicit "species evolution/survival" level motivation to specific action of specific specimen
- directly extrapolating individual specimen events to the level of species, not paying attention to emergent higher-order system behavior
>Some folks have done those complicated statistical calculations,
such calculations is like law logic - can be bent both ways.
Have any idea about chaotic dynamical systems and the effect even small perturbation can cause?
>Does anyone have experience with screaming squirrels?
if you hike in the SF Bay Area hills, you'll see the high social organization they exhibit, including "watch duty".
Edit: for illustration, as somewhat related in principle - recently "Schneier on security" posted
http://www.schneier.com/blog/archives/2011/05/status_report_...
about "dishonest minority" - their individual situation is better, yet the species consisting entirely of the "dishonest minority" type specimen would obviously lose.
Do those models account for the fact that selfish individuals might get punished by the group? What if silent squirrels get systematically excluded from the screaming squirrels group? Wouldn't the loners have a lower life expectancy?
Does anyone understand enough to tell us what the "math" is all about? According to the article, the math is key to this ... yet they don't give any info at all about what that math says and how it is relevant. If anyone can enlighten us, please do!
It's up behind a paywall (probably... with my university connection I'm never sure what has it and what doesn't) here: http://www.nature.com/nature/journal/v466/n7310/full/nature0... . If anybody is more mathematically sophisticated than I, wants to tackle it, but can't get through the paywall, send me an email (in my profile).
Skimming through the supplemental information, it seems like they assume a Markov process where individuals adopt one of two strategies (with a preference towards adopting the strategy that their parent adopted), then look at which one is more likely to be dominant in the population.
The math is about the relative costs and benefits of altruistic behaviors. Altruistic behaviors must be a net benefit to an individual or group in terms of reproductive success, or else it would die out of a population. For a business analogy, the behavior needs to turn a profit.
In the case of kin selection in particular, the math is described by Hamilton's Rule, an inequality that must be satisfied in order for kin selection to explain a given altruistic interaction between individuals (http://en.wikipedia.org/wiki/W._D._Hamilton#Hamilton.27s_rul...).
Did Richard dawkin's "selfish gene" not move us beyond this point in the debate? With the focus on survival of the fittest _genes_ rather than the fittest organisms altruism seems far more logical.
The problem is that nobody knows what a "gene" is. It's a pseudoscientific term that implies we know where to find traits (which could be conserved by evolution) in a genome. We don't.
Regardless of how you define what a "gene" is, the "unit of selection" is a fragment of an organism's dna, not the organism itself. In this context Dawkin's explores an ESS(evolutionary stable strategy) that exhibits altruistic behaviour. It's also worth noting that group selection is shown to be unstable as an evolutionary strategy. This review of The Selfish Gene gives some good background http://www.miketuritzin.com/writing/review-the-selfish-gene/.
It's a lot like a binary diff, where phenotype is the source code, natural selection the programmer, and the genome the binary. Natural selection works on the phenotype, and you are saying that at all times the binary diff is compact, or sensible to talk about. Untrue. In fact it's trivially untrue because the genome is too small.
Group altruism is stable and is a very strong attractor of population behavior. All that's required is that group members be able to recognize one another.
http://people.brynmawr.edu/twong/models/pseudoaltruism.html
Individuals born into such a group may learn this or it may be genetically inherited (instinctual), or both. Of course the distinction between learning and instinct is also a bit fuzzy.
Note that in the tournaments that Axelrod organize, in order for a cooperate strategy to thrive, there must be a critical mass of other cooperate strategies too.
Kin selection, which Edward O. Wilson is attacking, is the current explanation of how this critical mass of cooperative entities come about. If kin selection is wrong then an alternative explanation must be found to explain the appearance of this critical mass of cooperative entities.
Two things: first, kin selection is one possible mechanism by which a critical mass of cooperators can arise, but it is not the only such mechanism. And second, the mechanism that gives rise to cooperation may have evolved for some other reason, with cooperation as an ancillary effect.
For example, inter-species cooperation (as between dogs and humans) is almost certainly not a result of kin selection.
So even if Wilson is right and the theory of kin selection is flawed (I can't imagine how he could be right, but I haven't read the paper) that is not an indictment of Axelrod.
I wonder if there is any type of controversy more exciting than this? - Memes and Dawkins are overrated... Btw other very interesting scientists/thinkers on Anthropology, Evolution theory related to humans, are Dan Sperber, Boyd & Richerson "Not by genes alone", Michael Tomasello "why we cooperate"...
I've often explained aspects of human society to my non-scientist friends in terms of group selection. (Short version: Humans who formed tribes out-competed the loners.)
On an unrelated note, I don't see how group selection and kin selection are mutually exclusive.