More Recent Comments

Tuesday, August 31, 2010

The Mutationism Myth, VI: Back to the Future

This is the eighth in a series of postings by guest blogger, Arlin Stoltzfus. You can read the introduction to the series at: Introduction to "The Curious Disconnect". The first part is at: The "Mutationism" Myth I. The Monk's Lost Code and the Great Confusion. The second installment is: Theory vs Theory. The third part is: The Mutationism Myth, II. Revolution. The fourth installment is: The Mutationism Myth: III Foundations of Evolutionary Genetics. Part five is The Mutationism Myth, IV: Mendelian Heterodoxies. The sixth installment was The Mutationism Myth, V: The response to Mendelian heterodoxies.

This is Arlin's last contribution. The entire series has been an excellent introduction to the history of evolutionary theory and the concept of mutationism. There are many ways in which the so-called "Modern" Synthesis has to be revised and extended. One of them is to reinstate the concept of mutationism which was purged from evolutionary theory in the 1940s. If you want to understand why this is important then these articles are the place to start.

The Curious Disconnect

Today on The Curious Disconnect (credits), we wrap up our 6-part series on the Mutationism Myth, and set the stage for the future by locating the primary weakness of the 20th century neo-Darwinian consensus in its theory of variation.

I'd like to thank Larry Moran for hosting this series of posts on Sandwalk. If it still seems like a good idea later in the year, I will continue the Curious Disconnect on my own blog site (to be announced).

The Mutationism Myth, part 6. Back to the Future

The Mutationism Myth, a story about how the discovery of genetics affected evolutionary thought, continues to be part of modern neo-Darwinism's monologue with itself (e.g., Charlesworth and Charlesworth, 2009), being used even by leading thinkers calling for an "Extended" Synthesis (e.g., Pigliucci, 2010 1). Since April, we've been deconstructing the Mutationism Myth by exploring what the early Mendelians actually thought, and how their view was replaced by the Modern Synthesis (MS).

Today we'll take the opportunity to review what we've learned and start unpacking its relevance for the future of evolutionary biology.


In the Mutationism Myth (see part 1 for examples), the founders of genetics misinterpret their discovery, concluding that evolution takes place by large mutational jumps, without selection. The false gospel of these "mutationists" brings on a dark period of confusion and error that lasts until the 1930s, when theoretical population geneticists (Fisher, Haldane and Wright) prove that Mendelian genetics is not only compatible with selection, but provides the missing link that completes Darwin's theory and unites all the biological disciplines. The "Modern Synthesis" combining genetics and selection becomes the foundation for all subsequent evolutionary thought.

As we discovered, the Mutationism Myth isn't very accurate. Heredity is not missing from Darwin's theory; selection is not missing from the Mendelian view. Darwin had a theory of heredity both in the sense of a set of phenomenological laws, and in the sense of a mechanism to account for them (part 2). Both were wrong. The Mendelians synthesized genetics and selection, rejecting Darwin's "Natural Selection" theory due to its dependence on fluctuations or "indefinite variability" (defined by Darwin as the subtle variations that arise anew each generation in response to conditions of life). As we know today, such enrivonment-induced fluctuations are non-heritable.

In part 3, we found that the Mendelians laid the conceptual foundations of evolutionary genetics (later formalized mathematically), while part 4 addressed how their view diverged from Darwinian orthodoxy. The Mendelians assumed that new hereditary variants arise rarely and discretely, by mutations whose effects may be large or small. Each new mutation is likely to be rejected, but it may be accepted by chance, especially if it improves fitness. Because, in this view, change depends on discrete events of mutation, the Mendelians (part 4) considered the process of mutation to be a source of initiative, discontinuity, creativity and direction in evolution (Stoltzfus, 2006). This view expanded the role of variation well beyond the subordinate role of raw materials that Darwin had imagined.

The Mendelians were unable to convince naturalists (the majority of their biologist peers) to accept their new view of evolution, nor even a new view of inheritance. Many naturalists remained wedded to Lamarckian and Darwinian views of "soft inheritance".

As we found out in part 5, the "Modern Synthesis" (modern neo-Darwinism) claimed to reconcile Darwin's own view with genetics, though it quietly ignored Darwin's errors while depicting the Mendelians as foolish "saltationists", dismissing their "lucky mutant" view and their ideas about the role of mutation in evolution. In the MS view, each species has a "gene pool" that automatically soaks up and "maintains" hereditary variation, providing abundant "raw materials" for adaptation. The key innovations of this view were to define "evolution" as "shifting gene frequencies" in the "gene pool", to erase the link between "Darwinism" and Darwin's own theory of soft inheritance, and to develop a theory of causation in terms of population-genetic "forces", in which continuous shifts in allele frequencies are the common currency of causation. The new theory put mutation in a subordinate position of supplying infinitesimal "raw materials" for selection. As a result, the MS created a consensus where the Mendelians had failed: naturalists such as Ernst Mayr found that they could accept Mendelian genetics without giving up adaptationist preconceptions.

A bright line

The backdrop for this whole discussion (in case you missed it) is that the MS is strikingly wrong in its neo-Darwinian departures from the Mendelian view. I've implied this several times, and perhaps I've waved my hands and pointed vaguely to mountains of molecular evidence contradicting the MS, but I haven't made this point perfectly clear.

In a moment, I will do that, but first I want to make clear what is at stake.

The Mendelians allowed that evolutionary change could be initiated by an event of mutation, and they interpreted this to mean that mutation was (to an unknown degree) a source of initiative, discontinuity, creativity and direction in evolution. The MS represents a very deliberate rejection of this view, and proposes instead that evolution is a complex sorting out of available variation to achieve a new multi-locus equilibrium, literally by "shifting gene frequencies" in the "gene pool". The rate of evolution, in this view, does not depend on mutation, which merely supplies the "gene pool" with variation; evolution is not shaped by mutation, which is the "ultimate" source of variation, but not the proximate source.

When I made this distinction at a 2007 symposium in honor of W. Ford Doolittle, Joe Felsenstein was in the audience and pointed out that, while Fisher may have looked at things in this way, Wright's stochastic view took into account random events, like new mutations. It's true that Wright's "shifting balance" model assigns a prominent role to random genetic drift, while Fisher's view was deterministic. However, these are just two different flavors of the same "shifting gene frequencies" paradigm: neither view incorporates new mutations. The absence of new mutations from Wright's shifting balance process is apparent from the fact that Patrick Phillips (1996) extended it to include a new starting phase ("phase 0") of "waiting for a compensatory mutation".

The fact that contemporary evolutionary biologists, for the most part, don't understand this aspect of their intellectual heritage is not evidence of a cover-up. Scientists don't get much chance to learn history. The history that they absorb is mainly from stories that appear in scientific writings, like the Mutationism Myth and the Essentialism Story, stories that represent Synthesis Historiography (Amundsen, 2005), the discipline of telling history in ways that make things turn out right for the Modern Synthesis. Synthesis Historiography teaches us that "saltationism" (Mayr's pejorative term for the Mendelian view) and other alternatives to neo-Darwinism are nonsensical, "doomed rivals", supported only by "typologists", creationists, vitalists and other crazies. That is, Synthesis Historiography teaches the TINA doctrine: There Is No Alternative.

As contemporary research drifts away from the "gene pool" theory and the Darwinian doctrines of the MS, each evolutionary biologist remains confident that, due to the TINA doctrine, his own view must be "neo-Darwinian". In reality, alternatives are being explored with increasing vigor in molecular evolution, evo-devo, and evolutionary genetics.

A few folks today are in the reverse situation of being familiar with MS orthodoxy, but not with recent research. Dawkins (2007) stakes his critique of a book by "intelligent design" creationist Michael Behe entirely on his faith in the gene pool theory. Behe claims, in effect, that there was not sufficient time for all the mutations needed to account for evolution. Dawkins responds by attacking the premise that evolutionary rates depend on mutation rates:

"If correct, Behe's calculations would at a stroke confound generations of mathematical geneticists, who have repeatedly shown that evolutionary rates are not limited by mutation. Single-handedly, Behe is taking on Ronald Fisher, Sewall Wright, J.B.S. Haldane, Theodosius Dobzhansky, Richard Lewontin, John Maynard Smith and hundreds of their talented co-workers and intellectual descendants. Notwithstanding the inconvenient existence of dogs, cabbages and pouter pigeons, the entire corpus of mathematical genetics, from 1930 to today, is flat wrong. Michael Behe, the disowned biochemist of Lehigh University, is the only one who has done his sums right. You think? The best way to find out is for Behe to submit a mathematical paper to The Journal of Theoretical Biology, say, or The American Naturalist, whose editors would send it to qualified referees."

With his signature over-the-top rhetoric, Dawkins insists that "mathematical genetics" has proven that evolutionary rates are not limited by mutation. Allowing for some exaggeration, this is an accurate representation of MS orthodoxy ca. 1959, the approximate vintage of Dawkins's views. If Mayr had been alive, he might have said the same thing.

Meanwhile, no one who has been active in evolutionary genetics research in the past 15 years would represent the current state of knowledge in this way. If you want to know what a contemporary researcher would say, take a look at the most recent issue of Evolution in which the article by Douglas Futuyma (famous for his evolution textbook) gives many examples of how evolutionists (including himself) repeated the doctrine that mutation does not "limit" evolution, but argues that we are no longer making this dubious assumption. Another example would be the piece by Ronny Woodruff and James Thompson (1998) that introduces their symposium volume on Mutation and Evolution.2

Yet the MS and its "gene pool" theory have left their mark on evolutionary biology, even if the MS itself has largely disappeared from the collective memory of researchers. One indelible mark is what Gillespie calls "The Great Obsession" of population genetics to understand the "maintenance of variation", but that's a story for another day.

Another indelible mark is the long absence of mutationist models of "adaptation", a topic that has blossomed just in the last dozen years. Allen Orr has achieved well deserved fame for his innovations in this area, and we'll discuss his work briefly in the next section. For now, let us note how other researchers have pointed out the absence of such models:

"Almost every theoretical model in population genetics can be classified into one of two major types. In one type of model, mutations with stipulated selective effects are assumed to be present in the population as an initial condition . . . The second major type of models does allow mutations to occur at random intervals of time, but the mutations are assumed to be selectively neutral or nearly neutral." (Hartl & Taubes, 1998)

"The process of adaptation occurs on two timescales. In the short term, natural selection merely sorts the variation already present in a population, whereas in the longer term genotypes quite different from any that were initially present evolve through the cumulation of new mutations. The first process is described by the mathematical theory of population genetics. However, this theory begins by defining a fixed set of genotypes and cannot provide a satisfactory analysis of the second process because it does not permit any genuinely new type to arise. " (Yedid and Bell, 2002)

These authors are not trying to make a point about history or about the Modern Synthesis: they are simply claiming the novelty of their own models of adaptation that incorporate new mutations. And what they are saying is that the paradigm of 20th-century population genetics is "shifting gene frequencies": overwhelmingly, it's a body of theory about what happens to the variation that is present in a population as an initial condition, not about a larger-scale process in which there are new beneficial mutations.3

One small step for a phage, one giant leap for evolutionary biology

The actual role of mutation in evolution is not what is theorized in the MS. Many arguments could be made to support this contention, but I'm going to make just one argument drawing on one source, namely Rokyta, et al., 2005. I choose this argument because it is particularly compelling and concise. My argument addresses the lucky mutant view of initiative or (to put it another way) dynamics.

Rokyta, et al. is a study of parallel evolution in an experimental population of the bacteriophage phiX174, published in Nature Genetics. It was hailed as "the first empirical test of an evolutionary theory" (Bull & Otto, 2005), where the theory in question is Orr's (2002) ingenious extension of Gillespie's (1984) "mutational landscape" model to take into account predictions of extreme value theory.4

In spite of the fancy name, the "mutational landscape" model of sequence evolution is simple. Rather than considering all conceivable evolutionary changes from a starting sequence, we simplify the problem by considering only changes that occur via 1-bp mutations. That set of possibilities, by definition, is the "mutational landscape" or (my preferred term) the "evolutionary horizon". Each change will shift the evolutionary horizon, but it's easy to recompute the horizon, because it's easy to enumerate (theoretically) all the alternative sequences.

We are going to make this a model of beneficial changes ("adaptation"). A beneficial mutation is introduced into the population of N individuals at some total rate Nu, and faces acceptance with a probability of 2s, based on the classic formula p = 2s for the probability of fixation of a new beneficial mutation.5 For beneficial substitution i with selection coefficient si, the probability6 is just Nu*2si. If we divide an individual Nu*2si by the sum of all such values on the horizon, we get a normalized probability: the probability that the next step in our evolving system is step i. The factor Nu*2 is the same for every step, so it cancels out: only the si values matter. To evolve our sequence, we just sample from this probability distribution of possible steps, then recompute the new evolutionary horizon in preparation for the next step. Easy! 7

From past experiments, Rokyta, et al. know which steps on the horizon are beneficial, and they even know the selection coefficients. They know that sometimes, the same evolutionary steps happen in parallel, in replicate phage populations. They can compare the observed pattern of parallel evolution with the pattern predicted from theory.

Now, the preceding description suggests something fascinating: the cutting edge of evolutionary genetics today, with papers that get published in Nature Genetics with commentaries, uses experimental systems to explore the "lucky mutant" view of parallel evolution.

But the story gets even better. Rokyta, et al actually reject Orr's model, in its original version. They find more parallel evolution than expected. Why? Because the model treats all mutation rates equally. Note above that we canceled out mutation rates on the grounds that they are all the same. But that's not realistic. Some mutations are more likely than others, and this will affect the rate at which they are introduced into the population and subjected to acceptance or rejection. The more heterogeneity in rates of mutation, the more parallel evolution. Rokyta, et al. found that if they revised the model to take into account transition:transversion bias (I think it's about 5-or 6-fold under the experimental conditions), then the predicted amount of parallelism matched the observed amount.

Just let that soak in for a moment. We have an experimental study and a precise model. Evolution in this model is characterized by origin-fixation dynamics, dependent on the rate of mutational introduction of new alleles, and on their probability of fixation. Both factors affect the outcome of evolution; both factors affect the chance of parallelism. The experimental study eliminates (statistically) a model that lacks mutational bias in the introduction of new alleles. Thus the study clearly illustrates the dual causation of evolutionary change, in regard to its dynamics.

Back to the future

The MS is wrong, and not in a small way: it's wrong because reality just looks too much like the antithesis of the MS, i.e., the mutationist view. For instance, as we found out in part 4, Vavilov (1922; see Stoltzfus, 2006) understood the dual causation of parallel evolution, including the role of parallel variation. By contrast, Mayr famously said that the search for homologous genes or homologous mutations was foolish.

This mistaken prediction is repeated ad nauseam in the evo-devo literature. If you have been following along, you now understand why Mayr would make such a prediction. The MS makes substantive claims about evolution, among which are the claim that, while mutation is ultimately necessary to keep the "gene pool" from drying up, selection doesn't need to wait for a new mutation, but draws together a multi-locus optimum from the abundance of raw materials in the gene pool; "evolution" is so far removed from the process of mutation, with so many complex dynamic processes interceding, that the outcome of evolution does not depend on specific events of mutation. If evolution really were like that, parallel mutation would be unimportant. That is, Mayr's prediction accurately reflects the logic of the MS. But as the Rokyta, et al study (and many others) show, the prediction is not fulfilled.

According to Dawkins, "the entire corpus of mathematical genetics" from 1930 to "today" (i.e., about 1959, for Dawkins) would be "flat wrong" if one accepts the premise that evolution depends on new mutations, or that it is limited by the mutation rate. While this view is not often defended, that isn't because it's Dawkins's own personal opinion. Dawkins is accurately characterizing a theory that makes substantive claims about the world, a theory that most of us have forgotten. One of these claims is that "evolution" can be represented mathematically as a process of shifting the frequencies of alleles already present in an initial population, without new mutation; sometimes this doctrine is invoked by saying that "macroevolution" can be extrapolated from "microevolution".

If evolution actually worked like this, then evolutionary change would not exhibit a dependence on the rate of mutation, and Dawkins would be right in his criticism of Behe. But this is wrong. In fact, the dependence is so sensitive that effects of only a few-fold are noticeable, as the Rokyta, et al study (and many others) shows.

I'm not going to mince words. The MS is wrong, and not in a small way: reality looks too much like the mutationist view that we (the scientific community) rejected when we bought into the MS. We need another theory1, perhaps several others.

The road less traveled

What's wrong about the MS, and what its replacement(s) must replace, is its theory of the role of variation in evolution. In future posts on the Curious Disconnect, I intend to focus on this issue. The Mutationism Myth suggests a lesson about how to develop (or rather, how not to develop) a theory of variation.

Darwin knew that hereditary variation played a vital role in evolution. He studied the subject intensely. He found that organisms vary in many different ways, and on many scales, but the evidence on heredity was bewildering and inconclusive. Lacking the means to derive a mechanism of evolution by reasoning upward from genetics, Darwin reasoned downwards from his premises that 1) organisms are exquisitely and pervasively adapted to their niches, 2) selection must have played some role in this, and 3) Mother Nature never makes a jump. Gould argues that Darwin's willingness to posit precise restrictions on variation was a stroke of genius.8 Darwin knew that discrete "sports" (mutants) could be heritable, but he discounted them: they could not make his theory work as desired. Instead, he staked his "natural selection" theory on the heritability of fluctuations, because they were infinitesimal, indefinite (unbiased), and "everywhere present", being induced in abundance whenever organisms encountered altered "conditions of life". Inferring the heritability of fluctuations completed his theory and made it work.

But it was wrong: the fluctuations that made Darwin's theory work are non-heritable, as the Mendelians discovered.

The architects of the MS tried again, with advantages unavailable to Darwin. Not only did they know genetics, they had some mathematical tools to work out unforeseeable implications of genetic concepts. However, they didn't have the knowledge to distinguish among different, genetically consistent modes of evolution. They had to fill in this gap somehow. Their downwards Darwinian reasoning and their upwards Mendelian reasoning met in the middle with the "gene pool": a theory of population genetics that would supply abundant, infinitesimal, "random" variations, in order to rationalize their commitment to the same premises Darwin accepted. That was the genius of the MS.

But again, it was wrong.

If we look at Darwinism in Popperian terms, as a theory1 that takes risks and generates potentially falsifiable claims, then (counterintuitively) it is largely a theory of the role of variation in evolution. The claims that selection is "important", and that it has some inalienable role in adaptation, carry little risk and have been widely accepted for 150 years. By contrast, the restrictions that Darwinism places on variation, in order to make it a subordinate factor that supplies "raw material" to selection, are risky and controversial, e.g., the claim that variation is random with respect to the direction of evolution, or that the rate of evolution does not depend on the rate of mutation, or the "gradualist" claim that variation is not a source of discontinuity. The architects of the MS invested the "gene pool" with nearly magical properties in order to improve the prospects for adaptation. Problematic claims about the role of variation are, and have been for 150 years, the overwhelming basis for scientific criticism of Darwinism.

And this problematic view of variation is based on reasoning from the premise that organisms are exquisitely and pervasively adapted to their niches, to the conclusion that variation must play just the right role of supplying abundant raw materials to make this possible. I believe that there is something fundamentally wrong with this mode of reasoning. Perhaps it betrays a kind of subconscious Panglossian agenda. Every time I give a lecture on mutation-biased evolution, someone suggests that perhaps the mutation biases themselves are adaptive, as though this inference could restore one's faith that everything turns out for the best, and that "the ultimate source of explanation in biology is the principle of natural selection" (Ayala, 1970). Remarkably, the evo-devo-inspired view that seems destined for inclusion in the emerging "Extended Synthesis" is headed down much the same path, with a focus on the idea that the process of variation has been jiggered to make things turn out right for adaptation. What's revealing about this new view is how little attention its proponents have paid to understanding precisely, in terms of population-genetic causation, how the process of variation shapes evolution, before jumping ahead to the shadowy inference that the process of variation itself was shaped by selection for this very role.

We are not going to go down that same road here on the Curious Disconnect, which should make things all the more interesting.


Ayala, F. J. 1970. Teleological Explanations in Evolutionary Biology. Philosophy of Science 37:1-15.

Bull, J. J., and S. P. Otto. 2005. The first steps in adaptive evolution. Nat Genet 37:342-343.

Charlesworth, B., and D. Charlesworth. 2009. Darwin and genetics. Genetics 183:757-766.

Dawkins, R. 2007. Review: The Edge of Evolution. Pp. 2. International Herald Tribune, Paris.

Gould, S. J. 2002. The Structure of Evolutionary Theory. Harvard University Press, Cambridge, Massachusetts.

Hartl, D. L., and C. H. Taubes. 1998. Towards a theory of evolutionary adaptation. Genetica 103:525-533.

Medawar, P. B. 1967. The Art of the Soluble. Methuen and Co., London.

Nei, M. 2007. The new mutation theory of phenotypic evolution. Proc Natl Acad Sci U S A 104:12235-12242.

Orr, H. A. 2002. The population genetics of adaptation: the adaptation of DNA sequences. Evolution Int J Org Evolution 56:1317-1330.

Phillips, P.C. 1996. Waiting for a compensatory mutation: phase zero of the shifting-balance process. Genetical Research, Cambridge 67:271-283.

Rokyta, D. R., P. Joyce, S. B. Caudle, and H. A. Wichman. 2005. An empirical test of the mutational landscape model of adaptation using a single-stranded DNA virus. Nat Genet 37:441-444.

Woodruff, R. C., and J. D. Thompson. 1998. Preface in R. C. Woodruff, and J. D. Thompson, eds. Mutation and Evolution. Kluwer, Dordrecht, The Netherlands.

Yedid, G., and G. Bell. 2002. Macroevolution simulated with autonomously replicating computer programs. Nature 420:810-812.


1 Pigliucci, along with Gerd Muller, edited a book on "the extended Synthesis" with papers from a select group of thinkers who were invited in July, 2008 to a special meeting in Altenberg, Austria. The book is now available in paperback:

2 from p. 1 "Although mutation is a key parameter in the genetics of populations, the role of mutation as an evolutionary factor has been debated since the time of Darwin. Early geneticists, who held to the 'classical' view of the genome as being homogeneous with occasional mutant alleles, saw new mutation as a major determining force in adaptive change. When the classical view was replaced with the 'balance' view of the genome, i.e., highly heterogeneous, pre-existing variation became more important as the resource on which selection would act. Many, therefore, began to disregard new mutation as a significant force in evolution, since the level of genetic diversity is already so high that new mutants would generally be expected to add little to that resource . . . Mechanisms responsible for maintaining levels of genetic diversity became the focus of attention, and mutation pressure is now thought by many to have only minor significance, especially when compared to selection, recombination, gene flow, and similar factors. We think this position, like the classical view, is too extreme. While there can be little doubt that mutation per se is not the principle driving force it was once believed to be for phenotypic evolution, we see growing evidence that its role is under-appreciated in important situations. The rate and pattern of mutation can be influenctial variables in adaptive responses, and the role of mutation in evolution deserves to be reexamined."

3 Orr (2002) notes the absence of such models by making a far more sweeping claim that population genetics has ignored, not just new-mutations models of adaptation, but all models of adaptation, and instead has focused on neutral and deleterious alleles. That is an odd thing to say, given that the quantitative genetics of adaptation have been a topic for a long time. In any case, here is what Orr says: "Evolutionary biologists are nearly unanimous in thinking that adaptation by natural selection explains most phenotypic evolution within species as well as most morphological, physiological, and behavioral differences between species. But until recently, the mathematical theory of population genetics has had suprisingly little to say about adaptation. Instead, population genetics has, for both historical and technical resasons, focussed on the fates of neutral and deleterious alleles. The result is a curious disconnect between the verbal theory that sits at the heart of neo-Darwinism and the mathematical content of most evolutionary genetics. "

4 Also known as the theory of records—"record" in the sense of "pinnacle of achievement". Given a series of records, such as the world record in the long-jump, what's the interval of time to the next record, and by how much will it break the previous record? The theory of records addresses such questions. Can you see how this would be useful to make a predictive theory of adaptation?

5 Rokyta, et al. used a different formula for the probability of fixation, because the classic approximation only works for s << 1, whereas the phiX174 populations experience very large s, sometimes s > 1.

6 Formally Nu*2si is not a probability but a steady-state rate (e.g., for an infinite-alleles model). If we treat it as an instantaneous rate, and then compare it to all other instantaneous rates, this makes it a relative probability of choosing step i over a short interval.

7 For our present purposes, we don't need to explain Orr's addition to this model, which was a theory of the distribution of the favorable s values under generalized assumptions (oddly, the commentators on Rokyta, et al. did not mention that Orr's theory wasn't really needed, and that the study really was a test of the mutational landscape model itself).

8 Gould (2002, p. 140) is not endorsing Darwin's error about fluctuation. Darwin's followers think of that mistake as a trivial detail. Instead Gould is endorsing a more general inference. Here is what he writes. "Darwin reasoned that natural selection can only play such a role [as exclusive source of creativity and direction] if evolution obeys two crucial conditions: (1) if nothing about the provision of raw materials—that is, the sources of variation—imparts direction to evolutionary change; and (2) if change occurs by a long and insensible series of intermediary steps, each superintended by natural selection—so that "creativity" or "direction" can arise by the summation of increments.

Under these provisos, variation becomes raw material only—an isotropic sphere of potential about the modal form of a species. Natural selection, by superintending the differential preservation of a biased region from this sphere in each generation, and by summing up (over countless repetitions) the tiny changes thus produced in each episode, can manufacture substantial, directional change. What else but natural selection could be called 'creative,' or direction-giving, in such a process? As long as variation only supplies raw material; as long as change accretes in an insensibly gradual manner; and as long as the reproductive advantages of certain individuals provide the statistical source of change; then natural selection must be construed as the directional cause of evolutionary modification.

These conditions are stringent; and they cannot be construed as vague, unconstraining, or too far in the distance to matter. In fact, I would argue that the single most brilliant (and daring) stroke in Darwin's entire theory lay in his willingness to assert a set of precise and stringent requirements for variation—all in complete ignorance of the mechanics of heredity. Darwin understood that if any of these claims failed, natural selection could not be a creative force, and the theory of natural selection would collapse. "

Credits: The Curious Disconnect is the blog of evolutionary biologist Arlin Stoltzfus, available at An updated version of the post below will be maintained at (Arlin Stoltzfus, ©2010)


-DG said...
This comment has been removed by the author.
-DG said...

Another great post Arlin. I'm really looking forward to future posts when you construct what the successor theory/theories to the MS should look like.

I am wondering how many people will walk away from this post thinking you are saying Behe was right. He was still very, very wrong just not for the reasons that Dawkins gave.

Arlin said...


I'm not concerned about that, and I don't think you should be either.

One reason I choose not to be concerned is that this would be a distorting influence. Could you have an intelligent discussion of US foreign policy if one of the criteria is that you can't say anything that might provide ammunition to those outside the US?

The second reason not to worry is that creationists as a group have not developed the intellectual skills to sort out the truly devastating critiques of neo-Darwinism from the truly lame ones. Apparently they all look the same to a creationist. Although I'm not proud of it, CD has a Flesch-Kincaid readability score ( of about 31 (grade level of about 15). The readability is worse than the Harvard Law Review.

I would get more attention from creationists if I called Dawkins a poopy-pants.


-DG said...


That's very true, and I'm not too worried. These have been dense but very informative reads, even for someone with my background. Glad to see a new one posted and looking forward to future ones.

Steve Bunnell said...

I'm having some trouble seeing the kernel of your argument, so please let me know if I am on track here:

1) The mechanisms that produce variation are poorly understood and are likely to play crucial roles in evolution.

2) Infinitesimal changes are probably not the driving force in the development of major evolutionary innovations.

From the signal transduction perspective, this seems apparent from studying the family trees of modular domains and their incorporation into novel forms.

In short- To what extent was metazoan evolution facilitated by the relatively late evolution of tyrosine kinases that could be exapted for roles in cell-cell communication? Would it have been possible to evolve arms and/or legs without a prior duplication of the hox cluster driving specification along the body axis? To what extent was the diversification of the vertebrate immune system dependent on gene duplications that provided the distinct versions of the kinases and adaptors that contribute to the character of distinct immune cell lineages? To what extent do such unlikely events drive macroevolution?

Am I remotely on track?

Jim Menegay said...

An excellent summation of the story to date. I'm looking forward to more, whatever form the continuation takes.

If the choice is between publishing the historical analysis you already have, or trying to tack on some good and original "neo-mutationist" theory to illustrate why the history is important - well, I would vote for quickly publishing the history and letting the theory arrive on its own schedule.

Although I am an amateur, I found your historical account very energizing, and I suspect that there are plenty of grad students and post-docs who would feel the same. Energize them and then let them get some of the glory for constructing new theory.

Corneel said...

Thank you for another great post, Arlin. I am sympathetic to your ideas, but I think it may be hard to generalise your ideas. To explain this, I should return to my previous comments on the "gene pool"in your previous post.
Neither Darwin, nor the founders of the MS were aware of the mechanistic basis of variation. Whenever they mentioned variation, they meant phenotypic variation. One of the observations that shaped their thinking was that, no matter what trait you were considering, or what organism, there often was heritable variation present. I think this is why their theoretical models start off with a generous amount of variation (the gene pool).
It is true that there are organisms to which this does not apply. Phages have a life history that consists of a series of bottlenecks, and mutation will be an important factor in their evolution, but your phage example works, because it is a simplified model
In most plants and animals there are numerous mutations present that affect the phenotypic trait that is under selection. Transition:transversion bias and other mechanistic aspects of mutation are unlikely to influence the direction or magnitude of the phenotypic effect of any mutation. This is why the gene pool is a reasonable approximation of most systems.
Yes, I am playing devil's advocate :-)

Larry Moran said...

Corneel says,

In most plants and animals there are numerous mutations present that affect the phenotypic trait that is under selection.

Isn't this a circular argument? When you detect a trait that's "under selection" it's because there are variants with different fitness values.

What about all those traits that aren't under selection because there is no variant available? Most plants would benefit from having a better Rubisco, for example, but it ain't happening. Many plants would benefit from resistance to RoundUp but very few species have developed resistance and, even then, only in localized areas.

There are all kinds of possible beneficial mutations that would probably sweep to fixation if they had ever occurred. But they didn't.

Conversely, many of the phenotypic characteristics of modern species are probably just due to the accidental fixation of a nearly neutral mutation that just happened to occur in the lineage.

-DG said...


Excellent point. You can't have a variant under selection unless that variant is first produced via mutation and not all mutations are equally likely to occur (the evolutionary horizon). This is just as true in animals as in any other lineage under consideration.

I personally believe, with some evidence and a lot of opinion, that likely the majority of phenotypic differences observed within most lineages of plants, animals, and other large multi-cellular organisms were the result of fixation of neutral variants and are not adaptations at all.

Arlin said...


Thanks for playing Devil's Advocate.

When I witness a seminar in which a researcher is describing an instance of presumptive adaptive evolution, e.g., improved color vision, that has been mapped to the genetic level, I ask myself a question: does this set of genomic changes-- changes identified post-hoc as mediators of adaptation-- exhibit transition-transversion bias (or other kinetic mutation biases that I know about)?

I ask this question of myself, because I have a particular expectation, based on how I think about evolution. What do we expect based on the "gene pool" theory?


Corneel said...

Larry said:

Isn't this a circular argument? When you detect a trait that's "under selection" it's because there are variants with different fitness values

No it is not a circular argument. There can be selection without a response, e.g. when all phenotypic variation is environmentally induced.

There are all kinds of possible beneficial mutations that would probably sweep to fixation if they had ever occurred. But they didn't.

Is that how evolution goes? occasional mutations increasing fitness step by step? No, evolutionary novelty occurs in response to ecological opportunity (e.g. during adaptive radiation) or large genomic rearrangements (e.g. polyploidisation). At those moments, large quantities of fitness variation present themselves to natural selection.
A counterexample: In our lab we selected Drosophila for resistance to DDT, and they responded within generations. In many cases, the variation is present when selection is imposed.

Arlin said:

...does this set of genomic changes-- changes identified post-hoc as mediators of adaptation-- exhibit transition-transversion bias (or other kinetic mutation biases that I know about)? I ask this question of myself, because I have a particular expectation, based on how I think about evolution. What do we expect based on the "gene pool" theory?

I don't think "gene pool" theory has anything to say on this issue, unless there is any difference in the expected phenotypic effects of transitions and those of transversion. I don't think there is any a priori reason to suspect there is.

Arlin said...

Steve Bunnell--

Thanks for your comment. Its always good to try to take something complex and distill it into a few points.

I certainly agree with proposition #1, as would anyone in the "Extended Synthesis" camp and probably most other evolutionary biologists, because "crucial role" could mean many different things. For instance, some would agree with #1 on the grounds that individual "chance" mutations play a "crucial role" in introducing "contingency" in evolution (search "Lenski-R" in PubMed to access the literature on this). But I am more interested in understanding what is predictable, rather than what is unpredictable.

I'll respond to #2 separately.

Arlin said...

Steve B--

In regard to #2, this takes a strong non-traditional position on a fascinating issue-- fascinating because the architects of neo-Darwinism claimed to have settled this issue of gradualism, yet today's evolutionists are questioning the Darwinian answer.

The way that #2 invokes "driving force" and "major" makes it hard for me to evaluate, though. Can we re-phrase this so that its more tractable and has a clearer relation to underlying theoretical positions?

Your subsequent comments are more concrete, and provide some leverage for clarification. For instance, you ask "To what extent" was X (some evolutionary outcome or pattern) "facilitated by" or "depended on" Y (some mutational event or condition)? Can you tell me what is a possible answer, and how that answer would be determined?

Arlin said...

Corneel says

I don't think "gene pool" theory has anything to say on this issue, unless there is any difference in the expected phenotypic effects of transitions and those of transversion.

I agree that this is the correct interpretation of the MS theory. In Darwinism, selection is the source of bias or non-randomness, while variation is "random".

I don't think there is any a priori reason to suspect there is.

In the MS view, any systematic tendency for some type of change X to be favored by the process of evolution would be evidence that changes of type X are systematically favored by selection. While there may not be an a priori difference, Darwinists would seek out properties of transitions that would explain why they are somehow "better" than transversions.

This is, in fact, what Darwinists do when confronted with the evidence that transitions happen more commonly than transversions-- they propose explanations in which transition are more likely to be beneficial, or less likely to be deleterious.

Paul McBride said...

Corneel says: One of the observations that shaped [MS theorists'] thinking was that, no matter what trait you were considering, or what organism, there often was heritable variation present... In most plants and animals there are numerous mutations present that affect the phenotypic trait that is under selection... This is why the gene pool is a reasonable approximation of most systems.

Yet, there are also plenty of phenotypic traits in all populations that, conversely, do not have abundant variation, and plenty of traits that feature discontinous, rather than infinitesmal, variation. We are biased towards examining those copiously variant traits that lack obvious constraint. Constraints, and as Arlin points out, mutational biases are important features of evolutionary trajectories.

Ultimately, the view of a gene pool abound with variation for selection to act upon only gives us a biased and proximate view of evolution. You mention adaptive radiations. In such radiations we typically focus on examining how a presumably-selected-for trait has varied such that a set of species/subspecies fill some vacant niches. Too often the story stops there. Left unexplained is why other niches are left unfilled, why the trait in question and not another trait was the subject of sufficient variation, why one species and not another radiated etc.

The MS view of the gene pool is one that is prone to marginalising important features of biology, such as stasis and constraint - by assuming such things represent the absence of selection rather than the absence of suitable variation, and thereby forcing the view that change is initiated by environmental (i.e. external) factors.

Corneel said...

Arlin said:

While there may not be an a priori difference, Darwinists would seek out properties of transitions that would explain why they are somehow "better" than transversions.

I agree this view is wrong. Selected alleles will reflect any bias in the base population of alleles.
However, that does not invalidate the gene pool as a useful concept. It simply requires that we are aware of any non-adaptive biases that arise because of other evolutionary forces (such as mutation) The impression I got from your blogposts was that the concept of a gene pool itself was fundamentally flawed. I am still unconvinced that this is true.

Corneel said...

Paul McBride said...
The MS view of the gene pool is one that is prone to marginalising important features of biology, such as stasis and constraint - by assuming such things represent the absence of selection rather than the absence of suitable variation, and thereby forcing the view that change is initiated.

I disagree. The MS allows any absence of change to be either caused by the absence of an evolutionary force (including drift and mutation), or the absence of heritable variation (including those caused by constraints of any sort). If there are biologists that hold the view that you described, than the fault lies with their preconceptions, not with the theory.

Arlin said...

Corneel and Paul M--

With all due respect, I think your discussion is going to go off the rails for lack of agreement on what the MS represents. This happens to nearly every such discussion, not just online ones. Person A says "The MS is wrong because it says X" and person B says "The MS is right, and it doesn't say X". This has been going on for 30 years.

The solution to this problem is to focus on documentable claims. We don't have comprehensive survey data on evolutionists' beliefs, so proving claims of the form "{ all | almost all | most } evolutionists believe X" would require a major research project. The MS has never been axiomatized in an agreed-upon way, so claims of the form "the MS says X" cannot be documented in a direct way.

So, how do we proceed? In Stoltzfus (2006) and elsewhere, I frequently rely on arguments of the form "the architects of the MS repeatedly claimed X", documenting this with quotations from original sources. That's one way to proceed. There may be others.

If you were wondering, "why is Arlin focusing on these quirky notions of 'creativity', 'direction', etc, and not asking addressing vital questions about the relative importance of selection vs. drift?", the reason is that the architects of the MS took a clear position on one, but not on the other.

I'm not sure whether there is a documentable MS position on the cause of non-change.

Arlin said...


The architects of the MS argued that, because the "gene pool" is a kind of buffer or sink, it creates a wall of separation between "mutation" and "evolution". They argued that, because of this, the rate and direction of evolution will be found not to depend on the rate of mutation; mutation merely supplies the "gene pool" with variation; it is the ultimate but not the proximate source of change; individual events of mutation are unimportant (all these claims are sourced in Stoltzfus, 2006). But we know that, in fact, the rate and direction of evolution does depend on mutation, and individual events of mutation are important.

That is what is wrong with the "gene pool" idea: it leads to (or perhaps merely rationalizes) conclusions that we know to be false.

Now, can you tell me what is right about the "gene pool" idea? Every population has some diversity D from 0 to 1 at every locus. Where does it get us to say that this is a "gene pool" that soaks up variation and maintains it for use by "selection"?

For 70 years, neo-Darwinian population geneticists have pursued what Gillespie calls The Great Obsession, to understand how & why variation is "maintained" in the "gene pool". For 70 years, they have failed to resolve their self-imposed conundrum. I've never understood this. Why not just move on?

Corneel said...

Arlin said:
Now, can you tell me what is right about the "gene pool" idea? Every population has some diversity D from 0 to 1 at every locus. Where does it get us to say that this is a "gene pool" that soaks up variation and maintains it for use by "selection"?

First, let me emphasize that I have no problem with a more prominent role for mutation in the rate and direction of evolution. For me, it is plausible that biases in the introduction of genetic variation may have a substantial effect on the outcome of evolutionary processes.

Second, the concept of the gene pool as a "sponge" sounds like group selection to me, and I agree that sounds like a bogus argument.

But, come on, the presence of variation in natural populations is an interesting topic. Most populations harbor large amounts of genetic variation, sometimes because of bizarre and surprising mechanisms, and that is as fascinating as evolutionary change. In addition, it is useful knowledge, e.g. in medical sciences, where thinking is still very typological. The gene pool concept was not just devised to annoy the mutationists, it describes biological reality in many organisms.