# Selection on many levels....

I'm rereading Unto Others, David Wilson and Elliott Sober's argument for Multilevel Selection. One of the core planks in the book is that supra-individual levels of selection are necessary for the evolution of altruism, and much of the book details what Wilson & Sober perceive are the stretched and implausible explanations that scientists who are straight-jacketed into individual selection need to concoct. So how can selection between groups lead to the emergence of altruism?

Consider two groups:

Mixed & Selfish

Assume that both groups begin in generation 1 with 100 individuals.

Mixed:
50 altruistic individuals
50 selfish individuals

Selfish:
100 selfish individuals

Now, assume that the two groups are distinct & separate.

Selfish grows at a rate of 1% per generation. After 30 generations Selfish will have ~135 individuals.

Mixed is more complex. The altruists provide a benefit to the group so that the selfish individuals reproduce at a rate of 2% per generation. In contrast, the altruists have a cost so they only grow at 1.25% per generation (imagine that in pairwise interactions they sometimes have to deal with selfish free riders). After 30 generations you have ~163 individuals. Clearly the mixed group is "fitter" in terms of increase. But, where in generation 1 the altruists were 50% of the Mixed deme, they are now 44% of the deme. Though they increase the fitness of their group, they are selected against within group. Additionally, if you add all selfish individuals together (between groups) then you note that they've more than doubled. In contrast, the altruists have only nearly increased their numbers half again.

In the far off future of generation 300 in the mixed deme the altruists are 8% of the population and the selfish 92%. But, there is an interesting development, here are the absolute numbers:

~3015altruists in Mixed
~34437 selfish in Mixed

~2667 selfish in Selfish

Though the altruists grow far slower in number then non-altruists, over the long term they've managed get ahead of the total size of the Selfish deme, which in absolute numbers if you will recall consisted of 100 individuals vs. the 50 altruists in the Mixed deme in the initial generation. Simply counting across demes and lumping the altruists and selfish together as if they were an amorphous mass removes information about the dynamics of population expansion and growth. This is a form of what Wilson & Sober term the "averaging fallacy," just averaging fitness across groups of individuals with distinct phenotypes and eliding the substructure of arrangment. But, note that the percentage of altruists keeps declining. In the real world exponential growth eventually is restrained by resource constraints of some sort, and at that point one assumes that the altruists will slowly converge upon extinction as the selection coefficients favor the selfish.

But this is where step two comes into play: dispersal. Consider that at generation 300 the two groups disperse into sub-demes of approximately ~100, as in the initial scenario. Also imagine that of the 8% who are altruists they assortatively cluster with each other so that they are distributed in a 50/50 fashion with the selfish, while the vast majority of demes are exclusively selfish. What will occur is simple: the demes with the altruists will supersede those which are exclusively selfish...until once more we are at the point where the altruists are a benighted minority amongst their selfish peers.

And so on. The above was a "toy" example, real interdemic selection models are generally more complicated (e.g., imagine for example that the fitness of the selfish within the Mixed demes was dynamically dependent upon the proportion of the unselfish). The great evolutionary biologist John Maynard Smith did not accept the relevance of group level selective effects primarily because of the complexities of the dynamic interactions, in particular the necessity of fissioning in the manner alluded to above. Workers such as Smith implied that the "cheating" of selfish individuals, the power of within group/individual level selection always seems like it should dominate the between group variation (which relied on longer "generations"). Additionally, in most populations between group genetic variance is far less than within group variance, and evolutionary response to selection is proportionate to variation. These are some of the reasons that theories such s kin selection and reciprocal altruism emerged, they are fully reduced down to the level of the individual and so exhibit more apparent parsimony.

A more technical elucidation of these topics can be found at this post Defining Group Selection: Price's Equation. I will be posting on this topic more in the near future....

Tags

### More like this

##### Cooperation and multilevel selection
A few days ago I introduced how higher levels of selection could occur via a "toy" example. Obviously it wasn't realistic, and as RPM pointed out a real population is not open ended in its growth potential. I simply wanted to allude to the seeds of how Simpson's Paradox might occur, where…
##### Innate social aptitudes of man
Innate social aptitudes of man is a controversial paper. As noted in the biographical introduction to it William D. Hamilton states that his friend Robert Trivers referred to it as the "fascist paper" (see Natural Selection and Social Theory for Trivers' perspective on his relationship with…
##### What does not kill the group, makes it stronger!
I recently finished reading The Faith Instinct: How Religion Evolved and Why It Endures, a new book by Nicholas Wade, a science writer for The New York Times. Before giving it the "full treatment" I thought it behooved me to revisit some of the scientific literature which Wade relies upon to give…
##### Levels of selection & the full Price Equation
In the post below on the Price Equation I stayed true to George Price's original notation in his 1970 paper where he introduced his formalism. But here is a more conventional form, the "Full Price Equation," which introduces a second element on the right-side. Δz = Cov(w, z) / w One can…

Historians of many types find rise-and-decline patterns in social history (both in nations and, in China, in the great familes), where the early generations are upright, frugal, and diligent, and later generations hedonistic, lazy, and corrupt. This can be fitted to the "from altruism to selfishness" model, with a possible end result (if selfishness and "profit-taking" become proportionately too great) being a weakening of the society to the detriment of its individual selfish members. (But this is not necessary: large-scale, wealthy, well-developed societies have enormous advantages over poorer societies, even if the poorer societies are more "virtuous").

Roman writing about the Teutons, or Chinese writing about border peoples, often stress their virtuous nature. What this really means is that 1.) the Latin / Chinese writer was trying to shame their civilized peoples into being more virtuous, and 2.) the Chinese / Latin writer realized that barbarians make better soldiers, and were thus a threat, but also potential mercenaries. The civilized peoples ultimately had no real admiration for the barbarians.

I should add that the cycle I described is almost certainly not genetic, but cultural or behavioral. It has a sort of analogy to selfish genes increase generationally in a population, but what really happens is that the altrusit social norms imposed on the early, impoverished generations become unenforcible on later, wealthy generations. So this is really just an analogy.

But real populations cannot increase in size indefinitely -- there are limits to the number of individuals. The within population frequencies are important because even if the altruists outnumber the homogenious selfish, they make up less than 10% of the mixed population. It seems that the altruistic allele would reach a negligible frequency in the mixed population and be lost.

But real populations cannot increase in size indefinitely -- there are limits to the number of individuals.

i'm not a shill for group selection...obviously. but i'll explore some more realistic models soon....

(though, following up john's comments, a lot of simple group selection models which perpetuate altruism strike mas boom-bust cycle dynamics)

Not to be a stickler (read: jackass), but it's J Maynard Smith, not "JM Smith"

By Rich Lawler (not verified) on 05 Apr 2007 #permalink

No, I think the convention is that his first name is "J Maynard". Like "J. Michael Strazynski".

I realize it's not a very detailed model, but in the classic Prisoners' Dilemma Ecology simulation, a threshold number of the population had to possess the Tit for Tat strategy before it would take over the population. Too few, and it was destroyed by the selfish Always Defaulters who dominated the less ruthless strategies, then fell into inefficient (but very stable) competition.

There are good reasons to think that some traits need to become established among a small group before they can begin to affect populations. Look at the eusocial insects.

By Caledonian (not verified) on 05 Apr 2007 #permalink

As I understand it, his last name is "Maynard Smith" (no hyphen).

By Rich Lawler (not verified) on 05 Apr 2007 #permalink

I think there is something to this, especially since:

1. Altruistic people do seem to recognize each other and try to stick together (greenbeard).
2. Selfish people might protect altruists for their own selfish reasons (symbiosis).
3. Selfish people can pretend to be altruistic so as to be admitted to the club (meta-altruist phenotype).

Maynard Smith (without a hyphen) is the family name, e.g. his wife was called Sheila Maynard Smith, his father was Somebody Maynard Smith, his son is Somebody Maynard Smith, and so on. It probably started out as an affectation back in Victorian times - they thought plain Smith was too vulgar, so they added the Maynard, but hyphenated names are also a bit vulgar, so they didn't use a hyphen!

...on the issue of substance, the crucial question is why altruists should associate together. If it is because they are clusters of kin, it is difficult to distinguish from kin selection. If it is a genetic Green Beard effect, it requires a rather unlikely (but not impossible) bit of linkage or pleiotropy. If it is because altruists want to receive benefits from other altruists, it looks suspiciously like Trivers's reciprocal altruism. Sober and Wilson regard both kin selection and reciprocal altruism as forms of group selection, but this largely is a matter of semantics.

Kin selection is a form of group selection. It's just that it is often discussed separately from group selection because the generation times of both groups and individuals are the same. The issue is not just semantic, we can derive Hamilton's rule from the Price equation, treating the second term in the equation as "selection within a group" while the covariance term captures differential group survival and reproduction. This basically shows the "multilevel" structure of kin selection. In kin selection, we have multiple family units and variation in productivity among those units (relatedness, in Hamilton's equation serves as an approximation for "probability that another individual is also an altruist," which is a safe bet because if altruism is rare, then it's likely that two individuals that have the allele for altruism got it from a common ancestor, i.e., they are related.)

By Rich Lawler (not verified) on 06 Apr 2007 #permalink

I'm not a biologist at all, and think of this mostly in human / social terms. The problem that I fumble with is: How does biological evolution relate to social/cultural evolution, especially with regard to biological altruism (= sacrificing one's own personal fertility for the sake of the group's overall fertility)?

Following Boyd and Gintis, I think that for social coding for altruism is only possible if there already has been biological evolution making both socialization and altruism possible. All human societies demand some altruism, but some demand more than others. A group's socially-coded behavior (altruistic or otherwise) changes the environment for members of the group in ways whose consequences for genetic evolution are unpredictable. For example, social-economic-political success may or may not lead to reproductive success; in advanced societies, it seems not to.

I've looked briefly at Trivers' "Social Evolution", and there's a lot there, but it seems that culturalization complicates the story enormously.

It seems to me that the temptation to identify human social behaviors with the social behaviors described by Trivers is to be resisted, even though there are obvious congruences, because in human society you have two parallel evolutions (social and biological) feeding back to one another.

By John Emerson (not verified) on 06 Apr 2007 #permalink

Kin selection is a form of group selection. It's just that it is often discussed separately from group selection because the generation times of both groups and individuals are the same. The issue is not just semantic, we can derive Hamilton's rule from the Price equation, treating the second term in the equation as "selection within a group" while the covariance term captures differential group survival and reproduction.

Yes, this is exactly right. The typical derivation of Hamilton's rule is faulty because it only proves the increase in number of altruism alleles, not their frequency. For the latter the group structure needs to be taken into account. Hamilton published a book chapter in 1975 explicating the applicability of the Price Equation to group selection.

anyone interested in the group selectionist take on these issues should read the first half of unto others, a lot of it is devoted to hamilton's work and its relationship to multilevel selection....