Richard Dawkins (1) begins his review of E.O. Wilson’s (2) new book The Social Conquest of Earth with an appeal to authority—namely the 137 evolutionists who co-authored a reply (3) to an article by Martin Nowak, Corina Tarnita and Wilson in Nature magazine (4). Appealing to authority is a risky business in science, as Dawkins appreciates, because scientific progress often involves the few prevailing against the many. Nevertheless, Dawkins’ appeal to authority acknowledges something that everyone should know about group selection, kin selection, and all that: Dawkins and Wilson are only two of dozens of scientists who have been working on the issues over a period of decades. This is in contrast to their outsized images on the public stage, as if they are the only two figures meriting attention and all the important ideas sprang from them.
I mean Dawkins and Wilson no disrespect by calling them two among many. I trust that they would agree and would defer to others especially when it comes to mathematical models, which is not their area of expertise. If the public is going to become literate on the issues at stake—as well they should, because they are fundamental to the study of human sociality—then they will need to realize that both Wilson and Dawkins get some things right and other things wrong. Moreover, the entire community of scientists is in more agreement than the infamous exchange in Nature seems to indicate. Taking the argument from authority seriously can lead to a breakthrough in the public’s understanding of social evolution.
Wilson has written abundantly on his rejection of kin selection in favor of group selection, as he thinks of it. Dawkins’ review is the first time he has written at length on the topics of group and kin selection in many years, as opposed to little snippets here and there. I will therefore use Dawkins’ review to outline the zone of consensus that exists among the many, which both Dawkins and Wilson should abide by unless they provide compelling arguments and evidence to the contrary.
The Origin of Group Selection and Kin Selection Theory
The seeds of both group selection theory and kin selection theory are present in the work of Charles Darwin, and both were invoked to address a single problem—how traits that are “for the good of the group” can evolve when they are selectively disadvantageous within groups. Examples include the sterile castes of bees and the human moral virtues. As one answer, Darwin proposed that groups of individuals who behave for the good of their group would outcompete other groups, even if their solid-citizen behaviors were selectively disadvantageous within groups. This was the seed of group selection theory. Darwin invoked it repeatedly in the corpus of his work, as Elliott Sober (5) has documented in detail. Dawkins is wrong when he asserts that Darwin invoked group selection in only “one anomalous passage”; he needs to read and refute Sober’s article or acknowledge his own error on this relatively minor point.
As another answer, Darwin used the analogy of animal and plant breeders selecting traits that involve sacrificing individuals by breeding their relatives, which was the seed of kin selection theory. It is unclear whether Darwin regarded these as different or equivalent explanations, nor does it matter for anything but historical interest.
In the received history that just about everyone agrees upon, many biologists during the first half of the 20th century assumed that adaptations evolve at all levels of the biological hierarchy, including individuals, social groups, species, and ecosystems. Either there was no awareness that group selection might be necessary, or it was assumed that between-group selection would be strong enough to oppose within-group selection. Today this position is known as naïve group selectionism. It came under intense scrutiny in the 1960’s and a consensus emerged that while between-group selection is possible in principle, it is almost invariably weak compared to within-group selection in nature. If behaviors evolve that appear to be “for the good of the group”, then they must be explained without invoking group selection. All of the other theories for explaining the evolution of cooperative and altruistic social behaviors, such as kin selection, reciprocity, and selfish gene theory, were explicitly developed as alternatives to group selection.
As an influential example that is also described by Dawkins in his review, John Maynard Smith constructed a simple mathematical model in which mice live in haystacks for a number of generations before dispersing to colonize a new set of haystacks. The mice vary in their aggressiveness. Aggressive mice beat docile mice within haystacks, but groups of docile mice are more productive than groups of aggressive mice. The haystack model corresponds exactly to what Darwin described in words. Maynard Smith concluded on the basis of his model that the between-group advantage of docility was insufficient to counter the within-group advantage of aggressiveness. If docility exists in nature, it would need to be explained another way, such as W.D. Hamilton’s newly minted theory of kin selection—a term coined by Maynard Smith, which Hamilton called inclusive fitness theory.
Everything I have said so far falls within the zone of consensus of the many, including Dawkins and Wilson. I would be surprised if anyone knowledgeable about the subject disagreed with it, and I encourage them to speak up if they exist.
On Replicators and Vehicles
George C. Williams in Adaptation and Natural Selection (6), followed by Richard Dawkins in The Selfish Gene (7), articulated the concept of genes as replicators. According to their argument, sexually reproducing individuals are unique combinations of genes that will never reoccur and therefore do not qualify as units of selection. A unit of selection must replicate with high fidelity and only genes have this property.
Individual organisms don’t qualify as replicators, but they remain important as vehicles of selection, which explains their functional organization. A vehicle is a collection of genes that survive and reproduce as a unit. In an analogy made famous by Dawkins, vehicles are like rowers on a crew team that can only win the race by pulling together. Genes do occasionally evolve to succeed at the expense of other genes within the same organism, such as cancer and meiotic drive, but they are relatively uncommon and tend to undermine the functionality of the organism when they occur.
Williams borrowed the replicator concept from population genetics theory, where it is known as “average effects”—the fitness of alternative genes, averaged across all contexts. It is an important concept for many purposes, but it says nothing about group selection. Group selection models always posited genes for altruism (favored by between-group selection) and selfishness (favored by within-group selection). Whenever between-group selection prevails against within-group selection (however rarely or implausibly) the altruistic gene has the highest average effect and evolves in the total population.
A lot of confusion existed on this point, especially since calling genes the “fundamental unit of selection”, based on their status as replicators, made it seem as if this constituted an argument against group selection. By now the dust has fully settled, however. Everyone, including Dawkins in his review, agrees that if selfish gene theory has anything to say about group selection, it must involve something other than the concept of genes as replicators.
The only other major concept articulated by selfish gene theory is the vehicle. The paradigmatic vehicle is the individual organism, which provides very few options for genes to succeed other than as part of the collective. But here we encounter a problem: Recall Maynard Smith’s model, in which the groups are populations of mice living in haystacks. These groups are nothing like individual organisms. There is nothing restricting aggressive mice from outcompeting docile mice within the groups. Nevertheless, if the between-group advantage of docility was sufficient to counter the within-group advantage of aggressiveness, then it would count as an example of group selection. Nothing more organism-like would be required of the groups. When we survey other models and verbal descriptions of group selection prior to the advent of selfish gene theory, the assumptions about groups vary in their details, but all of them are mere collections of individuals in which altruistic traits are selectively disadvantageous but nevertheless can potentially evolve on the strength of the differential contribution of the groups to the total population. Organism-like properties of the groups were never required.
Thus, if we define the vehicle concept too narrowly in reference to individual organisms, then it fails to represent what has always been understood by group selection. Either selfish gene theory is inadequate for studying group selection, or the vehicle concept must be broadened to include the groups that have always been posited in group selection models.
As an aside, both high-fidelity replicators and individual organisms as airtight vehicles are products of evolution, which means that they are not required for the process of evolution to take place. The original argument that sexually reproducing individuals don’t qualify as replicators failed to identify single phenotypic traits as candidate replicators. Socrates might never come again as a large collection of traits, but the shape of his nose or aspects of his intellect might reoccur with the same frequency generation after generation. As long as the phenotypic distribution of a trait is replicated and responds to selection, high-fidelity replicators are unnecessary. The concept of “evolution without replicators” is required not only to describe evolution before the advent of genetic replicators (8), but aspects of ongoing cultural evolution (9).
The Advent of Equivalence
A watershed event took place in the 1970’s, when W.D. Hamilton encountered the work of George Price, which has been ably chronicled by Oren Harman in his book The Price of Altruism (10). Price developed an equation that partitioned selection in the total population into within- and between-group components. It was like Maynard Smith’s haystack model in concept, but much more general, and it showed that between-group selection could plausibly prevail against within-group selection. Moreover, when Hamilton translated his own inclusive fitness theory in terms of the Price equation, he realized that altruism expressed among kin is selectively disadvantageous within kin groups and evolves only by virtue of the differential contribution of kin groups to the total population. In other words, inclusive fitness theory was not an alternative to multilevel selection theory but included the elements of multilevel selection within its own structure. The reason this was not obvious before is because Hamilton’s method of calculating the net effect of an altruistic gene on copies of itself predicted whether it would evolve in the total population, but did not reveal its local selective disadvantage and the corresponding need for between group selection to evolve.
Other theoretical models published during the 1970’s and 80’s, including my own trait-group model (11), Michael Gilpin’s model that was the first to incorporate nonlinear interactions (12), and my re-analysis of Maynard Smith’s haystack model (13), showed that the wholesale rejection of between-group selection as a significant evolutionary force was just plain wrong. Hamilton’s realization for inclusive fitness theory goes for all evolutionary theories of social behavior, regardless of what they are called. All assume that social interactions take place in groups that are small compared to the total population. The traits labeled cooperative and altruistic typically are selectively disadvantageous within the groups and require the differential productivity of groups to evolve in the total population. This is expressed in the Price equation as a negative within-group term and a positive between-group term that sums to a net positive average effect for the altruistic trait. The Price equation can accommodate many kinds of groups, including those that have always been the province of group selection models.
The Price equation also helped Hamilton realize that his original interpretation of r, which was largely restricted to genealogical relatedness, could be generalized to include any correlation between the genes of the donor and genes of the recipient(s), for any reason. Moreover, Hamilton’s original formulation remained useful for calculating what evolves in the total population, even if it didn’t partition selection into within- and between-group components along the way. The choice of which framework to use became largely a matter of preference, with any given result from one framework translatable into the other framework.
Thus was born the era of pluralism and equivalence in sociobiology. It has become part of the zone of consensus of the many, but Wilson and Dawkins are not among them. Both fail to recognize that the era of “kin selection vs. group selection” has passed. Most of the important questions can be asked within either framework and can be translated between frameworks.
What prompted the 137 co-authors to respond to the Nature article was not based on what Nowak et al. said about group selection, but their denial that it could also be framed in terms of inclusive fitness theory or that ideas framed in terms of inclusive fitness theory had ever proven to be useful.
Holding Dawkins Accountable
Curiously, while the many have spoken against Wilson’s outdated views about kin selection, they remain largely silent on Dawkins’ outdated views about group selection. I will therefore list points made by Dawkins in his review (in italics) that have been rejected by the many long ago. Dawkins can argue on their behalf if he likes, but he cannot invoke the argument from authority.
1) “Group selection: the poorly defined and incoherent view that evolution is driven by the differential survival of whole groups of organisms”. Multilevel selection theory is defined as coherently as inclusive fitness theory.
2) “The American grey squirrel is driving our native red squirrel to extinction…” Choosing an example of competitive exclusion between species to illustrate group selection is poorly informed. Even the concept of species selection, which is different than the concept of group selection within species, is not represented by competitive exclusion (14).
3) “So biologists with non-analytic minds warm to multilevel selection…” Multilevel selection models are just as analytic as inclusive fitness models. In fact, proponents of inclusive fitness theory often praise it as more intuitive than multilevel selection theory. As an aside, accusing those who disagree with you of mental weakness is unbecoming to a scientist.
4) “With the exception of one anomalous passage in The Descent of Man…” As previously mentioned, Darwin articulated the concept of group selection numerous times in the full corpus of his work. If Dawkins wishes to claim otherwise, he should respond to the detailed scholarship of Sober. Historical revisionism is unbecoming to a scientist.
5) “Genes are the primary replicators, organisms are the obvious vehicles, but what about groups? As with organisms, they are certainly not replicators, but are they vehicles? If so, might we make a plausible case for ‘group selection’? This passage is followed by a list of examples involving huddling penguins, hunting lions, birds in flocks, fish in schools, racing cyclists, and hens in flocks. In most of these examples, the basic problem of cooperation and selfishness arises. Whenever the traits that maximize relative fitness within groups are not the same as the traits that maximize collective benefits, between-group selection is required to counteract within-group selection. Groups do not need to be organism-like in any other respect, as previously explained. Using inclusive fitness theory to study the same examples does not change any of the facts concerning within- and between-group selection.
6) “Convincing examples are vanishingly hard to find”. There is abundant empirical evidence for between-group selection, including extensive experimental evidence (15). Hens in flocks provide a compelling example, in which aggressiveness is favored within groups and docility between groups, just like Maynard Smith’s haystack model (16). The reader is invited to type “group selection” and “multilevel selection” into Google Scholar to consult both the theoretical and empirical literature.
7) “[Wilson] treated kin selection as a special case of group selection, an error that I was later to highlight in my paper on “Twelve Misunderstandings of Kin Selection”. Wilson was right to conceptualize kin selection as a special case of group selection during the 1970’s, which was consistent with Hamilton’s interpretation based on the Price Equation. Since then, inclusive fitness theory has been generalized to the point where it is equivalent to multilevel selection rather than a special case. Maynard Smith’s haystack model, for example, which he created to distinguish group selection from kin selection, can easily be seen as an example of kin selection.
8) “Bert Hölldobler (yet another world expert who will have no truck with group selection)…” Dawkins should consult with people before making such assertions. Hölldobler disagrees with Wilson’s strong rejection of kin selection but is perfectly receptive to group selection and understands the principle of equivalence. He represents what I have been calling the consensus of the many when he states (personal communication, cited with permission): “Almost everyone agrees that selection can also operate on the level of the colony. Indeed a colony can serve as a vehicle of genes, and one can model this by employing inclusive fitness theory or multilevel selection theory.”
Recovering the Important Questions
Dawkins is so intent on preserving his views on group selection that his review of Wilson’s book does not describe the main thesis of the book. Wilson claims that we are a eusocial species (which he defines in terms of cooperation and not reproductive division of labor), the primate equivalent of the eusocial insects, which accounts for our dominance of the earth. Mark Pagel advances the same thesis in his book Wired for Culture (17), which he frames in terms of selfish gene theory, calling human cultures vehicles of selection. The groups that serve as vehicles (as Pagel would put it) or targets of selection (as Wilson would put it) are not composed of close genealogical relatives. Heritable phenotypic variation among groups is achieved in different ways, such as norms enforced by punishment. Similar mechanisms suppress self-serving behaviors within groups, so that the balance between levels of selection tilts heavily in favor of between-group selection. It doesn’t matter that the same basic thesis can be framed in terms of multilevel selection theory, selfish gene theory, or inclusive fitness theory. It remains new and important either way.
Other issues that were originally framed in terms of group selection vs. kin selection remain important, despite the fact that they now can be addressed within either framework. Take the role of genealogical relatedness in the evolution of insect eusociality. When Wilson observes that species with high r values are common while the evolution of eusociality is rare, he is saying that high r values are not sufficient to explain the evolution of eusociality and might not even be necessary. Modest r values might be sufficient and there are even sensible models that work with very low r values. All of these possibilities can be framed in terms of inclusive fitness theory and Wilson is wrong to claim otherwise—but the issues remain interesting and important either way. If we overlook what Wilson says about kin selection, what he says about group selection deserves our attention.
Taking the Argument from Authority Seriously
In this article, I have described Richard Dawkins and E.O. Wilson as two among many who have been studying kin selection and group selection over a period of decades. I also have claimed that there is a zone of consensus of the many and that both Dawkins and Wilson are outliers who fail to recognize that the days of pitting kin selection against group selection are over.
Dawkins invoked the argument from authority to criticize Wilson’s position and to support his own. If my analysis is correct, then both Dawkins and Wilson should bring their views into alignment with the many or at least stop invoking the argument from authority to support their own views. Of course, it is possible that I have misrepresented the consensus of the many. If so, then I’m sure that many will be willing to correct my own errors.
1. Dawkins, R. (2012). “The Descent of Edward Wilson.” Prospect. May 24 2012
2. Wilson, E. O. (2012). The Social Conquest of Earth. New York: Norton.
3. Abbott, P., Abe, J., Alcock, J., & al., e. (2010). Inclusive fitness theory and eusociality. Nature, 471, E1-E4.
4. Nowak, M. A., Tarnita, C. E., & Wilson, E. O. (2010). The Evolution of Eusociality. Nature, 466, 1057-1062.
5. Sober, E. (2010). Darwin and Group Selection. In E. Sober (Ed.), Did Darwin Write the Origin Backwards: Philosophical Essays on Darwin’s Theory (pp. 45-86). Amherst, NY: Prometheus.
6. Williams, G. C. (1966). Adaptation and Natural Selection: a critique of some current evolutionary thought. Princeton: Princeton University Press.
7. Dawkins, R. (1976). The Selfish Gene (1st ed.). Oxford: Oxford University Press.
8. Godfrey-Smith, P. (2000). The Replicator in Retrospect. Biology and Philosophy, 15, 403-423.
9. Henrich, J., Boyd, R., & Richerson, P. J. (2008). Five Misunderstandings about Cultural Evolution. Human Nature, 19, 119-137.
10. Harman, O. (2010). The Price of Altruism: George Price and the Search for the Origins of Kindness. New York: Norton.
11. Wilson, D. S. (1975). A theory of group selection. Proceedings of the National Academy of Sciences, 72, 143-146.
12. Gilpin, M. E. (1975). The theory of group selection in predator-prey communities. Princeton, NJ: Princeton University Press.
13. Wilson, D. S. (1987). Altruism in mendellian populations derived from kin groups: the haystack model revisited. Evolution, 41, 1059-1070.
14. Jablonski, D. (2008). Species Selection: Theory and Data. Annual Review of Ecology and Systematics, 39, 501-524.
15. Goodnight, C. J., & Stevens, L. (1997). Experimental studies of group selection: What do they tell us about group selection in nature? American Naturalist, 150, S59-S79.
16. Muir, W. M., Wade, M. J., Bijma, P., & Ester, E. D. (2010). Group selection and social evolution in domesticated chickens. Evolutionary Applications, 3, 453-465.
17. Pagel, M. (2012). Wired for Culture: The Natural History of Human Cooperation. New York: Allen Lane.
John, Noses don’t replicate—either high or low fidelity. Noses are relevant to fitness (as are kin and non-kin groups, and lots of other things), but, ultimately what is selected are not noses, but the genes that produce noses.
There may be slightly genetically different groups of beavers that built slightly different beaver dams. But, like noses, the dams don’t replicate, the genes that produce this (extended) phenotype do.
I think you basically nailed it here. Usually when anyone from either the group selection or kin/inclusive fitness camp blasts the other side, they are usually just simply uninformed attacks on the early forms of the theory. Nowak et al. and others nuke the original r, not the current r. And when Dawkins, Jerry Coyne and friends go after group selection, they are still stuck in stone age arguments, which is pretty embarrassing as a scientist. And sadly, many assume these voices to be the most informed due to their earned capital as good scientists. That being said, the only thing I disagree with here is your quote “the entire community of scientists is in more agreement than the infamous exchange in Nature seems to indicate”. I would say the “community of informed evolutionary biologists”, which is actually a tiny fraction of the community. The majority of the scientific community has no idea what we are talking about most of the time. Obviously we make the point that many don’t ‘know’ what multilevel selection is, but what cruises under the radar, are the die hard inclusive fitness supporters that have no clue what modern inclusive fitness entails. This, amazingly, is also true for some of the 137 authors as well. One aspect of the Nowak et al. article I liked, was that they at least made this discrepancy apparent. People were upset Nowak et al. attacked inclusive fitness theory but didn’t even know why they were upset.
Thanks for the constructive comments so far. I didn’t have space to discuss the extended phenotype concept, but it also has no bearing on multilevel selection. Two examples of an extended phenotype are a bird’s nest and a beaver dam. The nest influences the fitness of an individual bird and the dam influences the fitness of a group of beavers, introducing all the problems associated with altruism and selfishness. The issues associated with group selection are confined to social behaviors and require identifying groups of individuals who influence each other’s fitness. Insofar as extended phenotypes apply equally to social and nonsocial behaviors, they don’t bear upon the issues.
On evolution without replicators, it is possible for a phenotypic trait such as a pointy nose to reappear generation after generation. Thus, what Williams and Dawkins said about individuals as collections of traits doesn’t hold for single traits. The next question is whether the phenotypic trait can reappear without elements such as genes that replicate with high fidelity. The answer, at least in principle is yes. The concept of hypercycles provides an example. In high-fidelity replication, A->A. With hypercycles, A->B->C->A. Please consult the references that I provided for more, especially the one showing that cultural evolution can take place without anything corresponding to memes.
These topics are interesting, but let’s not let them distract from the main point of my article, which is to establish the consensus of the many.
D. S. Wilson loses me here:
“Socrates might never come again as a large collection of traits, but the shape of his nose or aspects of his intellect might reoccur with the same frequency generation after generation. As long as the phenotypic distribution of a trait is replicated and responds to selection, high-fidelity replicators are unnecessary.”
Huh? That is a stretch.
” …replicators are unnecessary?” The 3 components required for natural selection to occur: replication (with heritability), variation, and non-random selection. There just ain’t no evolution without replicators.
Also, what is being selected for are the genes. The phenotypes themselves are effects of gene selection (interacting with the environment), they are not themselves replicators. Noses don’t replicate. Beaver dams don’t replicate.
The terms “kin selection” and “group selection” are confusing and deeply misleading.
There is no “kin selection” because “kin” does not reproduce. There is no “group selection” because groups don’t reproduce. There are genes that promote altruism toward kin, and there can be genes that predispose altruism toward one’s in-group.
Kin and groups are (dare I use Dawkins’ term), “extended phenotypes”—phenotypes produced by genetic adaptations that extend beyond the body to form what we call “kin groups” or “non-kin groups.”
Imagine how silly a term like “beaver dam selection” would be. Of course, there is no beaver damn selection because beaver dams don’t reproduce. But the genes that produce that extended phenotype do, and, so beaver dams may look like they evolve over time.
Asking is it beaver dam selection, kin selection, or group selection? is asking a malformed question that leads thinking down the wrong fork in the road to a dead-end path. Dawkins is correct: it confuses replicator and vehicle (or, extended phenotype).
Natural selection can only “see” replicators—it is ultimately blind to everything else.
As much as I agree with David Wilson’s article here, the whole discussion is a terrible real-life dose of Groundhog Day. By discussion, I mean not David’s commentary upon the fight, but the fight itself. It is truly tiresome. The elephant in the room is that a number of academics have their careers (and maybe other things) wrapped up in a particular viewpoint (on both sides). As a commentator said above, it’s politics, not science. The 2010 paper by Nowak et al seems provocative (albeit with intellectual value), the response from the 137 (plus others who actually wrote their own responses) was lame and did not engage with Nowak et al, and now we see it replicated (dare I use the term at all here) between Wilson and Dawkins. And once again DS Wilson tries to point out that there are areas of agreement, but too many are too busy kicking sand in each other’s faces.
Dawkins review is nothing but a polemic, and it just regurgitates what he has always said. There is no intellectual movement or engagement from him at all. I’m not actually sure the review warranted any attention.
What is really needed is someone to provide a nice in-depth resolution of the two positions—actual meaningful points of disagreement, along with points of agreement, as can be found here.
For example, it seems to be that whereas Nowak et al argue that other traits (such as communal nest-building) need to be in place BEFORE eusociality can arise (which is likely to depend on kin effects), the traditional IF view is that kin effects will facilitate the cooperative dispositions, which rollup together traits such as nest-building along with eusocial reproductive strategies. This begins to get to the heart of the problem I have with kin selection vis-a-vis MLS: kin selection in most cases is simply capturing a way for those with similar traits to aggregate (such as daughters staying with their mother’s nest and helping her reproduce further). But if altruists can spot altruists, then kin aggregation isn’t a limiting factor. Yes, Greenbeards. With humans at least, we all ‘wear’ such greenbeards—it’s in our behaviour. If you spend a little time around me, you can begin to evaluate how altruistic I am. It won’t be perfect but it’ll probably work quite nicely.
Michael Mills. You have selectively quoted DSW. You quoted ”—- replicators are unnecessary”. In this instance the “—-” is vital to the argument. The missing word is “high-fidelity”.
Nose shapes do replicate in families; but with relatively low fidelity. Nose shapes will indeed be relevant to fitness – and therefore selected – in certain environments. This is immediately apparent from a comparison of nose-shapes between geographically distinct human sub-populations.
I suggest that the assumed need for high-fidelity is the crucial mistake made by some evolutionary theorists when discussing aspects of multi-level selection.
I think this is a very good description of the state of affairs in the ‘kin selection’ vs. ‘multilevel selection’ debate. There is no clear distinction that separates the two. Kin (a kind of group) and genes are levels of biological organization above and below the individual organism, after all. It might be difficult to end a debate without declaring a loser, but I hope we can move on soon.
137 authors on a comment to a Nature paper? Sorry, that’s politics, not science. In experimental high-energy physics, yes, collaborations can include hundreds of researchers, making author lists bloom to multiple pages, but this is not a CERN team hunting for the Higgs boson. Such a manifesto, weighted down by signatures, might be helpful when the issue itself is political, like convincing people that climate change is real and anthropogenic. But it’s a shoddy way to do science.
I went through all the responses to the Nowak, Tarnita and Wilson paper (no doubt this speaks to some masochistic streak in me). My chief impression was that most everybody involved in writing them hadn’t read the “supplemental” information to the paper, which was (a) long enough to be an article by itself, (b) where all the mathematical content was, and (c) free to read online. When a depressingly politicised area of science meets poor publishing practices, up surges a perfect storm of miscommunication.
I would argue that this entire year-in, year-out fracas narrows the intellectual horizons: by arguing over how best to slice and dice Hamilton’s rule, one naturally grows to think that such a rule is evolutionary theory. But there are more things in heaven and earth than are dreamt of in your r, b and c.
P. Bijma and M. Wade (2008), “The joint effects of kin, multilevel selection and indirect genetic effects on response to genetic selection” Journal of Evolutionary Biology 21, 5: 1175—88. DOI: 10.1111/j.1420-9101.2008.01550.x
J. D. Van Dyken et al. (2011), “Kin Selection–Mutation Balance: A Model for the Origin, Maintenance, and Consequences of Social Cheating” American Naturalist 177, 3: 288–300. JSTOR:10.1086/658365
J. A. Damore and J. Gore (2012), “Understanding microbial cooperation”. Journal of Theoretical Biology 299: 31–41. DOI:10.1016/j.jtbi.2011.03.008. PMID:21419783.
B. Simon, J. A. Fletcher and M. Doebeli (2012), “Hamilton’s rule in multi-level selection models” Journal of Theoretical Biology 299: 55–63. PMID:21820447.
Just because two theories make the same predictions, it doesn’t mean that they are equally good. There’s also Occam’s razor, convenience, and other factors to consider. Kin selection is widely-accepted orthodoxy. Group selection is a controversial theory with a long history of associated confusion. The fact that it adds nothing to inclusive fitness theory is hardly much of a point in its favour.
Kin selection work typically emphasizes close relationships – which is highly appropriate. Group selection work typically emphasizes more distant and numerous relationships – which is not appropriate at all. Kin selection seems to be better. No wonder it wiped out group selection as an explanation for altruism decades ago.
Nowak-Tarnita-Wilson’s main points are:
1. Hamilton’s Rule only works if B and C are NOT B and C in the standard payoff matrix and R is not the standard genetic relatedness. (I’m not sure whether they realised that many of the Inclusive Fitness people had understood this some time ago, but generally kept it rather quiet).
2. You can only find out what will (probably) happen by doing proper Evolutionary Dynamics calculations. These will tell you a lot more than “Hamilton’s Rule” or “Inclusive Fitness”. And once you have done these calculations you don’t need the “rule” except as a rule of thumb/accounting method.
Indeed you can’t even calculate B, C and R in general without doing the proper ED calculations.
I am not convinced in group selection nor in kin selection. I am convinced only in self selection. Since I suppose that brain is the part on which evolution works. Brain is very selfish in its function. Genes are responsible only for the structure of the brain. That structure is built during evolution to promote it’s own benifit. This own benifit can occure in different shapes like the so called “kin selection” or the other “group selection”. Even the desier to inhance reproducibility would be stoped if the brain came to a conclusion that this trait is against it’s benefit without any connection to spreading genes.
Tim: MLS and IF aren’t just two theories that make the same predictions; mathematically they are considered equivalent (aside from Nowak et al’s point that there are specific cases where IF falls short). Kin selection as commonly used and popularised relates to close relationships, but Inclusive Fitness Theory, as currently refined (and which is the real theory here, not the subset kin selection framework), is not limited to close relationships, because r is no longer limited to genealogical relations. IF makes the same predictions about altruists helping other altruists as MLS does. And MLS applies to close relationships as much as more distant relationships.
@Tim. I’m afraid you have some of that backwards. Yes, r, is now shared phenotype, because that is what matters for selection. With MLS/group selection, this has always been the case. The Price Equation clearly demonstrates how it is the partitioning of variance at the within and between group levels that matters to selection. Not necessarily shared genes. This is why Hamilton updates his theory, and why IFT has been further revised to reconcile this oversight. MLS/group selection has since still relied on the fundamentals of the Price Equation, that selection acts where the variation is. So MLS was/is/will be a general framework as kinship, kithship, punishment [which reduces which dynamically reduces group variation], and assortative interactions are just ways to influence the partitioning of variance. When r took on serious overhauls, I would say that is where IFT ran in to some issues. Certainly, the updating of r made IFT a general framework but not everyone followed this change. Very few know of it. In fact, few knew Hamilton even wrote other papers since then. The intuitiveness/catchyness of Hamilton’s theory was with the idea of the inclusive fitness effects from shared genes, only he later realized the importance was where variance was partitions. If after I read Hamilton 1964, someone told me, btw, that r term, yeah that can also mean cultural relatedness, because r as shared genes doesn’t jive so well. I would have said but wasn’t the shared genetic aspect that the point? When that concept of r changed, many did not follow, and some [eg. Nowark et al.] rejected this alteration altogether. I for one accept these updates, but it is not hard to argue that once the fundamentals of a theory becomes overhauled, it is not just an update, but a new theory [consider kin selection theory as coined by Maynard-Smith and what that term means now]. So in other words, IFT essentially became MLS. At that point came equivalence, and complete niche overlap leading to heavy competition. This is simply an objective account of the history. So using your logic, IFT should essentially be dropped as that was a specific framework which expanded under the influence of Price and his covariance perspective. Again, I agree with the IFT approach when considering what it is now, but Nowak et al. had a point, r is “something” but not what it originally was, and it can mean different things to different people and in different scenarios, and expanded in its definition to become more general. This was really the first time Kin selection/IFT got taken to task and it elicited a knee jerk response, and caused other to be upset when some of them do not even know what “r” is. Partitioning of phenotypic variance, however, is very intuitive and was there from the start, this is why Hamilton had his AHA moment [which many have no idea of]. So if you want group selection people to distinguish MLS from Kin selection, well thats easy. Keep r as genetic relatedness and MLS will keep with variance partitioning, and problem solved. But I do not see how MLS needs to distinguish itself from kin theory. I do not care that Kin selection/IFT and MLS are equivalent, as in, that doesn’t bother me so much. I use MLS because to me, it is a more intuitive way to observe selection, while IFT may be more intuitive to others. All I care about is that people have an understanding of the frameworks. The field of social evolution is plagued by those that rebrand old ideas as new, strong opinions backed up by little to no knowledge of the topic, and poor scholarly efforts to read the background literature. Tim, I am not accusing you of this, it is a more general comment to why there is so much beefing in this field.
This dispute is largely a reflection of the sociology of science as a human endeavor—in a sense, it’s politics as an earlier commenter said. It’s about egos, struggles for attention and credit, group selection (that is, competition among groups for preferential recognition of their point of view) and so on. There is as usual in such situations a variety of concepts and terms that are used in various ways by different authors, and changeably with individual authors, and so on.
EOW’s book is heavily rapped in the current NY Review of Books for slipshod human history evidence, but EOW’s sloppy scholarship is not itself the point except to the extent that it directly vitiates his generalizations as regards human nature.
DSW’s long discussion here is very informative and measured. I agree with the comment that the vote-count of people signing on to a view, even if to a ‘middle ground’, is not necessarily relevant to truth, so from a scientific point of view it doesn’t really help to say that there is a kernel of consensus.
In a sense little of this is actual science, and it’s mostly discussion of personal views of the world. That’s because all of these theoretical positions require a certain rigidity, long duration, or replicability of cause—of success of individuals or groups based on a particular trait. The models are not really testable in a serious way: the math may be correct if the conditions pertain, and it may be possible to retrofit aspects of the model to presumed scenarios, but we know that many scenarios can fit the same data to different aspects of theory comparably well, so that’s not the same as proof of their truth. The models tend to be categorical when llfe may be much more fluid.
DSW’s ideas make room for lots of nuance and variation. The basic idea seems patently true, because whatever succeeds preferentially succeeds, and there are many paths to success, and we know they all are relevant. He points out something else that may be quite important.
If we fixate on individual narrow traits (like pointy noses) and Mendelize them as being due to single alleles, then things may seem quite simple and fit to a formal theory. Mendelian segregation led to phenomenal research strategies but confounded inheritance of traits with inheritance of genes, which has been a problem we’ve had in genetics ever since. Many or most traits are not monogenic, and the individual contributing alleles mainly have very low penetrance, but nothing stops one treating the trait essentially as if it were monogenic. One can ask what happens to a ‘gene for altruism but is altruism due to ‘a gene’?
As I understand his post, DSW discusses recurring traits in the population distribution of trait values to help get past the hypercategorical arguments at play in this dispute. That’s a useful way to think about complex traits of which altruism and its like would be exemplars.
If traits of this sort are due to the effects of many contributing genes, each varying among contemporary individuals (and over time), and most alleles having little individual effect, then we can develop a better understanding in terms of phenotypes than genotypes. The individual alleles come and go, but if there is some advantage, whatever its source(s), in being toward one tail of the distribution, the distribution can move in that direction in a way that favors the group or individual without the net fitness effect at any one contributing gene being very tight or precise or even detectable in any serious way at any given time.
If being willing to help the contemporaries we might meet, even at some personal expense, is polygenic, then no two rescuers have the same rescuer genotype, even if the trait itself becomes more common. There need not even be a cost to the rescuer for his actions. Maybe the same phenotypes make better warriors as well as helpers (multilevel effects!). One needn’t be able to attribute the changes in the population distribution to any particular gene, even if the net genomic change in aggregate will have changed in a direction compatible with the trait distribution.
In principle we shouldn’t need to be so rigid in our models, or our disputes….but maybe we do, because we’re human and we evolved to do it! We’re Homo disputensis.
It’s nice to see some slightly more constructive discussion of these issues, now that some of the original hysteria (on both sides) is subsiding.
Rick O’Gorman: the ‘lame’ response by the 137 authors (myself included) was space limited, and concentrated on the claims of NTW that inclusive fitness has no empirical support (table 1) and has made no predictions (table 2). Limited space was available to deal with the myriad technical problems. The same holds for several of the other Nature commentaries published at the same time, many of them by experimental biologists. The focus was on brief comment in the original venue.
On the other hand more detailed technical commentary has been published, e.g. Lion et al. TREE 2011, Gardner et al., J Evol Biol 2011, Marshall, TREE 2011, Bourke, Proc B 2011. One point that almost all the commentators (myself included) missed out on was what was highlighted by NTW reply, and by a careful re-reading of their original paper – they think that the neighbour-modulated version of inclusive fitness (a mathematical convenience to aid analysis) is actually the same as classical Darwinian fitness, i.e. Darwin’s theory never needed to be extended, and Hamilton’s gene’s eye view was just a tortured reconceptualisation… this is a serious misunderstanding of the history of evolutionary thought.
@Tim Tyler. I do not want to single you out as an example of the lack of understanding the two theories but your response is certainly loaded with misconceptions of both. Kin selection/IFT is not about close family ties while MLS are the other cases. Both, are about how variation is partitioned in the population. IFT [as originally coined by Hamilton] uses the term r to demonstrate this. It was originally envisioned that shared genes would have an indirect fitness consequence to the actor by going disproportionately more to its shared genes in others. This had since been updated, not by the MLS folks, but by IFT proponents for various reasons [genetic relatedness does not necessarily mean phenotypic relatedness and vis versa. Relatedness is just one way to partition variance, while assortative interactions, culture, punishment and other mechanisms can do the same, if not even more strongly]. Thus the modern variant of, r, is actually a measure of how [phenotypic] variance is partitioned at the group vs within group level. This meaning of r is computed from the Price Covariance Equation, as you should know is the formal equation for MLS/group selection. MLS and IFT vary only in their approach to show the effects of the partitioning of variance. MLS shows the direction and strength of selection at the multiple levels on a particular trait. IFT more or less conceptualizes the population structure in r, showing the effect a trait proportionately has on itself due to the distribution of like vs unlike phenotypes. In other words, you can say an altruistic trait is disadvantageous within groups, but advantageous at the group level, and the relative strength of selection [partitioning of variance] at these various levels determines the balance of these forces, and the net effect. OR you can say, an altruistic trait does well if the positive effect it has on itself is increased because these effects are going proportionately more to other altruists [itself so to speak] because there is increased group vs within group level variance. Thus, the frameworks are entirely equivalent and uber general. What happens between groups matters. Whether you want it in pill form or the syrup doesnt matter. A perfect example of how one can translate back and forth between frameworks is my paper [the role of multilevel selection in the evolution of sexual conflict 2010 Evolution] which is explicitly in MLS terms, and its companion paper [Sexual conflict in viscous populations: The effect of the timing of dispersal 2011 Theoretical Population Biology] by IFT folks who do the same thing but use IFT terms. I recommend you read up a bit more because your perception of both frameworks and the history of the field is off. That being said it is 2am here so I hope this makes sense and I will leave it to others to add to my comments.
Adding to Tim and Omar’s comments, in the sexual conflict example, the variation among groups is caused by behavioral sorting and not genealogical relatedness. The major thesis advanced by E.O. Wilson and Mark Pagel in their respective books involves human groups in which genealogical relatedness is low (even in small hunter-gatherer groups) and the balance between levels of selection is determined by other factors (norms, policing, etc.). Thus, the claim that the only important cases of group selection involve groups of close genealogical relatives cannot be sustained in my opinion.
Hamilton’s Rule is not the same thing as “inclusive fitness theory” it is a built on a simple model of inclusive fitness that illustrates the mechanisms. If Wilson (and co-authors) just said “the assumptions needed to derive Hamilton’s rule don’t hold in most cases” I don’t think many people would argue with that. But instead they take that well-understood observation (amongst theorists) and use it to argue that inclusive fitness theory is wrong.
To those that are annoyed that the 134 co-authors did not address the math in the appendix: the math in the appendix is besides the point.
Omar, it’s not really correct to think of kin selection as being
based on shared phenotypes – as you say by claiming that “r, is now shared phenotype”. Kin *recognition* is based on shared phenotypes – but there’s more to kin selection than kin recognition.
In particular, maternal care is not given based on shared phenotypes. The baby of the lady down the road looks much the same and behaves much the same as the mother’s baby – but gets practically no maternal care from the unrelated mother. The mother is going by beliefs about genetic relatedness, *not* clues about phenotypic similarity coming from the baby’s phenotype. Cuckoos live by exploiting this difference.
As for Nowak, he fairly clearly endorses cultural kin selection in the “Tides of tolerance” paper from 2001 – saying: “So, the mechanism that leads to cooperation is a form of kin selection — either classical (if traits are inherited genetically) or social (if they are inherited culturally, like a dress code).”
FWIW, I’m well aware of cultural kin selection – as you will rapidly see if you Google the term. However, it is based on heritable variation – just like all other forms of kin selection are. It’s based on shared *memes* – not shared DNA genes. However, it is pretty bad practice to bundle genetic and cultural relatedness together into a single ‘r’ – because of “meme quarrantine”. ‘r’ between symbionts – including cultural symbionts – is not the same as ‘r’ between hosts. Symbionts are related primarily to *other symbionts*. There are cultural kin (sometimes called ‘kith’) – not just kin based on shared DNA – and I never suggested otherwise.
Group selection proponents should work to distinguish their theory from kin selection, IMO. Equivalence between the two theories is not good for one of them. Imagine you invent the “oogolrome” – only to find that people tell you that you are just renaming the wheel. At the momemnt, group selection is looking a lot like a oogolrome. We know it is equivalent to 30 years of orthodoxy – what is less clear is what makes it worthwhile.
Applying the idea to family groups – where relatedness is significant is what the kin selection folk do. Applying the idea to all manner of groups – such as companies, tribes and nations – where relatedness is often much lower – seems to be what the “group selection” and “good of the group” folk do – and it is often folly. The long history of this is what has given group selection a bad name. Group selection had better have some impressive redeeming features if it wants to be taken seriously, after all these decades of causing this sort of problem. I can see some redeeming features – but it isn’t clear whether they compensate for the decades of muddle, confusion and bad models that it has been responsible for.
At the moment, if all the fans of group selection and multi-level selection would switch to using kin selection / I.F.T. a whole bunch of nonsense involving cases where relatedness is low would rapidly go up against the wall.
Incidentally, the idea that I.F.T. is more general than kin selection isn’t worth too much, IMO. When genes are identical, in practice, they are identical by descent. Genes being identical by chance – without descent being involved – is so unlikely as to not be worth bothering about. So “kin selection” is a perfectly general term.
Modern inclusive fitness is perfectly fine. What Nowak et al. did was misguided in their representation of inclusive fitness theory. Very similar to how group selection gets misrepresented by many, including some of those 134 authors. I am assuming that Rick, considered it as “lame” because of 1) how it was a signed paper 2) many of those authors misrepresent group selection all the time so it is a bit hypocritical 3) when group selectionists respond to their framework being misrepresented, their responses are often resisted against. Many group selectionist rallied against that Nowak et al. paper because having your theoretical framework misrepresented is wrong, not scholarly, and just sucks. I would at some point love to see that sort of vigilance from the inclusive fitness side for papers that blatantly misrepresent group selection such as “Sixteen common misconceptions about the evolution of cooperation”. In any case, we all study evolution so are well versed in what comes from competition. I like that notion of “Homo disputensis”.
I think when Rick meant the 134 author paper was “lame” [or maybe these are just my thoughts] he wasn’t necessarily making a point about the content of the paper but other aspects of it. Such as 1) authoring papers as petitions is not the way to go. 2) because of this, some of the authors on that paper do not even study MLS or IF/Kin selection and are barely even tangential to the field. 3) The response was to criticize Nowak et al. for misrepresenting a framework [which is fine with me] BUT some of the authors are guilty of doing this very same thing when discussing group selection, and to me, this hurt the paper and made it seem hypocritical. I certainly have issues with how Nowak et al. represented IF/Kin selection theory and many of the MLS proponents joined with the IF/Kin folks and stated their dissatisfaction with the article, hence this very blog. What is troubling is how many times MLS/group selection gets misrepresented and it is business as usual, and response articles are often met with resistance. I am not accusing all the 134 author papers of this by any means at all, but coming from the group selection side of things, I was like oh someone misrepresented your framework in Nature? That happens to me so much I am confused what Group Selection is. Either way, we all must be knowledgeable of our field and take better care of our scholarship. Misrepresenting theories seems rampant in the field of social evolution and it is embarrassing for all of us, no matter what framework one prefers. Hopefully it improves.
oh no, I pulled a Ken Weiss.
Thanks not only for the well-done ETVOL-exclusive post, but also for the comments, which I read with interest and delight.
I’d love to read a piece by Martin Nowak on the matter whose “SuperCooperators” impressed me deeply (and made me recommending it both on my German & English scilogs). Is there amy chance to ask Nowak if he could do a post on ETVOL, too?
When Dawkins says “group selection” – he uses the term to refer to “interdemic selection”. He doesn’t approve of redefining “group selection” to refer to kin selection, saying:
“The so-called “new group selection” is just kin selection or in some cases reciprocal altruism under another name. For reasons best known to himself (which I can’t understand) D. S. Wilson thinks it’s helpful to rephrase it in terms of group selection. How it can be helpful when he’s reviving a word which has been debunked and is simply grafting that word onto the very thing that did the debunking – namely kin selection and reciprocal altruism and various other things – it seems to me be to be utterly unhelpful, to be totally misleading to students and it’s deeply regrettable that E. O. Wilson should have teamed up with him in this way.”
To James: Thanks for your informed reply. One of my main points is that the “people working at the cutting-edge of the field”, as you put it, HAVE reached a consensus, so we agree on that point. I certainly regard Stuart West and Andy Gardner as part of that consensus.
Another main point of my article is that the evolutionists who loom so large in the public eye need to abide by the consensus (or provide compelling arguments to the contrary) and that there should be symmetry in this regard. Dawkins’ outdated views should be called out, along with Wilson’s. In your message, you treat it as common knowledge that “Dawkins is not active or even up-to-date in the field”. Then why isn’t he criticized by his own colleagues, as Wilson was?
The same goes for claims that a given framework isn’t useful. What Nowak et al. said about inclusive fitness in this regard is much like what West and Gardner said about multilevel selection. In both cases, it’s offensive to those employing the other perspective, who publish in the very same peer-reviewed journals. A good test is whether a given result derived from one perspective is also new and interesting from another perspective. Omar’s work on sexual conflict, derived from a MLS perspective and recently translated into a IFT perspective by West and others, provides an example. The main thesis of Wilson’s “The Social Conquest of Earth” provides another example.
The main benefit of clarifying a “zone of agreement” is to focus on a “zone of legitimate controversy”. There’s lots to disagree about, including the concept of equivalence at an advanced level. I’d like to think that this article, along with others that have appeared in ETVOL, has helped to clear the air so we can concentrate on the current frontier of controversy.
@JamesMarshall: David Wilson has eloquently stolen my point here, but suffice to say that Williams is in my opinion conveniently revising what he said in his 60s book. More importantly, many others have used his book to say that GS is irrelevant. And what’s worse is that Williams’ argument in that book is just a (bad) thought experiment, no maths, no data. Yet on the back of that GS was suppressed for years as a legitimate viewpoint.
@TimTyler: This is getting old. It doesn’t really matter what Dawkins says, he’s wrong. Even the most biased proponents of Inclusive Fitness, such as Stuart West and Andy Gardner, acknowledge that Inclusive Fitness and MLS are equivalent. So if Dawkins wants to define group selection as ‘selection between groups except kin’ then so be it. But it puts him at odds with the leading people on this issue. Aside from Nowak & EO Wilson to some degree, who also define group seleciton in a way that retains kin selection.
@Omar: Thanks for interpreting my ‘lame’ comment. It was a bit lame itself. But yes, I think that it did not take 137 of them to write the article! They are signatories, and thus it is an ideological point.
Whoops, in my last post I should have said: Aside from Nowak & EO Wilson to some degree, who also define group seleciton in a way that retains kin selection as a separate process.
West and Gardner clearly acknowledge that group selection has had multiple definitions in practically every paper on the topic that they write – typically calling them “new group selection” and “old group selection”. Dawkins too, used the term “new group selection” in my quote from him above. Everyone knows about the new and the old forms, but part of the group selection muddle arises when people mix these concepts together.
@TimTyler: On your last point I think we can totally agree. What I would point out is that those who are familiar with MLS are generally quite clear on what is group selection. And it is actually questionable what new v old means. This has been invented in my view primarily so that people who have been gung-ho on Inclusive Fitness can save face and accept that MLS is legit. The main problem with ‘old’ GS is that there was an assumption that GS could overcome individual selection (so sometimes its referred to as ‘naïve group seleciton’). That then flipped around with George Williams who suggested (with little evidence) that individual selection will almost always undermine and overpower GS. The reality (‘new’ GS) is that it just depends on the selective pressures. Though no-one has yet referred to the outgoing Williams dogma (that IS will always trump GS) as naïve. Which to be fair, it seems we now should.
@O’Gorman: I think ‘naive’ group selection arguments were ones of benefit to the group/species… i.e. they didn’t even consider the possibility that selection at the within-group level might undermine selection at the higher level. This is what G.C.Williams reacted against. Interestingly, the Nowak, Tarnita and Wilson model of the origin of eusociality appears to be precisely such a ‘naive’ group selection model, in that the strategy set for daughters excludes the free-rider option; that is, virgin daughters can either undertake a nuptial flight and try to found a nest (risky) or stay at the home nest and raise their mother’s offspring. The third strategy, amply demonstrated e.g. in honeybees, is not modelled; namely to stay in the safety of the home nest, unmated, and produce male sexuals (with a direct fitness benefit). NTW’s model is thus ‘naive’ group selection because it considers the queen/colony to be the exclusive unit of selection and ignores the levels below; the language they use is very telling (e.g. workers are described as ‘robots under the queen’s control’ or similar).
@O’Gorman: PS perhaps best to let GC Williams defend his reputation (from the preface to the ‘96 edition of Adaptation and Natural Selection, p.xii) – “it also became fashionable to cite my work […] as showing that effective selection above the individual level can be ruled out. My recollection, and my current interpretation of the text […], indicate that this is a misreading. I concluded merely that group selection was not strong enough to produce what I termed biotic adaptation: any complex mechanism clearly designed to augment the success of a population or more inclusive group.” The question of group-level adaptations, such as we see in the eusocial insects, is an interesting one… see, e.g. Gardner and Grafen’s ‘Capturing the Superorganism’, and Okasha and Paternotte’s critique.
Here is where good scholarship becomes important. If we’re going to talk about “old” and “new” we need to review the literature. Darwin, Wright, Fisher, Haldane, Williams, Maynard Smith, Price, Hamilton, Wade, and lesser-known figures all wrote about group selection in words and formal models. This was not a tight-knit community, so a given author often developed his or her ideas independently of the others. We need to examine their work to find common denominators and to see if there was indeed a discontinuity that can be called “old” and “new”.
Elliott Sober and I did this in Unto Others (1998). For those who distrust our objectivity, there are other books by Samir Okasha, Oren Harman, and Mark Borello that broadly reach the same conclusion. The various authors conceptualized groups and multi-group population structures in different ways. The haystack model of Maynard Smith is different than the permanently isolated demes imagined by Sewall Wright or the “tribes” imagined by Darwin. What the models all share in common is this: traits that are “for the good of the group” are selectively disadvantageous within groups and require the differential productivity of groups to evolve in the total population.
The way that groups are defined in the Price equation provides a telling example. The Price equation is so general (that’s why people like it) that it can accommodate many specific conceptions of groups, including ephemeral trait groups and more durable multigenerational groups. Hamilton and Price didn’t care about the specific conception of groups and certainly didn’t say that some but not others quality as group selection models. When Price asked Hamilton “Have you seen how my formula works for group selection?” He was pointing out the within and between group components for any kind of group, and that is how Hamilton saw it also.
There is no single arbiter for defining group selection, because the community of scientists was never sufficiently close-knit to agree upon a single definition. This is true for most major topics in science. Good scholarship is required to find the common denominators.
Is there any evidence for a discontinuity between an “old” and “new” form of group selection? Not by my estimation. In Chapter 2 of Unto Others (titled “A Unified Theory of Social Behavior”), Elliott and I describe some of the early models by Wright, Williams and Williams, Maynard Smith, and Price. The model of Williams and Williams, published in 1957, assumes that groups are ephemeral family groups. Anyone would regard this as a kin selection model today, but Williams and Williams borrowed it directly from Wright’s 1945 model.
Why do folks such as Dawkins, West, and Gardner talk about an “old” and “new” group selection if there is no historical evidence for it. In my opinion, this is a convenient way to claim that the wholesale rejection of group selection in the 1960’s was correct and that what passes for group selection today is different. While I am always careful to acknowledge the existence and error of naive group selectionism, the idea that group selection is defined differently today than in the 60’s cannot be justified by the scholarly evidence, and anyone who wishes to claim otherwise needs to develop a strong scholarly case for it.
This brings us to the subject of historical revisionism—the re-writing of history to justify current favored positions. We need to be vigilant about historical revisionism in science in addition to politics, especially since since we are all unintentionally prone to it, requiring a social process of holding each other accountable through good scholarship. When Dawkins says that Darwin refers to group selection in only one anomalous passage, he’s wrong and must yield to a more scholarly account, regardless of whether his revisionism is intentional or unintentional. In the same way, distinguishing between an “old” and “new” group selection must be justified by good scholarship; it can’t just be asserted. It doesn’t matter that Dawkins, West, and Gardner have said it consistently—what’s their evidence?
(continued in next message)
(continued from last message)
Maynard Smith’s haystack model provides an exceptionally clear example of poor scholarship and historical revisionism. It played a large role in defining group selection during the 1960’s. Folks such as Dawkins still regard it as a legitimate group section model and cite it as an example of how group selection doesn’t work. What they don’t say is that the haystack model DOES work, when the same assumptions about altruism that Hamilton employed in his inclusive fitness theory are used. I established this in 1987, in an article published in a top journal (Evolution), so you’d think it could not be ignored. Yet, ever since, Dawkins has continued to describe the haystack model as an example of how group selection doesn’t work, simply ignoring the subsequent literature.
I don’t want to dwell excessively on Dawkins. The point of my article is that EVERYONE needs to be held accountable. There is enough good scholarship and clear thinking to establish a consensus, which will enable us to move on to consider the substantive issues, using multiple perspectives and knowing how to translate among them.
@DSW: perhaps conflating West and Gardner with Dawkins is a bit confusing… as you point out Dawkins is not active or even up-to-date in the field. W and G however understand the equivalence that Price and Hamilton remarked on using Price’s equation; W and G’s criticism mainly seems directed at two things: i) claims that GS/MLS is a different theory to IF, and ii) the potential confusion of group selection arguments (‘Social semantics: how useful has group selection been?’; written as a reply to your commentary on their original article in JEB). I think new vs old group selection is usually used as pre-Williams vs post-Williams… in other words ‘old’ is used in the same way as what I interpret to be your (and my) use of ‘naive’ group-selection. I’m not aware of the people working at the cutting-edge of the field making the errors you describe Dawkins making with respect to, for example, JMS’ haystack model. If there are examples on either of these two points I would of course be happy to hear of them. As you know Okasha’s 2006 book goes into forensic detail over the various claims made and positions held/modified as part of the GS/MLS developments.
@DSW: I look forward to reading more of your thoughts on the ‘legitimate zone of controversy’
Regarding the individuality agenda, my interpretation is that these authors want to nail down what it means for an evolutionary entity to be adapted. It may be of interest also to note that Andy Gardner has recently done some work on formalising selfish gene theory (A formal theory of the selfish gene, JEB, 2011).
There are plenty of resources out there documenting that the “old” form of group selection was quite different from the “new” form. To give an example, here is J.M.S. 1975 “Group selection” paper:
“It is useful to distinguish as sharply as possible the processes of ‘kin’ and ‘group’ selection. The terms “Group selection” should be confined to cases in which the group (deme or species) is the unit of selection.`
There are numerous other resources along similar lines – there can be no doubt.
@TimTyler: Jeez Tim, can you not pick up from what D. Wilson said that JMS might not be quite the best person to cite in support of the idea that kin and group selection are different? Or that maybe you need at this point to just go read the (recent) literature and stop trolling.
@DSW: yes, I agreed with the substance of your original article, and with much of what you write now. No-one should be above criticism if they misrepresent the theory/facts. Why is Dawkins not criticised as Wilson is? Interesting question; both have intriguing similarities in having popularised Hamilton’s ideas around the same time… I think they key difference is that Wilson is now a vocal and prominent critic of IF theory, so the IF community have responded. Dawkins might make some conceptual mistakes, but doesn’t say IF is a load of rubbish, hence these are largely ignored. That’s why I appreciate the substance of your article, which is quite impartial.
@O’Gorman: well, I see some (but not much) maths in Williams’ book, as well as reviews of evidence from published studies (i.e. data). Anyway, I don’t think well-read people in the field use Williams to say MLS is wrong… quite the opposite from my own usage (see the introduction of my 2011 TREE paper, for example) and my discussions with people like Kevin Foster and Andy Gardner. If less informed people mis-use Williams book in the way you describe that’s not the fault of the book, but those who are using it, and that can happen to -any- text and has done so repeatedly throughout history. I think the reasoning in Williams, as a critique of naive-group selectionism (i.e. Wynne-Edwards type thinking) is sound… you seem to disagree; does that imply you think that naive group-selectionism is logically sound? In having these exchanges it’s -really- useful to explicitly differentiate between the different theories rather than just say ‘group selection’, hence here I use MLS to mean the modern, IF-equivalent, incarnation of the theory. It’s also helpful to be aware that group-level adaptations aren’t easily explainable by vanilla IF/MLS theory. The potential for confusion between different variants of the concepts is at the root of e.g. West and Gardner’s critiques of the conceptual approach (whose equivalence they do not deny), Okasha’s forensic examination, etc.
PS @DSW: yes I also agree that we definitely need to move on beyond the decades old aspects of the debate on MLS/IF… Lion et al. TREE 2011, ‘Beyond the kin versus group debate’ offer some nice ideas about things that -are- worth looking at.
@JamesMarshall: I think the maths in Williams’ book (from memory) isn’t related to the discussion on GS. What is there isn’t worth dwelling on (given modern levels of analysis).
It doesn’t happen so much any more, so perhaps I am having a go at a past sin, but in not-so-distant discussions (in the lit) on cooperation, for example, Williams has usually been cited as a reason why GS can be ignored. Perhaps you read different lit. But your own quotation makes clear that Williams himself was aware that his 1966 book was used regularly as an argument against GS. If you read stuff on cooperation, there are repeated cases of citations of Williams as to why GS doesn’t need to be considered. And I don’t know how uninformed you consider people like John Tooby and Leda Cosmides to be, but they’re a good starting point (but again, this doesn’t really happen so much now in print, though see a paper by Omar Eldakar for how it still happens with reviewers—I’ve experienced it myself first-hand).
“you seem to disagree; does that imply you think that naive group-selectionism is logically sound?”
To be honest, if you’ve read what I wrote above in full, I don’t see how you could in good faith ask me that question. I thought you were being reasonable in your recent post but I am usually too generous in that regard. But to answer…Williams picks as thought experiments (again from memory) rather odd examples. And he essentially argues that GS is very likely to be irrelevant, based solely on his verbal arguments. I think that’s what I said before, I don’t see how that says I think that naïve GS is valid. I find Williams 1966 argument against GS as compelling as prior arguments in favour of (naïve) GS.
Thus, I do find Williams 1966 at fault for what it says, and I do think that Williams has done a bit of revisionist history later on. In fact, what he writes in your quotation is a bit disingenuous in my view—he’s trying to say that he didn’t take a position against GS per se, but that it won’t lead to ‘biotic adaptations’. So what does he think it can lead to, and why isn’t that mentioned in his 1966 book, because it really isn’t? Perhaps this is semantics, that GS can lead to some shaping of a population’s trait(s) but not out-and-out group-level adaptations? How do we draw that line? If cooperation can evolve by kin selection, then it evolves by GS (within MLS framework). So is it that Williams (and you) don’t want to call that a ‘group-level adaptation’?
Elsewhere, I have used MLS to refer to the framework, and GS to refer in shorthand to group-level selection (contrasting with individual selection, which I didn’t abbreviate to IS for clarity). As I mostly wasn’t debating with you, and was debating the issue of kin selection v GS, I thought it was clear enough in the context.
@O’Gorman: it was an honest question, since I’m confused about the logic of your position, and whether we’re interpreting the same words in the same way… if we’re to have productive discussions on this, and this blog is to engage IF people (which I think is the original intention), I think questions need to be answered in the spirit in which they were asked, rather than giving the impression that contributions from outsiders are merely to be generously tolerated!
Anyway, I think I now understand your position to be ‘Williams was criticising (modern) MLS all along, with unsound logic, then pretending he wasn’t’… my reading is that Williams was criticising naive-group selection all along, with sound logic. If that’s an accurate summary we’ll just have to agree to disagree. I’d note that Williams was responding to his critics in 1996… I think the field has moved on a lot in the last 15+ years.
The group adaptation versus group beneficial trait is precisely what I was trying to get at. I don’t know if you’ve read Samir Okasha’s excellent book, or the papers it’s based on, but he precisely tries to disambiguate whether altruism in the vanilla IF/MLS framework (i.e. trait-group models) is a group-level adaptation or not, and concludes that it is not (for reasons I won’t even try to do justice to here). Okasha discusses the evolving positions of DS Wilson, and others, on this at some length… it would be interesting to get David’s input on his position about group-level adaptations versus group-beneficial traits now… (or a good recent reference).
A really fascinating and important topic now, which the original article above touches on, is the issue of organismality (vehicles), Queller and Strassman, West et al., etc have written about this in the last few years… your body (for example) is much more than a collection of bacteria producing some public good like siderophores (which could be explained by MLS/IF)… quite apart from the extreme reproductive division of labour between cell lineages (which can be explained by MLS/IF) there is the issue of the amazing specialisation of different cell lineages to form complex, interconnected organs… these are the kind of group-level adaptations Williams was talking about. While MLS/IF might help some way here (Darwin’s ‘well-flavoured vegetable’ example in the evolution of different worker morphs in social insect colonies) my impression is there is more to it than this… hence I cited Gardner and Grafen’s ‘super organism’ paper above, and Okasha and Paternotte’s recent critique. The other interesting thing about organismality (vehicles) is (as the aforementioned reviews point out) that the delineation between organism/non-organism is not clear… bits of vehicles sometimes start acting in organismal ways, competing for reproduction, etc. (e.g. meiotic drivers, cancer cells, etc. etc.) and of course at some point independent organisms started getting together and acting like vehicles.
@Marshall: Apologies, clearly I read the tone of your post wrong.
Yup, I think we’ll have to agree to disagree on Williams ‘66. Perhaps you are reading it in a more appropriate historical context, and seeing Williams as narrowly responding to what came before. I am perhaps reading it as more than this—a response to Wynne-Edwards et al., but also setting out the framework going forward. That is certainly the outcome, whatever the intent.
I confess I haven’t yet read Samir’s book, though I have browsed it. Given the difficulties of defining an adaptation (in the real world at least, if not in theory), I wonder about the value of arguing about group-beneficial trait v group-level adaptation. While the maths may lag, I can’t see that the group-level adaptations will require going outside of the frameworks we have (surely group-level adaptations have to be the product of group-level selection, as simple as that is to write and maybe much harder to show!). I guess I need to bump up Samir’s book in my reading list.
But the messiness doesn’t surprise me. Maybe it’s the teasing apart that is ultimately very difficult?
@O’Gorman: I think Okasha and Paternotte, 2011, Group adaptation, formal Darwinism and contextual analysis (JEB) is perhaps the most up-to-date and direct route into the maths and logic of studying group adaptations… I read it before publication, and still feel I need to go back and look at it again (probably several times): http://onlinelibrary.wiley.com/doi/10.1111/j.1420-9101.2012.02501.x/full
It seems like a bit of a straw man to contrast group selection with individual selection. Group selection competes with kin selection. Kin selection partitions selective forces into “individual” and “other” partitions – which explains why it has been so much more useful than group selection – but it isn’t really the same as individual selection. For an example of a genuinely individual-based model, perhaps look to economics.
By all means, let’s keep up both the cordial tone and high level of discourse—a refreshing contrast to the usual internet fare.
I admire Samir’s book and his excellent scholarship, but his distinction between type 1 and type 2 group selection is problematic. The distinction, if I remember correctly, is that type 1 involves traits that can be measured in individuals while type 2 involves traits that can only be measured for groups (such as population density). The problem is that ALL of the canonical examples of group selection are type 1 (e.g., the evolution of an altruistic trait, which can be measured in individuals). In addition, type 2 traits can evolve by pure within-group selection. For example, Michael Wade’s experiments on group selection in flour beetles selected on the trait of group size, which can only be measured for groups. His treatments included group selection for high and low group size, along with treatments in which group selection was absent (all groups contributing equally to the next generation, regardless of their size). Within-group selection resulted in smaller group sizes because of cannibalism, an effect that was augmented by group selection for small group size and countered by group selection for large group size. Since all groups have a size regardless of the balance between levels of selection, the mere fact that it is a type 2 trait says nothing about group selection.
Thus, although the type 1 vs. type 2 distinction might be useful for some purposes, it can’t be used to classify cases as examples of group selection or not. One point made by Okasha is that type 1 group selection can always be rendered as adaptive at the individual (or gene) level by average fitnesses across groups, whereas this isn’t possible with type 2 group selection. But that’s no argument against group selection for type 1 traits such as altruism, as in all the canonical models. I think that Samir would agree with this and he would be a good person to interview.
@Marshall: Many thanks for the Okasha ref. Does it get me out of reading the book?
On folks such as Stuart West, Andy Gardner, and Alan Grafen, they’re obviously smart and well-read, but they’re committed to a kind of individualism that strikes me as problematic and in any case needs to be distinguished from individual-level selection in multilevel selection theory.
Grafen’s goal is to create a formal mathematical framework in which natural selection is fitness-maximizing at the individual level. This is doomed to failure, because cases of intragenomic conflict (selection among genes within individuals) cannot be interpreted this way. Grafen’s only solution to this problem is to claim that such cases aren’t very common, which is lame.
Whenever selection operates above the individual level in multilevel selection terms, it can be rendered as advantageous at the individual level by averaging the fitness of individuals across groups. And selection at all levels can be rendered as advantageous at the gene level by averaging the fitness of genes across individuals and groups. This might be useful for some purposes, but it’s no argument against group selection, which is why Elliott Sober and I called it the “averaging fallacy” in Unto Others. If Grafen wants to create a mathematical framework that is fitness maximizing at a single level, it should be the gene level, but that is already the point of selfish gene theory.
Another move by Gardner and Grafen is to articulate a concept of group-level adaptation that is distinct from group-level selection. They acknowledge that group-level selection is common (as approximately represented by the group term of the Price equation), but assert that group-level adaptations are rare. Elliott and I have replied to this argument in an article titled “Adaptation and Natural Selection Revisited”, which is dedicated to the memory of G.C. Williams. We show that, ironically, Gardner and Grafen’s argument is at odds with the legitimate core of Williams’ book.
I can’t help but think that these folks have a prior commitment to individual self-interest as a grand explanatory principle, which leads to a “if at first you don’t succeed try, try again” form of theorizing. I’m aware that it’s dangerous to speculate about cultural influences and ideological commitments, and that the arguments must be considered on their own terms. Nevertheless, we know that Darwin and his contemporaries viewed evolution through the distorted lens of Victorian culture, and we should be mindful of the distorted lens of our own individualistic culture during the last half century.
My last two comments are definitely in the “zone of legitimate controversy” rather than the “zone of agreement”. I look forward to focusing on the legitimate controversies in future ETVOL articles, now that the zone of agreement has been established.
I didn’t cite J.M.S to establish that kin and group selection are different – as Rick claimed – I cited him to illustrate that there are multiple meanings of the term “group selection”.
The form of “group selection” that Williams, Dawkins and Maynard Smith criticised was interdemic selection – which is quite different from the new group selection (which makes all the same predictions as long-established orthodoxy, kin selection).
As I said, using the term “group selection” withou making it clear which kind you are talking about is the source of much of the confusion surrounding the topic.
A short explanation/ definition of what you understand by (1) interdemic selection and (2) the new group selection would be much appreciated Tim. I suggest that we need to have a common understanding of these terms if we are to get rid of the confusion you, quite rightly, refer to.
Hello, non-biologist here. I would like to ask, could any of you provide me with a link on a paper that deals with what exactly is (could be) ment by “group selection” (e.g. a good review of definition(s) of group selection)? Many thanks!
Hey DMark, check out this from David Wilson:
Its a compendium of blog posts exploring the issue, and I think covers what you are looking for.
Rick O’Gorman, thank you very much!
I am amazed by the fierceness of the attacks against group selection. The detractors seem to think that every posible evidence or explanation has been either ruled out or mistaken as kin selection.
If this is true I would like to ask them if there is any mathematical model which analyzes the possibility that group selection has evolved not only through a process of multilevel natural selection but also through a process of individual sexual selection which could have fostered genetic instincts such as altruism or a tendency towards male heroic acts under certain circumstances (specially when we are watched).
I am talking about a kind of Fisher´s runaway process like the one that Geoffrey Miller uses to explain the uniqueness human mind.
Peacock tails can´t be explained only in terms of natural selection. Perhaps neither can group selection.
Just think of certain human female tendencies: Why women get excited by men in certain uniforms? Why women fell in love with men that make heroic acts (apparently altruistic)? Why women prefer to get involved in a monogamous relationship with men who are empathetic and generous? and so forth…
A little introspection make things even more clear. Would be we human males be equally motivated to do a heroic feat if we knew nobody is watching and nobody would ever know about it?…
In my opinion sexual selection might be the key to explain instincts and groupishness tendencies (gene-based) that seem to go againts the individual fitness. If these instincts and tendencies make individuals more atractive, group benefits can align with individual benefits.
1)Hello! Seeing that some evo. biologists came and posted here from time to time, I thought this would be an ideal place to ask my question regarding Hamilton´s rule and possibly be given an expert advice. I am not a biologist, but a curious high-school student:) and am having a hard time finding in-more-depth discourse on Hamilton´s rule, its theoretical implications and how these should be reflected in its math.
2)I explain “my problem” by following an example given in Campbell (2008): “Imagine that a young man is close to drowning in heavy surf, and his brother risks his life to swin out and pull his sibling to safety….The benefit to the recipient of this altruistic act is…two offspring (B=2; average reproductive output for humans). Let´s say that in this kind of surf an average swimmer has a 25% chance of drowning. We can calculate the cost of the altruistic act as 0.25 times 2, the number of offspring expected if the altruist had stayed on the shore: C = 0.25×2=0.5….(Since) full siblings…share half of their genes on average (r=0.5)….rB=0.5×2=1 and C=0.5. This satisifies Hamilton´s rule; thus, natural selection will favour this altruistic act of one brother saving another.”
3)One of my questions is: Does Hamilton´s rule apply only in a case when one individual saves another individual (and thus his hypothetical (unborn) progeny)? Or, to put it differently, how would the math in the above example “looked like” if not one, but say TWO brothers were drowning, and “the altruist” could have saved both (i.e. trying to save the second does not (significantly) endanger them all)?
4)When it comes to “the cost of the provider” you can figure out the probability both brothers will die, but how would “the-benefit—to-the-recipients” side of the inequality look like? Can you add the benefit of the recipients?:
rB > C
(r1B1 + r2B2) > C
5)There are some issues with doing this, but I am not sure whether these are relevant. To explore, r1=r2=0.5 – this is correct mathematically, but not “biologically” for r1 and r2 are two different sets of the 50% of genes the two brothers share with “the provider”. Therefore, when comparing “the gene content” r1 and r2 represent, these might be the same in 99.99% of genes, but will differ in that 0.01% of genes. This leads to “the provider” saving not only more than 50% of its genes, but many of those in double sets (i.e. “the provider” has genes A,B,C,D; bother1 shares with him genes A,C and brother2 genes A,D – saving them means saving genes C, D and 2xA). I know that “r” is “quantitative (and approximate)” and “not qualitative”, but if Hamilton´s rule is about “how many more of my genes will be passed” should not this matter?
6)My last question: Can I make a “special use” of Hamilton´s rule and consider scenario “parent saving offspring”? One reaction I got to the to be proposed example was that “you do not need Hamilton for parent saving his offspring as this can be explained under “classical” definition of fitness” – I know that, however I wanted to demonstrate this using mathematical model.
7)Thus imagine: parent can save its children (we think of a parent in the terms of his life-time) by dying himself. In this case, the “cost of parent” (C) (as I understand it) can be measured as the probability of offspring´s death as a result of loss of parental care (x) times the number of potential offspring (y). For species needing parental care, the probability of death of offspring in rearing period (depending on stage) will be approximately 1. However, parent will also have mature offspring in whose case the probability will be 0. x will be the sum of all “death probabilities” (∑x) divided by the number of born offspring (z; y >z). Logically, ∑x will be less than 1. With r being 0.5 and B being the same for both parent and offspring (B; y=B; B1=B2=…=Bz), the adjusted formula would look like as follows:
rB > C
(r1B1 + r2B2 +…+rzBz) > C
rBz > C
‹0.5;1) Bz > ((∑x)/z) y
‹0.5;1) Bz(squared) > (∑x)B
‹0.5;1) z(squared) > (0;1)
8)Now, in this example I connected the two problems I encountered while dealing with Hamilton´s rule: 1) Can you add the benefit of recipients (in here I used range ‹0.5;1) to suggest more than 50% will be passed, however, this – again – does not completely overcome the whole “bio problem” I mention in paragraph 5)? 2) Can you use Hamilton´s rule as a tool to provide “mathematical proof” of cost/benefit to the parent “sacrificing” himself for his own offspring?
9) I realize the answer might be “time-consuming”, but if some of you could provide me not with a direct answer, but a link to a paper dealing with this issue I would be more than happy.
(I am not English native, so if something is not clear, let me know – I will try to clarify.)
I FORGOT TO INCLUDE this in my previous post – please, consider this an ending of PARAGRAPH 8:
“In Cambell, the definition of Hamilton´s inclusive fitness is: “the total effect an individual has on proliferating its genes is by producing its own offspring and by providing aid that enables other close relatives who share many of those genes, to produce offspring”. So, – when it comes to Hamilton – could we think not just about individual producing offspring and aiding to non-offspring (other close relatives), but about individual producing offspring and ading to offspring as is the case in our example (parent “aids” offspring by dying)?
So no one knows the answer?:)
@DMark: Funnily enough I was just looking at your post last week and waiting to find time to reply. Briefly though only, I’m afraid, but hopefully enough to help.
3) Yes, it can apply to more than just saving one person. As you have gone on to do…
4) Yes, as you have done if I am reading it correctly.
5) You are going wrong here (as many do). It is a trait by trait process. That is, you focus on one particular trait. Inclusive fitness does not say that an altruism gene works for all the other genes in the same body. They may get carried along but the altruism trait is favoured only if it benefits copies of the same gene in other individuals (other copies). Thus the r only applies to the likely relatedness between individuals for the specific trait/gene in question. Note that in practice it is the trait that matters but that for inclusive fitness we end up talking about selection on genes. One of the pros of MLS is that it does not drive the focus to the gene, skipping the phenotype (not that an IF approach has to do this, it’s just how it tends to be framed).
6) Yes, but it won’t yield any of the asynchrony that parents & offspring exchange of help yields in reality (i.e., we would expect more help to go to the offspring from the parent(s) than vice versa).
7) Essentially this is correct (I haven’t carefully parsed your maths though). Some offspring will be older and more independent so the benefit gained by parental sacrifice is lower.
8) 1) As discussed above.
8) 2) I’m not sure what you mean here—it is only a model, so no, you can’t achieve true mathematical proof of the reality. Only true within the assumptions that you make. If you limit the system, then you can prove something mathematically here, but then you’ve limited the system, so it cannot be a proof for the real world, where there is much more to include (e.g., age diffs limiting future reproduction, experience, value of older helping offspring, etc.).
Hope that helps!
Dear Mr. O´Gorman,
A) First of all, thank you very much for your response. The information you provide in 5) was new to me. In my school book “r” is defined as “the coefficient of relatedness which equals the fraction of genes that, on average, are shared”. If I understand correctly what you wrote, then this definition is incorrect (?) and “r” should be only thought of as the probability that the given gene of “the provider” is also shared by “the recipient” (and therefore, my calculation in 7) should use 0.5 instead of the range ‹0.5;1)). (One thing strikes me as odd – if only one gene is considered, what is the relevance of Hamilton´s rule in relation to individual´s kin or precisely, to being more willing to behave “altruistically” towards members rather than non-members of one´s own family? Within humans there is a large percentage of genes that are shared, so if “my objective” was to ensure “the one gene” I have will be passed it would not matter whether to save “my brother” or “my neighbour” as there is a high probability that “the one gene” in question is “the one own to all/majority of people”. There is, of course, an assumption that it is true that individuals do rather save members of their kin (group) than others. If we omit the issues such as “the degree of social attachment” (“I like my brother more than I do my neighbour, so I would rather save him”), how to answer this from a solely biological (in this case maybe genetic) perspective?)
B) As to the 8) 2), I was only trying to rephrase the problem/question that 6) and 7) deal with. My question was not whether Hamilton´s formula (or the formula as adjusted in 7)) provides a way of how to reach “an absolute truth” (or, as you say “proof of the reality”; I am aware it does not), but rather whether Hamilton´s formula can be used as a model in which parent “sacrifices” himself (dies) for his own offspring. You answered that in 6; 7, so thank you.
C) I would like to ask, do you happen to know where I could find in more-depth information on Hamilton´s rule; its theory and implications (For example, I find it interesting that even though IC approach – as you say- “drives the focus to the gene”, the act of “helping” another individual is still regarded to as “altruistic” even though – when genes are considered – the benefit to “the recipient” is, in fact, the benefit to “the provider” (and thus selfish). Is it the result of a conflict of an individual being thought of – in this case – on an organismal as well as genetic level? I mean, the rule needs individual-organism (“organismal level of an individual”) to provide help, but it is the spread (survival) of the gene own to the “helping” individual (“genetic level of an individual”) that determines the cost of its act…) “The discussion” on Hamilton´s rule in my high-school book is covered in one paragraph and even that is a verbal version of the inequality. I am very curious about whether the math in 7) is correct (i.e. be it even the most basic (simplified) mathematical representation of the problem described, but without logical errors), but it is impossible to work with the rule mathematically if I lack the biological background. Should you have a link to a good article to provide I would very grateful.
(PS: I do not suppose that any of my “questions” is “new” (they do not raise any point/argument/..that was not raised and addressed before) and therefore worth your time – or anyone’s for that matter – to reply (they are simply a result of my lacking knowledge – I am fully aware of that), but if you could provide me with a reference to a book or an article that would contain an answer to any of my questions, I would be – as I said:) – very thankful.)
your high-school definition of r is wrong… r is the probability that the partner contains the same allele (for altruism) as the donating individual – above the frequency of the allele in the population -… this resolves the problem that so much of the genetic material in biological populations is identical. You could write this simply as Prob(partner has allele)=r + (1-r)f where r is the ‘relatedness’, f is the population frequency of the allele, and both are between zero and one inclusive. What this makes clear is that if the allele is fixed in the population (f=1) then regardless of ‘relatedness’ the probability your partner has the allele is 1 (since r+(1-r)=1).
Now, ‘relatedness’ as described above has just been defined rather than derived. Various people in the 70s and 80s did foundational work deriving relatedness from first principles… what I described above is actually Alan Grafen’s (of Oxford University) geometric view of relatedness. Andy Gardner (also at Oxford) with colleagues recently wrote an excellent summary of the generality of the theory, entitled ‘The genetical theory of kin selection’. The PDF is available for free from the author’s website (http://www.zoo.ox.ac.uk/group/gardner/publications.html)
James Marshall has already answered you before I could with a great response. I’ll add a few things. The textbooks often get definitions of Hamilton’s Rule and Kin Selection wrong, unless they are EvPSych/Evolution textbooks. Intro Psych textbooks can be quite bad, for example (I teach, among other things, intro psych).
To try to address your ‘own odd thing’ aside, briefly: Natural Selection is about traits rather than a unified individual (this is a simplification but for here it is fine). Thus, we ask questions such as how does natural selection favour running speed in cheetahs (which itself may actually be several traits, depending upon how you divide things up). For running speed, we don’t expect there to be a cooperation angle to it. But for any behaviour that has a social dimension, this is possible. Classically this is simplified to an ‘altruism’ trait. I say simplified because altruism does not have to be a single unified trait. But for the purpose of working through models, it is okay.
So, for an altruism trait (passed on by a single gene), the question is whether that trait prompts it’s vehicle (the individual in whom it is found) to act in a way such that the altruism trait becomes more common in the population. Altruistic traits achieve this not by making the vehicle/individual more fit (eg, faster) but by transferring a benefit to others. Depending upon the complexity of the trait, that may be as simple as randomly helping others. Or directing that benefit to kin. The trait prospers (increases in frequency at the expense of an alternative version/allele) if the benefits tend to accrue to others with the same trait, because they then translate that benefit into reproductive success.
Now, why just focus on one trait? Well, that altruism trait also helps lots of other traits that occur in the recipients, but most of those never ‘pay back’ the altruism trait. For example, if the recipients also have a ‘run faster’ trait, then they will also pass that on thanks to the help gained from the act of altruism but the run-faster trait is not other-oriented. So others with the altruism trait do not gain from the success of the run-faster trait.
Moreover, usually we assume that traits are independent of each other. So the run-faster trait is no more nor less likely to be in a body that has an altruism trait than a selfish trait. Thus, the gains made by the run-faster trait above when in the recipient of altruism is just chance.
Does that make sense?
As for further reading, it depends on the level you want to delve into. Samir Okasha’s book mentioned way above (Levels of Selection) could be a good read, but it will be technical, as will the paper that James M mentioned from Gardner (plus I find Andy Gardner’s work excessively pro-IF and anti-MLS but I may be prematurely harsh). The Selfish Gene by Richard Dawkins remains a great intro to understanding IF without the maths as long as you remember the details here, that Dawkins is virulently anti-MLS (though not to the same level that he is anti-religion, but maybe close and with the same level of rationality!). David Wilson has a raft of pubs on his website which are also very readable and also avoid the maths and put Hamilton’s Rule into a broader framework of how it will relate to MLS. http://evolution.binghamton.edu/dswilson/publications/ [Full disclosure: I am a former PhD student of David’s.]
And as James says, keep questioning! Already your questions are deeply impressive for a high school student.
Dear James Marshall and Rick O’Gorman, thank you for your response (and the links provided)!
I forgot to especify: when I’m talking about phenotypic traits, i meam herritable phenotypic trairs.
when you say heritable phenotypic trait, do you mean perfectly heritable? If so, this is just the phenotypic manifestation of having a particular gene. Various people, starting with Dave Queller, and more recently Jeff Fletcher and Michael Doebeli, have argued for a phenotypic measure of relatedness as being the most relevant for the evolution of altruism. Dave Queller did the original and in my opinion the best work on this, in particular explaining the quantitative genetics approach to studying selection on conditional social behaviour phenotypes; crucially, for an evolutionary response to selection acting on a phenotype there must be a non-zero correlation in some direction between phenotype and undying genes, hence the breeder’s equation of quantitative genetics. So yes, selection acts on phenotypes, but evolutionary change is change in gene frequency (as Fisher, Price, etc. pointed out) and everything must be related back to that… your proviso that you’re considering heritable phenotypes means you at least implicitly understand that their genetic bases are crucial, although this contrasts with your initial wording “no matter if there are or not a genetic base for this behavior”
Hi James, thanks for responding.
When I used the sentence “no matter if there are or not a genetic base for this behavior”, I was thinking in two possibilities:
1- Learned behaviors. Some primates, including us, some birds, other mammals, etc., can learn a behavior and pass then on through social learning or imitation. I know that in practice is difficult to have evolution without changes in gene frequency, but if a learned behavior is heritable in this way, the genes associated of the learn capacity don’t have to change, although the phenotype could. I’m not sure if this really happens, but some authors i.e. Eva Jablonka works theoretically in these lines. Altruism surely depended of genetic basis for evolve, but maybe not only this. Maybe one generation receive the phenotype of altruism by genes together with learned behavior, so there is inheritance of the whole phenotypic trait.
2- For the initial evolution of altruism in hymenoptera, there is plenty of evidence of the impotence of high genetic relatedness, but in time, some trensgenerational epigenetic inheritance, maintained by food ingestion or pheromone control could have played the high “r” without this meaning necessarily a genetic “r”, helping to solve the problem of nowadays multiple matting queens and low genetic relatedness.
All need to be empirically tested of course, I’m just speculating…
I’m a under graduated student (almost graduated) in biology, and I have a real interest in those questions debated here, so much that I’m familiar with the literature cited, although I don’t know if I’m qualified for give my opinion in this subject because I don’t worked with this. As far as I see, the question on differences between IFT or MLS is really becoming a matter of preference. The long dispute is turning into a revival of the same arguments from both sides. Sometimes seems that we have only a semantic disagreement, related to how to measure and what the “r” really represents. Of course high “r” values are agreed to be important in IFT or MLS, but the last do not find it necessary. Thinking In the light of the new evidences of transgenerational epigenetic inheritance, and cultural evolution, I think we should consider the “phenotypic trait” the heritable unit o selection, not a “gene” anymore, and then assume that the MLS works fine with low genetic relatedness IF we have phenotypic relatedness for a trait shared in a population. There’s more than one genetic way to achieve a phenotype. A phenotypic trait of sacrifice for the members of your group will work, as long the group consists of a bunch of “sacrificers in potential”, no matter if there are or not a genetic base for this behavior. But if there is, of course we get a spread of that behavior too. But that’s not necessary. At least in theory… What we need now is a case to case empirical study.
Sorry for the bad English (I’m from Brazil), and sorry if I’m talking the obvious.
OK, I see you have in mind cultural evolution of social behaviour. People like Rob Boyd have worked on this for a number of years, and Alex Mesoudi has just published a book on cultural evolution that I think was very well received. It’s not really my field, but happy readings and ponderings…
if you really believe within-genome conflict to be relatively infrequent, you may know very little about transposable elements, B chromosomes, driving sex chromosomes and a host of other examples covered in our massive book on the subject (which can be downloaded in its entirety under Publications on my webpage, roberttrivers.com)—Genes in Conflict: The Biology of Selfish Genetic Elements. Harvard. 2006. Austin Burt and Robert Trivers.
That this page is called “evolution-institue” is worrisome, as it may confuse visitors in to thinking that David Sloan Wilson is an authority on biology. He is seemingly, or rather conveniently “middle-way” in his opinions. Fortunately for him this means they are right because everyone else is a radical. It is important to keep in mind that the “Dawkinsian” side he argues against is basically molecular biology. Which basically is biology nowadays. However, it does give rise to astonishing questions, perfect example is cited by Trivers above with regards to Transposable Elements. How did they evolve? Another was recently published in Cell, called “What’s Luck Got to Do with It: Single Cells, Multiple Fates, and Biological Nondeterminism” by Raj and Symmons (2016). Now isn’t that an exciting title? It discusses multilevel natural selection in detail, yet not a single reference to DS Wilson and his “accepted theory”. Nor even a hint at it or, as is claimed by its humble inventors; profound implications for life, the universe and everything.