Everyone interested in evolution should read the famous critique of the adaptationist program by Stephen Jay Gould (1941-2002) and Richard Lewontin (1929 - ). Whether you agree with them or not, it's essential that you become informed about the adpatationist-pluralist controversy—also known as the neutral-selectionist controversy. That controversy is still very much a part of the debates over evolution, although the adaptationist side tends to argue that the controversy has been settled. It's no secret that I'm a huge fan of Gould and if I had my druthers I'd make students read every one of his books, including, The Structure of Evolutionary Theory. I'm a pluralist.
My friend John Wilkins, a philosopher, visited St. Mark's Square and the Basilica last year. He's on the opposite side of this debate and he offers the best defense of adaptationism that I've seen in recent years. You should keep an eye on his blog, Evolving Thoughts, it's a must-read for anyone who's serious about evolution. I blogged about John's version of adaptationism [An Adaptationist in Piazza San Marco].
This is a rich topic for undergraduates and there are many potential essay topics.
Michael Ruse Defends Adaptationism
Richard Dawkins' View of Random Genetic Drift
Naked Adaptationism
Gould, S.J. and Lewontin, R.C. (1979) The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme. Proceedings of the Royal Society of London. Series B, Biological Sciences, Vol. 205, No. 1161, The Evolution of Adaptation by Natural Selection (Sep. 21, 1979), pp. 581-598. [AAAS reprint] [printable version]
20 comments :
Thanks once again for your links to the heart of these scientific debates/discussion/brawls.
There nothing more exciting than to dig into the real tussle within science. I think the best science writing opens such to general audiences.
Al
Thanks Larry. It's very civil of you to link to me :-)
I had a piece lately: Does Teleology Hang on in Venice? on this topic.
That's a very interesting paper.
You have a long list of must-reads, Larry. You could write it down and keep it linked on the sidebar. Don't feel obligated to include Aristotle ;-)
John Wilkins:
Where does contingency (particularly the sorts of contingencies that wipe out or bring into dominance whole classes of organisms) fit in the discussion in your linked article? Does it deserve separate treatment?
Though the effects of contingent events can be made to fit into the "adaptation and drift at the same time, we'll figure it out post-hoc" framework you propose, there was just something in the tone of your description that made me think of gradualism (whether or not that's what you intended).
I'm wondering if perhaps the problem is that genetic drift and natural selection a presented as two separate processes in the first place.
Michael M says,
I'm wondering if perhaps the problem is that genetic drift and natural selection are presented as two separate processes in the first place.
Why is this a problem? What's the alternative?
Are you thinking that adaptationist just-so stories are merely random genetic drift by another name?
Larry-
I'm saying that genetic drift and natural selection are really two different descriptors of the same process: population resampling. Natural selection is a statement about the expected value of allele frequencies under resampling, whereas genetic drift is a statement about the variance of allele frequencies under resampling. Both are contained in the distribution of allele frequencies under resampling: the former is the first moment and the latter is the second moment.
Quoting Michael M: ... genetic drift and natural selection are really two different descriptors of the same process: population resampling. [snipped details]...
I like.
Michael M says,
Natural selection is a statement about the expected value of allele frequencies under resampling, whereas genetic drift is a statement about the variance of allele frequencies under resampling.
I don't understand this distinction. Random genetic drift is a process that results in loss or fixation of alleles in a population. Why do you call that "variance"?
Larry, I think your definition of drift (“a process that results in loss or fixation...”) is too limiting by focusing solely on the eventual outcome. Even prior to fixation, the allele frequency will drift: for a neutral pair of alleles the expected F1 frequency will be the same as the parental generation, but the probability of any actual F1 frequency will vary in a statistically describable way (e.g., “50% +/- 3%”), where the variance is a function of the current frequency and population size. The variance is a consequence of the fact that nature is randomly sampling the parent population to construct the F1 population.
If one allele, initially present at 50%, provides a fitness advantage or penalty, the variance due to the random chance will for the most part remain the same, but what will change is the expected frequency. (e.g., 52% +/- 3% for an allele that conveys roughly an 8% advantage). A fitness advantage provides a bias to the drift, but drift (as characterized by the variance in outcome frequency relative to expected frequency) will still be present. IOW, drift describes the variance in the outcome, whereas the relative fitness advantage determines the extent to which the expected frequency will differ from the current frequency.
It’s analogous to pointing out the random chance remains a factor even though the roulette wheel has a house advantage, and that diffusion remains a factor in the movement of ions even though one imposes an electric field.
I think the important point is that it is never one or the other. Rather, the best way to think about what is going on is that NS provides a bias to drift.
A relatively overlooked paper in the recent history of evolutionary biology is the paper by Clutton-Brock and Harvey (1979), "Comparison and Adaptation" in which they list several problems with the adaptationist programme as well as several remedies.
This paper was presented at the same symposium (and subsequently in the same issue of Proc R Soc Lond B), just before Gould and Lewontin gave their paper. Seems like these folks have been talking past each other for some time.
Quoting Michael M: ... genetic drift and natural selection are really two different descriptors of the same process: population resampling. [snipped details]...
I like.
Thanks, but I can't take all the credit for the description. I am most familiar with it from a poster on the now-defunct Richard Dawkins forum, TalkRational, and Rational Skepticism.
Divalent says,
Larry, I think your definition of drift (“a process that results in loss or fixation...”) is too limiting by focusing solely on the eventual outcome.
What ever gave you the idea that's how I define random genetic drift? Here's how I explain the process: Random Genetic Drift.
Natural selection requires differences in fitness between alleles and it leads to adaptation. While you could stretch your definitions to encompass both process, such a stretch doesn't really serve any useful purpose. Adaptation is quite different from evolution by accident and everybody knows it.
Natural selection requires differences in fitness between alleles and it leads to adaptation.
No-one is disputing the fact that natural selection is driven by fitness differentials. However, I think that you need to define "adaptation" before you claim that "natural selection leads to adaptation". As Arlin Stoltzfus (who has commented on your blog before) has has noted in some of his publications, mutation can drive less advantageous mutations to fixation (e.g., Stoltzfus (2006)).
While you could stretch your definitions to encompass both process, such a stretch doesn't really serve any useful purpose.
It's a question of being able to "stretch your definitions to encompass both process [sic]"; it's a question of decomposing resampling (which is technically the only statistical process occurring in evolution) into change in expected value and variance (which are not processes that occur independently of one another in evolution).
For instance, if a population consists of 10 individuals, five of phenotype A and 5 of phenotype B, and the two phenotypes are selectively neutral, the probability that resampling with replacement will yield exactly five of each phenotype is ~.2507. In other words, it is ~3 time as likely that that the next generation will not be comprised of the same proportions of phenotypes that the previous generation was. Yet, if, for example, the next generation were to consist of 8 A's and 2 B's that would not necessarily be evidence of natural selection, because it is possible under genetic drift.
Adaptation is quite different from evolution by accident and everybody knows it.
This is an argumentum ad populum and therefore irrelevant to the discussion.
Michael M says,
This is an argumentum ad populum and therefore irrelevant to the discussion.
Nonsense. When there is widespread scientific consensus about the meaning of certain words and the understanding of certain concepts, then attempts to form an argument based on different definitions and different understandings have to work extra hard to make the case.
So far you haven't provided any evidence that your unpopular view is superior to the consensus.
Scientific definitions are not arbitrary. They are arrived at by consensus and they are, in a very real sense, relevant to the discussion. Quibbling over the meaning of adaptation (or typos) isn't going to get us anywhere.
I understand that it's very difficult to tell the difference between selection and drift. After all, that's one of the main points of the Spandrels paper. But just because it's difficult doesn't mean that the two processes are indistinguishable.
Selection and drift are distinguished by causation. Confusion arises, however, because they have similar statistical properties in the sense that both result in differential persistence of things, such as alleles. However, selection occurs when the persistence of alleles is causally influenced by variation in a phenotype--there is a causal effect of phenotype on fitness (which results in the persistence of alleles). When there is no causal effect of the phenotype influencing fitness, we get drift. This is why you can derive the Price equation to capture both the dynamics of selection as well as the dynamics of drift (e.g., Rice's 2004 book on Evolutionary Theory). While Michael M.'s distinction between mean and variance applies here (since we often model the expected covariance between fitness and phenotype for selection and we model the variance of the covariance between fitness and phenotype for drift), this still requires that we figure out why some phenotypes influence fitness while others do not. It's not just the mathematics, it's also about determining causes.
I never said that natural selection and genetic drift were indistinguishable; I just said that they were two different descriptors of the same process. In fact, I've been nothing but clear that natural selection and genetic are not identical, I said in my first post that natural selection is the change in the expected value of the allele frequency distribution upon resampling and that genetic drift is the variance of the allele frequency distribution under resampling. Expected value and variance are fundamentally different properties of a probability distribution, so I don't understand how you can say that the way I have described natural selection and genetic drift has made them indistinguishable.
The issue is that the differential reproduction of phenotypes "causes" natural selection as much as it "causes" drift. If this weren't the case, advantageous alleles would never be eliminated from a population and deleterious alleles would never be fixed.
By the way, why wasn't a comment that was submitted yesterday published? I might not have actually submitted it but I thought I did, and I was under the impression that Larry publishes all comments unless they are abusive or spam.
Note that Gould is pretty deceptive in The Mismeasure of Man.
Michael M, thanks for acknowledgement, but can we please NOT say "mutation can drive less advantageous mutations to fixation". This isn't correct. Mutational and developmental biases in the introduction of variation can impose a bias on the course of evolution. That is the way to say it.
There is no way to say this using the conventional theory of population-genetic causes or "forces" of evolution, in which a force is something that causes mass-action shifts in frequency and, ultimately, fixation. Mathematically, we can write a dynamic equation in which mutation has the role of a mass-action force that shifts frequencies continuously-- but this process is impotent as a cause of fixation because mutation rates are so small. QED, mutation is not a force that drives alleles to fixation. Fisher & Haldane made this argument 80 years ago and its still in your textbook. Selection and drift can cause fixations, but not mutation.
This "forces" view is inadequate to conceptualize evolutionary causes, and leads to demonstrable errors in reasoning about causation in evolution. Fisher and Haldane made a fundamental error when they concluded, from the fact that mutation cannot drive alleles to fixation, that mutation is not a cause of direction in evolution. Molecular evolutionists make an error when they assume that only neutral evolution can be biased by mutation. Evo-devoists make an error when they imply that biases in variation must be rigid "constraints" in order to be effectual.
A correct theory of causation would have terms that recognize the role, not just of the fixation process, but of the introduction process, and of biases in the introduction process. It is this concept of biases in the introduction of variation, as a cause of bias in evolution, that is absent from theoretical treatments and from the day-to-day reasoning of most evolutionists.
Post a Comment