Chance and Selection

Darwin’s conforming of his theory to the old vera causa ideal shows that the theory of natural selection is probabilistic not because it introduces a probabilistic law or principle, but because it invokes a probabilistic cause, natural selection, definable as nonfortuitous differential reproduction of hereditary variants.

Chance features twice in this causal process. The generation of hereditary variants may be a matter of chance; but their subsequent populational fate is not; for their physical property differences are sources of causal bias giving them different chances of survival and reproduction. This distinguishes selection from any process of drift through fortuitous differential reproduction in the accumulation of random or indiscriminate errors of sampling. To confirm the theory of natural selection empirically is to confirm that this probabilistic causal process exists, is competent, and has been responsible for evolution. Such hypotheses are both falsifiable and verifiable, in principle, if not in practice.

Natural selection has been accepted and developed by biologists with very diverse attitudes toward chance and chances. But the theory and its acceptance have always involved probabilistic causal judgments that cannot be reduced to correlational ones. So, the theory has contributed to a probabilistic shift within the development of causal science, not to any probabilistic rebellion in favor of science without causes.

– M.J.S. Hodge
– Natural Selection as a Causal, Empirical, and Probabilistic Theory
– The Probablistic Revolution: Volume 2
– p. 233

Thoughts?

130 thoughts on “Chance and Selection

  1. Joe Felsenstein: I think that this phenomenon involves the interaction of natural selection with genetic drift, and that it does not involve a change in the natural selection, which remains constant even as genetic drift overwhelms it.

    If it does remain constant (close, sometimes, I suppose), doesn’t it remain a constant probability?

    It’s really mostly semantics, of course, because the point is not whether or not it’s “probabilistic” or not, just whether or not it is “just luck.” Clearly not the latter, and I think one could consider NS a non-random bias (even if selective pressures are a bit of potluck itself), but it seems that one can’t deny that NS is a probabilistic process as another legitimate interpretation.

    Glen Davidson

  2. Rumraket: Looks like something’s got phoodoo all riled up today. Oh well…

    Over-sensitivity to stupidity. He would apply for disability but that would requiring having contact with the government which would only exacerbate his condition.

  3. Joe Felsenstein: It’s extremely clear why creationists and design advocates are so insistent on labeling natural selection as “random” or “chance” or “lucky accidents”.

    Flog that straw man Joe. Flog it good. The question isn’t whether natural selection is random, it’s whether it is probabilistic and whether it is stochastic.

    I have made this clear several times. Please try to do better.

  4. GlenDavidson: Clearly not the latter, and I think one could consider NS a non-random bias (even if selective pressures are a bit of potluck itself)…

    Yes, I have plainly told Joe that I agree that natural selection introduces a bias, that some outcomes become more likely than others.

    Will I roll a seven every time I roll the dice? No, but rolling a seven is more likely than rolling any other number.

  5. Rumraket: As with all quotes you bring, you do it because you want to present some sort of gotcha with statements you can interpret to contradict each other.

    So where have a claimed that Hodge has contradicted himself? Or are you reading chicken bones again.

  6. Allan Miller: So, that is the basic process, subject to a probability distribution. Selection is a bias in the distribution, not something separate from it.

    Nice post.

  7. J-Mac: So, if out of the 2 individuals the one with 10% greater viability doesn’t survive and the “inferior” one does, what are we to conclude?

    That shit has some probability to happen.

    J-Mac: Did natural selection eliminate the more fit or viable individual?

    Nope.

    J-Mac: Or, was it just pure random process?

    Nope. If it was “purely random,” then the one with 10% higher probability of surviving, would not have had that higher probability of surviving. It would have the very same probability of surviving as the other individual.

    Think of a somewhat loaded dice. It has a higher probability of landing a 6, but it won’t always land a 6, sometimes the imbalanced weight will not be enough to guarantee the 6, but the 6 will come more often than expected by “pure randomness.”

  8. Mung,

    Nice post.

    Cheers. Tell your friends!

    I came to my viewpoint through taking what I now think was the wrong side in a lengthy series of threads on the old Richard Dawkins forums. Many were critical of their host’s insistence that NS is ‘non-random’. IIRC one such thread was entitled “Is there a definition of ‘random’ such that Drift is but NS is not?”. I spent a long time arguing that there was, but ended up conceding defeat. The ‘population resampling’ thing, with selection as a bias in the distribution; a shift of the mean, is due to Simon Gunkel, posting as ‘susu.exp’.

  9. Joe Felsenstein:
    It’s extremely clear why creationists and design advocates are so insistent on labeling natural selection as “random” or “chance” or “lucky accidents”.

    J-mac: I’m neither a creationist nor a design advocate…

    Joe Felsenstein:
    So that they can then argue that it is just random wandering, unrelated to the direction of changes that would improve adaptation.

    J-mac: If that is true, as it appears to be, that there is a strong, inescapable element of randomness in natural selection, they have every right to argue it…Wouldn’t you do the same if you had exposed a strong, inescapable element of randomness in a process of design of life?

    Joe Felsenstein:

    These folks are terribly insistent on Clarity, but when the creationist debaters get up and start arguing that evolutionary biology has a theory that explains remarkable adaptations as purely accidental, as occurring by pure random wandering, somehow you won’t hear any of these advocates of Clarity standing up and correcting the creationist debaters.

    J-mac: It’s all about getting to the truth Joe, don’t you think? It’s not about who wins the argument…At least that’s the way it should be… Isn’t it? But that’s a theme for another OP…

  10. What does it mean to say that selection is probabilistic? Seems to me that is like equating selection to the whole process involving random variation, drift and also selection, in the sense Joe has been using it. Perhaps we’re just talking past each other here (while Mung, as always, tries hard not to understand)

  11. dazz: Perhaps we’re just talking past each other here (while Mung, as always, tries hard not to understand)

    I’m trying hard not to understand people talking past each other?

  12. Entropy: If it was “purely random” …

    … the 6 will come more often than expected by “pure randomness.”

    So what is impure randomness? Are you ok then with saying that natural selection is impurely random?

    Natural selection is a random process in which not all outcomes are equiprobable. We call this kind of random process an impurely random process.

    We good now?

  13. J-Mac,

    These folks are terribly insistent on Clarity, but when the creationist debaters get up and start arguing that evolutionary biology has a theory that explains remarkable adaptations as purely accidental, as occurring by pure random wandering, somehow you won’t hear any of these advocates of Clarity standing up and correcting the creationist debaters.

    Someone needs to nail down what “remarkable adaption” means.

  14. phoodoo: It means lucky accidents! Enough already!

    As one pair of authors put it, “by a stroke of good luck anthropoids were able to ‘win’ the sweepstakes.”

    – Fleagle and Gilbert

  15. dazz,

    What does it mean to say that selection is probabilistic? Seems to me that is like equating selection to the whole process involving random variation, drift and also selection, in the sense Joe has been using it.

    I think its more that drift and selection have a combined effect that renders the two effectively inseparable – we’re not including variation generation as well.

  16. Mung: So where have a claimed that Hodge has contradicted himself? Or are you reading chicken bones again.

    So where have I claimed that you have claimed, that Hodge contradicts himself? Or are you reading chicken bones again.

  17. phoodoo: I don’t think Mung ever labeled selection as lucky accidents.

    He says they’re random, and you think ‘random’ means “lucky accidents”. Since Mung has not given any indication that he disagrees with your definition of random, nonsensical as it is, one is entirely justified in inferring that Mung thinks both mutations and selection are “lucky accidents”.

    Now if I’m wrong here then perhaps you and Mung could spend some time really fleshing out exactly what the words ‘random’, ‘stochastic’, ‘lucky’, and ‘accident’ means. It’s not at all clear that you understand those words to mean the same thing.

  18. Mung: Quote of the week. We couldn’t agree more.

    Good, I’m glad to see you agree with me that when phoodoo defines ‘random’ as “lucky accidents” that this doesn’t explain anything.

    I will leave you two to duke this one out then.

    So why is it the basis of your theory.?

    Yeah phoodoo, why is the basis of the halluscination of evolution you think we are positing, based on a nonsensical term? Neither me nor Mung can make sense of it.

  19. Allan Miller:
    dazz,

    I think its more that drift and selection have a combined effect that renders the two effectively inseparable – we’re not including variation generation as well.

    OK, but the point is that it’s the whole process (of allele fixation), involving selection and drift, that is stochastic. I don’t think saying that fitness is relative is the same thing as saying that selection is stochastic.

  20. dazz: OK, but the point is that it’s the whole process, involving selection and drift, that is stochastic. I don’t think saying that fitness is relative is the same thing as saying that selection is stochastic.

    Not sure what the “relative” thing means.

    I’ve been over this with Mung repeatedly, to no avail, so I will repeat myself. The whole process of selection+drift may be modeled as a stochastic process. That does not mean that the selection part should be regarded as “random” or “chance” or “probabilistic”.

    Individual deaths may be modeled as random, and similarly individual redproduction. But the bias that favors one genotype over another is, in straightforward models, consistent. It is just like gambling in a casino where the odds are against you. You will lose your shirt more certainly the more times you gamble. If the casino owners find that the odds are actually “random” or “stochastic” they will be very upset and will close down the game. They depend on the odds being always biased in their favor. Straightforward models of natural selection and drift have a bias in favor of some genotypes. Mung has admitted that there is such a bias. Why Mung then insists on describing this bias as “random”, “stochastic” or “probabilistic” is a mystery.

  21. Joe Felsenstein: Straightforward models of natural selection and drift have a bias in favor of some genotypes.

    As we are talking of casinoes, does the analogy of a roulette wheel help? With equal slots acting as neutral selection and drift (all outcomes equiprobable), you can then widen some slots (representing beneficial alleles) and narrow others (deleterious) so outcomes are biased towards the wider slots but there is still an element of chance.

  22. Alan Fox: As we are talking of casinoes, does the analogy of a roulette wheel help? With equal slots acting as neutral selection and drift (all outcomes equiprobable), you can then widen some slots (representing beneficial alleles) and narrow others (deleterious) so outcomes are biased towards the wider slots but there is still an element of chance.

    OK for one spin per individual, and modeling a whole generation by multiple spins. But where the analogy breaks down is that the mix of genotypes found does not then undergo random mating and reproduction, and does not become the starting point for the next generation, in the roulette analogy.

  23. Joe Felsenstein: Not sure what the “relative” thing means.

    It’s a reference to this

    Allan Miller:
    Joe Felsenstein,

    I’m still not sure how those processes are really separable. If two competing alleles have equal selection coefficients, we have drift alone. Then if we add a tiny increment to the one, such that the other goes down relative to it, I don’t see that we have materially changed what’s going on, other than that we have introduced a bias, a directional arrow, to the neutral fixation probability and rate (I realise that’s a slight variation on my original ‘near-neutral’ case). We’ve shifted the mean of the distribution.

    I know that there is some reluctance to concede randomness because of the hay Creationists inevitably make of this. But they make hay of anything they can lay their hands on, so let ’em.

    If we see the process as one of generational resampling, a probabilistic process with varying degrees of bias, I don’t see that this is a fundamentally wrong way of looking at it.

    I think I understand what you mean, and what Allan means, and I don’t see a problem between both positions to be honest.

  24. dazz,

    I wasn’t saying that “fitness is relative is the same thing as saying that selection is stochastic”, though!

    When two alleles are segregating in a population at the same locus, relative fitness is the key. If their absolute fitnesses are equal, all is drift – it’s the neutral case. It’s just inevitable that turning one coefficient up will turn the other down, relatively speaking. But if one imagines doing this by the tiniest increment, one has hardly moved at all from the position where all is neutral. One hasn’t cranked a knob all the way to ‘selection’; there’s still loads of drift. One can imagine turning the knob by further such infinitesimal increments; drift incrementally goes down while selection goes up; there is nonetheless no point at which the randomness stops.

    Of course we can separate them conceptually – if we couldn’t, I couldn’t make statements about the one or the other. But if, say, there are 1000 births to 999 in favour of one allele over the other, you can’t say for example that one of those births was due to selection, the rest to drift.

    It’s like a biased dice. On a given roll, was it the bias that made it come up? Or is it not just probabilistic, with bias?

  25. J-Mac:

    Joe Felsenstein:
    So that they can then argue that it is just random wandering, unrelated to the direction of changes that would improve adaptation.

    J-mac: If that is true, as it appears to be, that there is a strong, inescapable element of randomness in natural selection, they have every right to argue it…Wouldn’t you do the same if you had exposed a strong, inescapable element of randomness in a process of design of life?

    This confirms what I suspected about how the label “randomness” would be used.

  26. Alan Fox: As we are talking of casinoes, does the analogy of a roulette wheel help?

    Not really. 🙂

    Maybe think of it in terms of colors rather than individual numbers At least in that scenario there is an unequal distribution and a clear bias. Red and black may have an equal chance of increasing or decreasing in the next generation but their chance is better than that of green. So it’s a clear bias against green, but it’s only a probability, not a certainty, that green will not increase.

  27. Allan Miller: I wasn’t saying that “fitness is relative is the same thing as saying that selection is stochastic”, though!

    I didn’t mean to misrepresent what you said, sorry about that

    Allan Miller: Of course we can separate them conceptually

    I think that’s the crux of the matter. Selection and drift are both intertwined in practice, but Joe is referring to selection conceptually, as a “parameter” in evolutionary models. That’s why I see no fundamental disagreement here

  28. Joe Felsenstein: The whole process of selection+drift may be modeled as a stochastic process. That does not mean that the selection part should be regarded as “random” or “chance” or “probabilistic”.

    …the decomposition of an evolutionary process into its deterministic and stochastic components may seem rather contrived.

    – Elliott Sober. The Nature of Selection. p. 115.

    🙂

  29. dazz,

    It’s true I think we are largely talking semantically. There is but one parameter though: selection coefficient. It can essentially take the value 0 (neutral; drift) or non-zero (beneficial/detrimental; selection+drift).

  30. Mung: So it’s a clear bias against green, but it’s only a probability, not a certainty, that green will not increase.

    Do you think that, in population genetics, an allele that has lower fitness is certain not to increase?

  31. Joe Felsenstein: Do you think that, in population genetics, an allele that has lower fitness is certain not to increase?

    No. It’s fate is probabilistic, not fatalistic. 🙂

  32. Joe Felsenstein: J-mac: If that is true, as it appears to be, that there is a strong, inescapable element of randomness in natural selection, they have every right to argue it…Wouldn’t you do the same if you had exposed a strong, inescapable element of randomness in a process of design of life?

    Joe FelsensteinThis confirms what I suspected about how the label “randomness” would be used.

    So, you agree that there is an inescapable element of randomness in natural selection, you just don’t like how it could or is being used?

  33. Elsewhere Joe has argued that natural selection is deterministic, and this appears to be the reason that he thinks it cannot be probabilistic. Is there such a conflict though, in that both cannot be true?

    Can a process be both probabilistic and deterministic?

  34. I take back that thing about Mung trying hard to not understand. He doesn’t even need to try

  35. Fisher’s late-life changes for his 1958 Dover edition often confused as much as they clarified …

    – W.D. Hamilton

    meh

  36. I should call the book something like

    THE GENETICAL THEORY OF NATURAL SELECTION

    I cannot easily get the words statistical or mathematical into the title, but genetical is essential. My impudence in treating the subject as a branch of mathematics, I must justify in a preface; very short and historical, was my intention, not dealing there with methods, and only hinting at aims.

    There will be four or five chapters on Man as the subject is generally shirked by geneticists, and I know of no historian who knows what Natural Selection means.

    – R.A. Fisher

    A statistical theory?

  37. Mung: meh

    Hamilton in full:

    This is a book which, as a student, I weighed as of equal importance to the entire test of my undergraduate Cambridge BA course and, through the time I spent on it, I think it notched down my degree. Most chapters took me weeks, some months; even Kafka whom I read at the same time couldn’t depress me like Fisher could on, say, the subject of charity, nor excite me like his theory of civilisation. Terrify was even the word in some topics and it still is, so deep has been the change from all I was thinking before. And little modified even by molecular genetics, Fisher’s logic and ideas still underpin most of the ever broadening paths by which Darwinism continues its invasion of human thought. For a book that I rate only second in importance in evolution theory to Darwin’s “Origin” (this as joined with its supplement, “of Man”), and also rate as undoubtedly one of the greatest books of the present century, the appearance of a variorum version is a major event. Fisher’s late-life changes for his 1958 Dover edition often confused as much as they clarified and this new edition should allow some of the old enigmas to be at last disentangled. Unlike in 1958, natural selection has become part of the syllabus of our intellectual life and the topic is certainly included in every decent course in biology. By the time of my ultimate graduation, will I have understood all that is true in this book and will I get a First? I doubt it. In some ways some of us have overtaken Fisher; in many, however, this brilliant, daring man is still far in front. W. D. Hamilton FRS, Department of Zoology, University of Oxford.

    Here’s a nice take on William Hamilton.

  38. Joe Felsenstein: Do you think that, in population genetics, an allele that has lower fitness is certain not to increase?

    If the definition of lower fitness is alleles that are certain to not increase, than I believe an allele that has a lower fitness is certain to not increase.

    I think what can happen however, is if an allele of lower fitness increases, it gets called an allele of greater fitness. Thus the paradox is solved.

  39. phoodoo: I think what can happen however, is if an allele of lower fitness increases, it gets called an allele of greater fitness.

    I briefly forgot that the process is deterministic, in which case the allele would certainly be lost. Unless it wasn’t.

    But because I don’t want dazz to think that we’re complete IDiots, if an allele does get fixed in spite of it’s lower fitness, we can always blame that on drift and thus save our theory from falsification. How convenient.

  40. dazz: I take back that thing about Mung trying hard to not understand. He doesn’t even need to try

    Don’t read that paper I linked to dazz. It might break something very fragile in your psyche and we would not want that.

  41. Mung: if an allele does get fixed in spite of it’s lower fitness, we can always blame that on drift and thus save our theory from falsification. How convenient.

    Yes, I guess that is the definition of drift.

    All of life’s problems can be solved through random definitions, that then become selected.

  42. R.A. Fisher:

    It will be noticed that the fundamental theorem [of natural selection] proved above bears some remarkable resemblances to the second law of thermodynamics. Both are properties of populations, or aggregates … both are statistical laws …

  43. Mung:
    Elsewhere Joe has argued that natural selection is deterministic, and this appears to be the reason that he thinks it cannot be probabilistic. Is there such a conflict though, in that both cannot be true?

    Can a process be both probabilistic and deterministic?

    I have a distinct impression that Joe has a third option up his sleeve..Since he has recently become somewhat an expert in quantum mechanics, maybe he has cracked the secret code of the enigma of subatomic world?

    What do you think of this?

    https://www.inverse.com/article/12061-quantum-darwinism-is-where-natural-selection-meets-quantum-mechanics

  44. The examples that Joe has provided of the fate of an allele appear to mitigate against his claim that natural selection is deterministic.

  45. Perhaps I need to clarify. Perhaps Joe didn’t really say that natural selection is deterministic. Perhaps what he actually said is that in some models of natural selection, natural election is deterministic. In other models of natural selection, natural selection is not deterministic. So natural selection is deterministic, except when it isn’t.

Leave a Reply