Dawkins on Nowak et al. and kin selection

March 24, 2011 • 7:27 am

If you’ve read the critiques of the Nowak et al. paper on kin selection that I highlighted this morning, you may have noticed a conspicuous absence among the authors: the name of Richard Dawkins.  Why, as the most famous exponent of kin selection since W. D. Hamilton, didn’t he co-author one of the critiques?

It turns out that nobody asked him.  I think this was simply an oversight, because all of us simply assumed that Richard would be penning his own criticism.  He didn’t, but he did write a brief piece on Nowak et al. for New Scientist, which he decided not to publish.  I’m posting it here with his permission, along with a link to his excellent paper about common misunderstandings of kin selection.  I’m not sure whether Richard will answer comments from readers, but you can certainly pose them below. Here’s his take on Nowak et al.:

This is no surprise. Edward Wilson was misunderstanding kin selection as far back as  Sociobiology, where he treated it as a subset of group selection (Misunderstanding Two of my ‘Twelve Misunderstandings of Kin Selection‘: Zeitschrift für Tierpsychologie 1969). Kin selection is not a subset of group selection, it is a logical consequence of gene selection. And gene selection is ‘standard natural selection’ theory. Inclusive fitness theory is not some kind of supernumerary excrescence, to be ‘resorted to’ only if  ‘standard natural selection theory’ is found wanting (Misunderstanding One). On the contrary, inclusive fitness theory is one way of expressing what was logically inherent in the neo-Darwinian synthesis ever since the 1930s but had been largely overlooked because people didn’t think hard enough about collateral kin. ‘Standard natural selection theory’ MINUS inclusive fitness would be like Euclidean geometry minus Pythagoras’ theorem.

Another way of expressing what was logicially inherent in the synthesis is Hamilton’s rule, rB>C: a gene for altruism will spread if the cost to the altruist, C is exceeded by the Benefit to the recipient, B, devalued by the coefficient of Relatedness, r. If you think, as Nowak et al. do, that ‘Hamilton’s rule almost never holds’, that simply means you haven’t been measuring B and C carefully enough. r is not the only term in Hamilton’s inequality. B and C matter too, and your game theoretic considerations are subsumed within them.

Perhaps most irritating is Nowak et al.’s concentration on haplodiploidy, which, in Hamilton’s original paper was a throwaway side-issue, interesting enough to pique the interest of generations of students, but not in any sense central to his paper. Of course Hamilton was well aware that eusociality is present in diplo-diploid animals, exactly as inclusive fitness theory would predict given appropriate B/C ratios. Indeed, Hamilton himself put forward an ingenious theory of the evolution of eusociality in termites, predating by seven years the version usually attributed to Bartz (attributed by Hamilton himself, indeed, with characteristically absent-minded generosity as I described in The Selfish Gene, second edition p 317).

Finally, Nowak et al. do Darwin an injustice, in discussing his theory of the evolution of worker sterility in social insects. They paraphrase Darwin’s ‘well-flavoured vegetable’ analogy. Let me quote it exactly: “Thus, a well-flavoured vegetable is cooked, and the individual is destroyed; but the horticulturalist sows seeds of the same stock, and confidently expects to get nearly the same variety . . . I do not doubt that a breed of cattle, always yielding oxen with extraordinarily long horns, could be slowly formed by carefully watching which individual bulls and cows, when matched, produced oxen with the longest horns; and yet no one ox could ever have propagated its kind. Thus I believe it has been with social insects . . .” It is true that Darwin goes on to phrase his idea in terms of benefit to the colony, but his analogy of the long-horned (castrated) oxen could not be clearer. No colony is involved. This is early inclusive fitness theory. It is entirely clear that, if Darwin had been alive to read Hamilton on social insects, he would have embraced inclusive fitness, not as an add-on to natural selection theory but as the logical way to express it in the age of the gene.

64 thoughts on “Dawkins on Nowak et al. and kin selection

    1. Wilson has not been the only bright scientist to misunderstand kin selection and fail to see it as a logical and inevitable consequence of genic level selection (not group selection). After reading Stephen Jay Gould’s magnum opus, The Structure of Evolutionary Theory, I got a general sense that Gould too failed to see that it requires a “genes-eye” view to fully understand how it works.

  1. “The list of authors and their institutions, which occupies two pages of the three-page letter, reads like a Who’s Who of social evolution.” … that argument is uncomfortably reminiscent of an infamous book titled “Hundert Autoren gegen Einstein” (Hundred authors against Einstein) http://en.wikipedia.org/wiki/100_Authors_Against_Einstein#Hundred_authors_against_Einstein

    Shorter: JC is a reactionary Kintian.

    Pun too obscure? Let’s just say that we can’t declare the show to be over until it appears on the hallowed pages of Conservapædia.

    1. Or, for that matter, a recent paper in Taxon with 150 authors against cladism (2005; Taxon 54(1): 5-8). “Look, there are lots of us!” is not a great argument in science.

      P.S. “Kantian”. 🙂

  2. “Kin selection is not a subset of group selection, it is a logical consequence of gene selection. And gene selection is ‘standard natural selection’ theory. Inclusive fitness theory is not some kind of supernumerary excrescence, to be ‘resorted to’ only if ‘standard natural selection theory’ is found wanting…”

    That’s exactly what I’ve been saying.

  3. I doubt the geekspeak version of the response would be published by anyone: RTFM. Or, as a teacher of mine loved to say: “don’t you know how to read a book?”

  4. Just wanted to second the recommendation of Richard’s paper “Twelve Misunderstandings of Kin Selection”. I use it in a graduate level behavioral ecology course, after the students have struggled through Hamilton’s 1964 papers, and they find it extremely helpful.

    And it is also completely non-strident, non-combative, and very well written. I’ve never understood people who claim Dawkins’ writing is anything else.

    1. Have downloaded and will read carefully– some of Prof Dawkins’ best writing comes to us when he undertakes to clear up common misconceptions. And I wholeheartedly agree– I don’t understand how anyone can read Dawkins and come away from it thinking it was anything but calm, clear, and concise.

      1. Thanks to cognitive dissonance, disabusing people of foolish notions will always cause pain and therefore be interpreted as stridency.

        The only cure is for the afflicted to abandon those positions they hold that aren’t in agreement with reality, I’m afraid — regardless of how comforting their beliefs.

        See? Strident. Can’t be avoided, so why try?



          1. But to admit to chronic error is hard, as one must necessarilly face the ignominy of personal fault; first, to one’s self, then to one’s peers.
            It takes guts to do that.
            It is far easier to blame the messenger for their delivery of a reality slap.

    2. The epitome of this, to me, comes when people (like Midgley) criticize The Selfish Gene for allegedly asserting that the gene is literally selfish. When I read the book, I found that Dawkins’ every-single-chapter repetition that “please remember this is all a metaphor, and we need to return to the literal reality from time to time as a sanity check” was starting to get borderline tedious. And yet somehow people still missed it. It boggles the mind…

      1. It’s not really boggling at all.

        How can it matter how excruciatingly clear he was if the critic never read the book in the first place?

        Though it’s interesting that I don’t find absolutely ridiculous the notion that a gene can be actually selfish, if you follow the meaning of the term down to the bottom. When a person is selfish, it all boils down to molecules in action, so why can’t a shorter chain of interactions also hold the label? I don’t think it’s a particularly useful argument, but if a person insists on not understanding what a metaphor is, why not have some fun forcing them to attack the strawman properly?

  5. “If you think, as Nowak et al. do, that ‘Hamilton’s rule almost never holds’, that simply means you haven’t been measuring B and C carefully enough.”

    Hm. I think this translates to: “If the theory I’m advocating isn’t supported empirically, the data is wrong.” This argument has some difficulties. 🙂

    1. Actually that translates to “this theory has been empirically validated but critics often miss-represent the theory by focussing only on the relatedness and ignoring the fact that the respective cost/benefit analysis is also important.”

      1. No, you’re thinking of the next two sentences: “r is not the only term in Hamilton’s inequality. B and C matter too […].” If that’s the point he’s trying to make in the preceding sentence, he’s got a funny way of going about it.

        As for “this theory has been empirically validated…” that bit you just made up. It ain’t in there.

      2. Also, FWIW, that the cost/benefit relationship is really what’s important is one of the main points made by Nowak et al. The other main point being that relatedness–particularly as it is typically measured–isn’t the only factor that can tip the balance of that relationship.

        Regarding measuring relatedness–the typical approaches to this really make a hash of things when it comes to understanding inclusive fitness. For the very short version, consider the extreme case: there is no genetic diversity whatsoever in the population. What does a naïve application of Hamilton’s rule suggest in this case, and does it make any sense?

        1. Isn’t that Misunderstanding #5? Or possibly #7?

          I’m certainly not an expert at this stuff, but I took it as obvious that r applies only to the variable portion of the genome.

          1. Having just read that 12 misunderstandings paper, yes, he does cover the issue under misunderstanding 5.

            However, my overall impression after reading that paper is that the “misunderstandings” are indeed errors, but I’m not at all convinced that the error involved is misunderstanding kin selection.

            For instance, in misunderstanding 5 he says that:
            1) definitions of “r” are basically unworkable;
            2) it isn’t really overall relatedness that matters anyways, but sharing of specific genes involved in cooperation (the “green beard effect”; see further considerations in “misunderstanding 4”);
            3) evolutionary stable strategies are a more useful way to look at the matter.

            Just substitute “objection” for “misunderstanding”, and you’d’ve concluded by the end of the thing that kin selection was dead. Hell, he says it pretty directly; for the evolution of altruism “there is no need for the individuals to be kin”. There goes any naïve use of “r” and Hamilton’s rule in research on the evolution of altruism, right? Well, AFAICT, that’s precisely what Nowak et al. are suggesting, too.

            1. Oh, and on this specifically: “I’m certainly not an expert at this stuff, but I took it as obvious that r applies only to the variable portion of the genome.”

              It varies depending on how r is calculated. The simplistic “r of parents to offspring is 0.5” sorts of statements involve the assumption that the parents are dissimilar at all loci involved. There are various complications to account for this. Empirical measures run roughly along the lines of looking at whether two individuals are more similar to each other than they would be if chosen at random from the population*, but can extend to explicit analyses of paternity & so forth.

              *This can have interesting results. If there is very little variation in the combination of markers & population investigated, inferred relatedness among individuals is uniformly low. If similarity in the reference population as a whole is near one, no pair of individuals can deviate from it to any substantial degree. There are a couple of fun papers on Solenopsis (fire ants) that show this result. There’s damned little genetic variation in introduced populations. Consequently, measures of relatedness drop down near zero and researchers wonder why Hamilton’s rule isn’t holding. Doh!

              1. Perhaps Nowak’s greatest flaw in the original paper is his rather arrogant assumption that mathematical models alone must dictate what is to be accepted as true, and that verbal models or deduction from evidence has no place whatsoever in Biology. Worse than this, his own model stacks up a preconstructed sequential mathematical structure to get the end effect that he is aiming at. And even worse than both of these, he completely neglects testing if Inclusive Fitness would improve upon the results of his own model. What Nowak is doing in essence is mathematical curve fitting to his own preconceived ideas.

  6. Amusing quote from the “Twelve Misunderstandings” paper:

    We do not separate Neptune, Uranus and Pluto off from the rest of the planets simply because for centuries we did not know of their existence. We call them all planets because they are all the same kind of thing.

    Well, not so much anymore.

    1. Yes, but in all fairness Pluto was separated out later precisely because it *wasn’t* “the same kind of thing” (of the same population).

  7. I was listening recently to a discussion between Richard Dawkins and Lawrence Krauss in which they argued about who was the greatest scientist, Einstein or Darwin. Dawkins thought that natural selection was a simple concept and that therefor Darwin didn’t need to be very clever to come with it and that Einstein was therefor the more brilliant scientist.

    This argument reinforces for me what I thought at the time, which is that although the concept is very simple, it is clearly not one the human brain is capable of dealing with easily, and that therefor it was probably far more difficult to think of than it seems now that we are familiar with it. Like maths which is so simple that computers can be progammed to do it very easily, but which the human brain is not really designed for so we aren’t very good at it, natural selection is clearly more difficult than it seems (to Richard Dawkins at least ;-)). It certainly must be at the head of the queue for ‘most misunderstood scientific theory’. If I had a dollar for every time I’ve read errors about evolution in books and articles written by people who should have known better, I could take myself out for a very nice dinner!

      1. Second that–well said:-) It’s frustrating that human brains have such a hard time grasping anything that involves geometric progression. I *know* the answer to the grains-of-rice-on-the -chessboard problem, and I’m still gobsmacked every time I actually sit down and work it out. Sit there looking at the calculator like it’s orf its bleedin’ ‘ead.

    1. Actually computers are really good at crunching numbers, they are terrible at mathematics 😉 It took a human to prove Fermat’s Last Theorem.

      1. Your implication is somewhat ingenuous. Wiles could not have proved the theorem without a massive amount of automated computation.
        The disputes over the huge involvement in artificial electronic mathematics involved in his proof resonate to this day.
        Computers have become very good at mathematics, and an algorithm partly developed by the PhD student at Cornell: Michael Schmidt (along with Hod Lipson) is able to derive fundamental equations from raw data.
        (Science 2009, Vol 324, p81)
        A version is available on-line.

        1. To say an algorithmic machine is better at mathematics than a human is, to my mind at least, like saying an abacus is better at arithmetic than a human. An abacus needs direct human involvement, a computer, one can wind up and let go.

          Computers are much quicker at performing logical algorithms than humans, as long as the humans create the algorithms. I don’t consider this thought or mathematics. There are computational proofs, of course, but again these rely on the sheer capacity of computers to crunch numbers.

          Asserting that these logic machines are good at mathematics is like saying a Porsche is good at sprinting.

          1. In what meaningful sense is a Porsche not good at sprinting? Is it that “sprinting”, by definition, can only be done by a biological organism?

            1. In the sense that a Porsche can’t sprint without humans to design, produce and operate it.

              Is it the Porsche doing the actual “sprinting”?

              1. How is your post different from saying this:

                In the sense that a human can’t sprint without millions of years of natural selection to “design”, “produce” and “operate” it.

                Is it the human doing the actual “sprinting”?

              2. Touché, Sean. Although if sophism is being engaged in, then yes, it is the human doing it, just like it is the Porsche doing the sprinting. Maybe I’m just being stubborn – I can see the distinction you’re making, but I’m not sure I appreciate its significance.

              3. Bryan, the whole Porsche thing is really just a tortured metaphor.

                I think the idea of computers doing mathematics is as ridiculous as abacuses doing mathematics.

            1. I’m only to offer an opinion after perusing your designated reading material?

              Why don’t you simply respond to what I wrote, please?

              Or just ignore it, please?

              1. 1) I did answer it, but necessarilly briefly.
                To summarise the scores of academic papers adequately for one at your level would take longer than the number of compiled pages alone. You request the impossible.
                2) You continually forward your personal incredulity on a complex subject as a reason for disbelief.
                This is a clear logical fallacy.
                3) It is not my job to be your unpaid tutor. It is up to you to resolve your ignorance on automated math algorithms for yourself. That is if you actually care about what is true.
                4) I did not ‘ignore’ your post. I went out of my way to check the author’s names and the date of the paper for your edifictation.
                I hope that I did not waste my time.

          2. Sean, computers are capable of abstract mathematical reasoning and have in fact discovered novel theorems previously unknown to mathematics. These are genuine mathematical discoveries in the sense that the software was not pre-programmed to arrive at those specific theorems but was exploring proof space by following interesting lines of reasoning (for some quantitative definition of “interesting”).

            If that doesn’t meet your definition of what it means to do mathematics, perhaps you’d care to spell out a definition that excludes the products of artificial reasoning systems but includes the products of naturally evolved reasoning systems.

    2. I think quantum mechanics easily outranks natural selection as ‘most misunderstood scientific theory’.

    3. It is a silly argument – who is the prettier, which is the best film/song/novel etc. These things do not have to equate. Darwin would not, I think, mind being called a ‘plodder’ – in the nicest sense. He was an accumulator, an observer, a cautious man who had much to lose (to wit, his wife’s good grace & social acceptability). Einstein was a genius – an outsider though, & working in an abstract form. They are simply different.

    4. I usually grind my teeth with these “who was a better scientist” question (unless of course a kook or a mediocre scientist is being compared with one of the very obviously more brilliant people in history). Einstein and Darwin worked on very different problems. Darwin gathered a lot of information on animals and worked out a very convincing argument that all animals are related and that Natural Selection plays a part in the evolution of species. Einstein looked at reported measurements of physical quantities (positions of planets and stars, etc) and worked on reconciling differences between observation and the existing ideas about how things worked (Newtonian mechanics, light propagation, etc). They were people working on different topics; we can hardly compare them.

      1. For some reason this brings to mind the “what famous person would you like to fight?” scene from Fight Club:

        Tyler Durden: OK: any historic figure.
        Narrator: I’d fight Gandhi.
        Tyler Durden: Good answer.
        Narrator: How about you?
        Tyler Durden: Lincoln.
        Narrator: Lincoln?
        Tyler Durden: Big guy, big reach. Skinny guys fight ’til they’re burger.

    5. they argued about who was the greatest scientist, Einstein or Darwin.

      Most physicists would say Newton (before Einstein). 😀

  8. I do love Dawkins’ writing. So lucid. So grammatical. And his way with words! I will forever be indebted to him for the phrase: “supernumerary excrescence.” So useful. OT, but just the other day I was discussing the dire state of contemporary tonal music, which is in large part owing to the fact that many composers view voice-leading as “some kind of supernumerary excrescence,” that is, an unnecessary conceit of composers past.

      1. Oh, don’t even get me started! 😉

        (But seriously, it’s a problem. Not that there aren’t ways to justify parallel motion, etc. I’m thinking more of middle-ground, long-view issues, anyway.)

  9. Dear mr. Dawkins, I’m sorry but I absolutely don’t share your interpretation of Darwin’s statements about the evolution of the sterile caste in eusocial insects. The form of natural selection Darwin invoked, is not kin selection, but kin group selection. He doesn’t say that the fertile and the sterile share the ‘genetic’ base of the characters that are typical of the sterile, but he says that the fertile with sterile ‘siblings’ share with their own parents the tendency to produce a heterogeneous offspring (both fertile and sterile). This sharing makes the colonies with sterile castes to produce similar colonies, which allows natural selection (in the guise of kin group selection) to reward the colonies with sterile castes in the competition against the colonies without sterile castes. In Darwin’s opinion, therefore, the sterile caste doesn’t evolve thanks to a (genetic) benefit to its own members, but thanks to a benefit to the entire colony (and despite a cost to the members of the sterile caste). If Darwin had read Hamilton’s papers, he would have been surprised to know that the inclusive fitness of a sterile is positive and comparable to the one of a fertile! Darwin was not a precursor of inclusive fitness theory: he almost always made use of individual selection, but he sometimes made use of group selection (not only in relation to eusocial insects, and not only in the guise of kin group selection). This was also Hamilton’s opinion (Innate Social Aptitudes of Man, 1975). Sorry for my ugly English (I’m an Italian boy).

  10. Hamilton seems to have regarded:

    – inclusive fitness and kin selection as two different concepts/theories/…

    – inclusive fitness as the general theory and kin selection, group selection etc. as special cases.

    This, I gather from the following quote of Hamilton (1975: “Innate social aptitudes of man…”, quoted from ‘Narrow roads of gene land’, vol 1, p. 336):

    “The usefulness of the ‘inclusive fitness’ approach to social behaviour (i.e. an approach using criteria like (b K-k) > 0) is more general than the ‘group selection’, ‘kin selection’, or ‘reciprocal altruism’ approaches..”

    Hence, there are still different traditions/usages that can be confusing, despite the 12 misunderstandings sorted out.

    1. In the above cited clarification of 12 misunderstandings, Dawkins seems to take ‘kin selection’ as synonymous with ‘inclusive fitness’ (unless I missed something). Anyway, I’m sure this usage is widespread, though clearly not Hamilton’s. I’d count it as the 13th misunderstanding for which Dawkins left space in his clarification paper.

    2. Others call the general theory ‘multilevel selection theroy’ and take ‘inclusive fitness’, ‘kin selection’, ‘reciprocity’ etc. as special cases.

    2.1 As a sub-issue of point 2, some include the term ‘group selection’ into multilevel selection theory as the special case where ‘between-group’ selection prevails over ‘within-group’ selection;
    2.2 while others avoid the term ‘group selection’ alltogether, because it has been too dirreputed in the late 1960s.

    3. Still others seem to speak of ‘group selection’, no matter whether between-group selection or within-group selection prevails. That is, they seem to call the general theory ‘group selection’ rather than ‘multilevel selection’ theory.

    So, there are at least three different terms competing for the label of the genaral theory: inclusive fitness (sensu Hamliton), multilevel selection theory, and group selection.

    Did I get everything wrong?

Leave a Reply