I have just realized that I keep mentioning David Sloan Wilson a lot (see the list of links below), always in a positive light as I think he is one of the pioneers of modern evolutionary theory (as soon as those drunk on Williams 1966 and Dawkins' opus retire or die) but have never really written a good post on group selection. I'll have to do this one day soon - that may be my contribution to the Basic Concepts collection.
Anyway, Wlison just gave a talk in which he presented my favourite example of the test of group selection - in chickens:
Two experiments using chickens show another aspect of evolution regarding selection. In the first experiment, groups of chickens in cages were evaluated for egg-laying. The best egg-layer within each cage was chosen and put together with the other prolific egg-layers. The second experiment took the best caged groups of egg-layers. The result of the second experiment after a few generations was healthy, sociable, egg-laying chickens. The result of the first experiment after a few generations was fighting and anti-social chickens that maimed and killed each other.
"You pick the best," Wilson joked, "and in six generations you get sociopaths."
Those two papers actually came out in Poultry Science, not Evolution, thus they are not as well known by the evolutionary community as they should be. What Wilson did not mention in his talk is that the group-selection experiments resulted, over just a few generations, in a greater egg-production than ever achived in a couple of thousand years of selective breeding of chicken. It also resulted in a complete loss of need for de-beaking of chickens, which is a nasty procedure in poultry industry.
Perhaps these chicken don't peck each other to death because they all adopted the identical religious beliefs LOL!
You can see my earlier mentions of Wilson here (really this), here, here here, here, here, here, here, and here.
Update: Mike has more.
- Log in to post comments
Perhaps I'm confused, but how is this an example of group selection? You're just selecting for two traits instead of one: egg-laying ability as well as sociability (assuming, reasonably, that not spending time and energy fighting and/or being beat up increases actual egg-laying rate as well). Whereas when just picking the best egg layer, you do not select for sociability one way or another.
Picking whole cages is still also individual selection, just on a statistical basis: you pick hens that tend to lay more eggs (and that are sociable, or they would not lay as many).
Bora,
to really make sense of this experiment, we need to know some things that are missing:
-how many eggs are each of the chickens in both treatments laying?
-how many eggs are laid (total) in each of the cages for each treatment?
Without this, I can't really make sense of Wilson's data (with it, one could use ANOVA to analyze it).
There's also another problem here: the fitness of the 'uber-egg layer' is greatly amplified. Suppose, instead the eggs in both treatments were proportionately allocated to the next generation (or probabilistically so). In that case, 'psycho-chickens' wouldn't persist--they would increase, and then perhaps decrease (again, depending on Wilson's data, this might or might not happen--but it is possible). That would be frequency-dependent selection on individuals, not group selection.
I'm tired, so I'm sorry if this isn't coherent (I might just have to post about this later).
Has Wilson published these data?
Those are not Wilson's own data - but I can dig out the actual papers (I have had them somewhere in this mess for years) and give you the numbers.
The only trait selected was egg-number per group - all other changes came about as surprise to the investigators - they were by-products of selection.
The chicken work was done by Bill Muir. I guess this is the paper:
Muir, W.M. 1986. Relative efficiency of selection for performance of birds housed in colony cages based on production in single bird cages. Poultry Sci. 64: 2239-2247.
His motivation for the work was animal welfare rather than any theoretical concerns. I don't think he's a big fan of the (trait) group selection debate: and it certainly seems to generate more heat than light nowadays.
Bob
I also accept multilevel selection. However, it should be noted that Williams reconsidered higher-level selection in his 1992 'Natural Selection: Domains, Levels, and Challenges' and Dawkins accepts that there is clade selection for evolvability. Furthermore, Williams 1966 and Dawkins 1976 were not fundamentally wrong but merely incomplete. The gene is the primary unit of selection, but it is not the only one.
While endorsing DS Wilson's Unto Others, Richard Lewontin mentioned an unsavory aspect of group selection (NYROB, 10/22/98): namely, war is a mechanism of the differential survival and reproduction of whole groups. Out-group aggression goes hand in hand with in-group cooperation. It is very advisable to be mindful of the Naturalistic Fallacy when considering group selection.
Lewontin: "Success in war presumably enhanced group survival, however, so the altruistic act of the war chief, sacrificing himself for the group, would nevertheless lead to a survival and spread of the altruistic institution."
"it certainly seems to generate more heat than light nowadays"
I think group selection is no longer heresy, so I don't seem to get the essence of this comment.
This example uses artificial selection, which accompanied by models shows that group selection is theoretically possible.
Now, the current controversy (far away from the 70's controversy) regards to what extend this theoretical possibility seems to correspond with real world conditions. The parameter space that seems to make group selection important in the models appears to be restricted in the real world. Unfortunately there's a gap between theoretical and empirical works to shed light on this issue.
Nonetheless, it is now accepted that in some domains group selection played a very important role, particularly in human evolution.
This is a clear example that it can work. Artiticial selection at the level of groups yielded much more profit. I don't think it is much relevant whether the author is or not theoretically aligned with the group selection literature in this case. It is very common to see group selection arguments that yet even deny they're claiming group selection.
Correct. The theoretical models show it is possible. Lab experiments show it is possible. The search for real-world examples is still slow (but you knwo just demonstrating selection is super-difficult, even without teasing out the relative contributions of selection at several different levels).
The references I had in mind are, among else:
Muir, W. M., 1996. Group selection for adaptation to multiple-hen cages: Selection program and direct responses. Poultry Sci. 75:447458.
Craig, J. V., and W. M. Muir, 1996. Group selection for adaptation to multiple-hen cages: beak-related mortality, feathering, and body weight responses. Poultry Sci. 75:294302.
Effect of Genetic Selection for Group Productivity and Longevity on Immunological and Hematological Parameters of Chickens
H. W. Cheng, S. D. Eicher, Y. Chen, P. Singleton, and W. M. Muir
The work on Tribolium by Wade and Goodenough is probably the best, though Wade never mentions it in his talks (last I heard him speak).
A few years back Tim Clutton-Brock put up a slide that screamed "group selection" at the audience and said that he had no idea how to explain the data (on meerkats). Afterwards I asked him about group dseleciton as a likely/possible explanation, but he angrily refused to even talk about it, as he is one of the old-time genocentric hyperadaptationists.
I agree: that's why it generates more light than heat! The theory is pretty clear, and there are a few examples of it occuring, so why the fuss? It seems to be fairly rare in nature: there are only a few good examples of it, and I'm not aware of any in humans (although I think it's difficult, because our culture has evolved away from the conditions where it could have occurred).
Bob
Does it generate more light or heat?
My view is not that there are just a few cases. It may be the case, but these arguments seem to be rather weak since there's evidently a gap between theory and empirical work. Some models conclude that the parameter space is restricted, but this, to the best of my knowledge, is never linked to specific empirical evidence (particularly in humans), which I would expect given the nature of the argument.
Models specifically cultural unfortunately do not abound. Most early models tend to concentrate in the old controversy (i.e. show that it can happen), leaving out of scope a number of potential important mechanisms that may eventually well enlarge the parameter space for which group selection could work. One of those mechanism is direct warfare, for which an empirical test just recently appeared in science:
Group Competition, Reproductive Leveling, and the Evolution of Human Altruism, by Samuel Bowles (in Science 8 December 2006).
The idea is that since human group conflict was at some point particularly lethal, group selection could have been indeed very strong. This kind of direct warfare for instance is hardly considered in the models....an exception is a relatively recent model that I like very much:
The co-evolution of individual behaviorsand social institutions. Journal of Theoretical Biology. 223:135--147. 2003. Bowles, S. and Choi, J.K. and Hopfensitz, A.
It deals specifically with culture (institutions) and direct group conflict.
On these grounds it seems to me that the whole thing is not settled, and there's an exciting ammount of research mounting both at the theorerical and empirical level (that is at least when it comes to the case of humans).
Just as a reference, the model by Bowles et al. scores 60 references in google scholar. That ammounts to 20 references per year, which I don't think is that bad.
Most people, rightfully, focus on the toughest case - altruism. Sober and Wilson specifically do that in "Unto Others".
But step back for a second.
Think of the adaptive landscape.
Now place a population somewhere on the landscape.
Now, from the point where the species is sitting (A) draw two vectors: one, driven entirely by selection at the level of the individual takes the population to the point B; the other, driven entirely by selection at the level of the group takes the population to the point C elsewhere on the landscape.
Only in the unique case of altruism is the angle between two vectors exactly 180 degrees, directly opposing each other - this is what makes this special case so difficult.
In all other cases, the angle between the two will be much smaller, i.e., the selection at two levels is more-or-less going in the same/similar direction.
We do not yet have the mathematical or experimental tools to calculate the angle between the vectors (not to mention that selection act simultaneously at several different levels of organization, not just two - multiverse of vectors!), but that is the really interesting question in multi-level selection: how to tease different levels apart.
The selection at the level of the group is surely there - but at what angle to other vectors as well as how long is its vector (denoting intensity, i.e., strength of selection) in comparison to other vectors in any given case is what we need to know in order to figure out how important this level is in the real world.
Forgot to mention the obvious - addition of vectors will place you at a point D where the combined effort of selection at different levels will actually bring the population to on the adaptive landscape.
Perhaps (partially) conflicting selection pressures at multiple levels could help get a population unstuck from a local fitness peak so it can reach an even higher one. (Can't recall a cite.)
Why can't you just calculate the vectors independently for each level, and then add the vectors at the end?
Bob
The other day, I heard my friend Justin Werfel give a very interesting talk about his work on group selection. PNAS 101, 30 (27 July 2004): 11019-11024 is probably a good pointer into the literature.