Cornelius Hunter seems very confused.
…This brings us back to the UC Berkeley “Understanding Evolution” website. It abuses science in its utterly unfounded claim that “natural selection can produce amazing adaptations.”
In fact natural selection, even at its best, does not “produce” anything. Natural selection does not and cannot influence the construction of any adaptations, amazing or not. If a mutation occurs which improves differential reproduction, then it propagates into future generations. Natural selection is simply the name given to that process. It selects for survival that which already exists. Natural selection has no role in the mutation event. It does not induce mutations, helpful or otherwise, to occur. According to evolutionary theory every single mutation, leading to every single species, is a random event with respect to need.
He has forgotten what “adaptation” means. Of course he is correct that “Natural selection is simply the name given to [differential reproduction]”. And that (as far as we know), “every single mutation …is a random event with respect to need”.
And “adaptation” is the name we give to variants that are preferentially reproduced. So while he would be correct to say that “natural selection” is NOT the name we give to “mutation” (duh); it IS the name we give to the very process that SELECTS those mutations that promote reproduction. i.e. the process that produces adaptation.
Cornelius should spend more time at the Understanding Evolution website.
ETA: CharlieM points out below that…
When CH says that natural selection does not produce adaptations he is talking about individual organisms. He is discussing mutations in individuals and adaptations in individuals. Natural selection has nothing to do with the first appearance of an adaptation in an individual.
And that of course is the confusion – I hadn’t seen just where Cornelius’s confusion lay. Because, of course, the term “adaptation” is a population-level concept. At the level of the individual, the equivalent would be “advantageous mutation”. And that makes the Understanding Evolution website absolutely correct.
Allan:
Heh. But for any such space, there is a larger but still finite space that includes both types. 🙂
ETA: And there’s a smaller finite space that includes neither, so they’re really both extinct!
Allan,
I was careful to address that when I introduced the thought experiment:
Allan:
I agree. They’d be hard to find, but they’d still be there.
You’re right. He produced a pile of shavings.
That’s right! they are competing to out-compete each other.
Ergo, natural selection. The fittest survive.
Keiths said:
No, I haven’t. I’ve told you straight out that NS has an essential role in the production of adaptations, and I’ve said repeatedly – after you pointed his words out – that CH was wrong to say that NS doesn’t influence the building of adaptations. I’ve said repeatedly that having an essential role doesn’t necessarily mean it actually builds the features.
William,
Mutation and selection together build — or produce — complex adaptations.
I repeat:
You keep saying that selection doesn’t build or produce anything — only mutation does. That depends on a woefully narrow and unreasonable definition of “build” and “produce”.
I’ve offered counterexamples. By your logic, a woodworker turning a decorative table leg on a lathe hasn’t produced anything. A worker carrying bricks isn’t building a house; only the bricklayer is. Ditto for someone cutting boards to fit vs. the person nailing them in place.
What’s ironic is that even mutation doesn’t build adaptations according to your narrow definition. All it does is build sequences. The genes have to be expressed for the adaptation to be built, and mutation doesn’t express genes — it just changes them.
Could you name some of these scientists? I mean ones from the last sixty or seventy years.
I didn’t argue that there was no selection in the case of two flasks; I said that the situations were different. In the single-flask case, the bacteria form a single population because you will necessarily sample both alleles if you make an observation. In the two-flask case, it depends on how you’re defining your population. What conclusions you draw about the operation of natural selection can easily depend on how you draw the boundaries of your population. For example, consider a pair of alleles in P. falciparum. One allele confers a greater ability to invade red blood cells, while the other confers a greater production of gametocytes. If you look at a single infected host, the first allele will be increasing in frequency, since the number of parasites depends on the number of infected erythrocytes. The second allele, however, is more likely to be transmitted, and thus increases in frequency in the village. Which allele is being positively selected?
As far as I can tell, your claim is simply wrong. According to you, any mutation whose sole effect is to increase the absolute fitness of its owner — no effect on competition, no culling of competitors — will increase in frequency, and increase the absolute fitness of the population, but it won’t be because of natural selection. It will be some other, nameless process that also leads to adaptive evolution. No evolutionary biologist would restrict “natural selection” in that way. (That’s particularly true because it’s often very hard to tell exactly how a beneficial mutation is operating. If we know an allele has increased more than expected by drift, we conclude that selection has been operating — not that either selection or some nameless other thing has been operating.)
keiths:
Steve:
There is no two-flask case. We are talking about a single flask for the entire population versus one flask for each organism, as Allan described:
I’ll hold off on responding to the rest of your comment until it’s clear we’re talking about the same thing.
Steve:
keiths:
Steve:
You could look these up as easily as I.
Stephen Jay Gould:
Robert Hazen:
Ernst Mayr:
That should suffice. If you want more, feel free to do the research yourself.
Just wanted to point out that we have
a neuroscientist/cellist,
a biochemist/guitarist/drummer,
a population geneticist/phylogeny inferrer,
a computational biologist/particle physicist, and
an engineer/amateur rocket builder/future truck driver,
all discussing the terminological separability of heritable differential reproduction from natural selection.
I’ll bet that’s never happened before.
I’m trained in architecture too 🙂 Probably explains my predeliction for terms like “build” and “construct”.
except, no, I’ve not agreed that at all.
To be clear: In a world that does not exist, what you say might be true. But that world does not exist, and therefore what you say is not true. Therefore you cannot use it to bolster a case about what “build” means in that real world.
Mung,
Not strictly. If there is no differential in reproductive output, there’s no NS but there is Drift. The lucky also survive, and variation is still eliminated. Drifting alleles are ‘competing’ for these places in a finite population, but with equal strength.
I have seen people argue (with some merit) that the separation of NS and Drift is artificial. There is simply a continuum of selective differentials, which includes zero. What’s really going on is population sampling, with variable degrees of bias.
keiths,
🙂
Careful circumscription of one’s population of interest is of course very important. One has the thorny issues of how ‘the population’ is bounded in complex bacterial ecosystems … and sex. For many, sex-mediated gene flow itself draws the boundaries of the population – the difference between reproductive isolation and intercompatibility. Errors, quite common and fundamental ones, can arise when people regard asexual offshoots of such populations as still part of them in pop-genetic treatments. Hence some of the ‘mystery of sex’, IMO, when it is challenged to provide a selective offset for the assumed twofold cost. Selection inside the sexual boundary is not the same as that across it. There is no selective pressure, in the more conventional sense, for a sexual lineage to become asexual.
That to me is the best way of thinking about it. It deals with the fact that “natural selection” is a stochastic process, just as mutation is. I think the separation of evolution into “random” and “non-random” processes is unhelpful.
Mutation is probably not biased by what will work better than in the current environment than the current average, but is certainly biased in favour of what is similar to the parent genotype.
Drift is biased by what works best in the current environment.
A constant drip-feed of mostly near-neutral variation is all it takes.
Elizabeth,
Now you are calling drift, a process which is biased, but just barely? Nearly neutral, but not quite? Then what is the difference between that and NS?
Sounds like more definition confusion coming.
phoodoo,
No, drift isn’t biased. The difference between drift and NS is bias.
I s’pose. But it is a legitimate position.
Allan Miller,
Then your beef is with Lizzie use of words not mine, because that is what she just wrote.
What is legitimate about it?
There isn’t a difference. It’s another way of thinking about “natural selection”: a bias in the drift in favour of what does better.
Instead of each variant having an equal chance of making it through to the next generation, some have a greater chance than other, because that specific variant confers some advantage in the current environment.
That inequality, or “bias” is what is also called NS.
Take the drunkard’s walk. The drunkard starts at a lamppost and starts staggering along the street. But he has no direction – each step is as likely to be east as west. He might travel a long way, just by chance. Or he might end up back near the lamppost. That’s a genetic variant with unbiased drift.
Now imagine that there is a slight slope to the street. Sure, every step is still random, but because east is downhill, and west is uphill, on each step he’s slightly more likely to take an eastwards step than a westward one.
That’s bias, aka NS.
Now, he is quite a lot more likely to end up nearer the eastern end of the street than the west, and quite a lot more likely not to end up back at the lamppost.
Well, think about it.
Elizabeth,
Oh my God, I think everyone here knows that NS implies an advantage or bias (however barely you want to call it) whilst drift implies no advantage at all, strictly neutral.
You want to obfuscate even this point?
phoodoo,
I have no ‘beef’ with either of you. You don’t seem to have said anything which I can argue, merely raising an eyebrow at Lizzie’s words. But I agree with the approach of defining evolution as a stochastic sampling process, with a biased and an unbiased element.
Imagine you have a vibrating table covered in sand. It’s level. The sand just bobbles about. Now tilt it, just a teensy fraction. The sand tends to move downslope. But it still jiggles about as well. The jiggling, at any angle of tilt, is drift. The tilt is NS.
Although there is one point – level – when NS (bias) is absent, a fractional increase in tilt does not immediately destroy the jiggling pattern. Now, things are moving due to NS and drift. The more or less you tilt, the more one dominates. But it’s really only one process, with varying degrees of bias.
Now this is where you tell me that organisms aren’t jiggling sand.
phoodoo,
That’s wrong. See my response above. I have pursued exactly this point elsewhere, to my embarrassment. The penny finally dropped.
Allan Miller,
I saw your response above. It is absolutely meaningless to the point of calling one mutation favorable, and one mutation neutral. That is the only reason for having the two terms, NS and drift. One means its an advantage and that’s why it spread through the population, the other means it has no advantage, but just so happen to be the mutation that spreads.
Trying to make these terms means something else is just a silly exercise down a rabbit hole that I refuse to follow the two of you into.
And it most certainly does nothing to solve the problem of how all these meaningless copying mistakes just so happen to add up to a sophisticated system of complex intertwined parts which require precision from each working part.
It means a great deal. It means that the neutral mutation has the same chance of propagating as any other, while the favorable one has a greater chance.
Drift normally refers to the underlying probabilities of mutatations changing their prevalence under the assumption that all have an equal chance of propagating.
NS is a bias on that process, i.e. the deigree to which the chances are unequal.
It is a perfectly meaningful distinction.
Or didn’t. If all variants have the same probability of spreading, some will and some won’t. If some variants have a greater probability of spreading, they are more likely to do so. They still may not.
It doesn’t mean “something else”. Nobody has said so.
Yes it does, as AVIDA shows.
In AVIDA the copying “mistakes” i.e. the fact that not all offspring are genetically identical to their parents are random. Nor did anyone “mean” anything by them. They are determined by a random number generator. Yet the result is a complex and coherent sequence of instructions that produce a functional logic gate.
So you are simply wrong.
phoodoo,
If you are talking about two alleles of the same gene, it is impossible to call one favourable and the other neutral. It’s either both neutral or one beneficial/one detrimental.
If you are talking about two different loci, it is meaningless to say one was fixed by drift and one by NS. One can determine the correlation with the rate at which fixation occured. The faster the rate, the more NS is in charge.
But fundamentally, you are definitionally WRONG to restrict drift to the selectively neutral. This is basic evolutionary theory, not a definitional quirk here at TSZ.
You are not obliged to read or post anything on any website, anywhere.
No, of course not. This sub-discussion is about genotypes.
Elizabeth,
I wrote a computer program which told the computer to just make lots of mistakes and keep some. It produced nothing but a blob on the screen, so you are just wrong.
Lizzie I am finding your arguments to not even be sophisticated. You say things like: “Drift is biased by what works best in the current environment.”, and then try to obfuscate by claiming you mean NS.
You basically have two forms of argumentation-obfuscate meanings, then proclaim some sort of victory. Then claim the other person is distorting your pointless argument.
Post the code then.
Much confusion in this thread pretending ‘nature’ is an ‘agent’ like in social sciences and humanities. This is the problematic Malthusianism in Darwin’s evolutionary theory. As Darwin wrote to Lyell (1860): “Talking of ‘Natural Selection,’ if I had to commence de novo, I would have used ‘natural preservation’.” (http://www.darwinproject.ac.uk/entry-2931)
In contemporary parlance, I suspect this means that Darwin didn’t want to ascribe ‘agency’ to ‘nature’ improperly. The English verb ‘to select’ is too close to the verbs ‘to choose’ and ‘to decide’ for biologists to be able to monopolise it (within an evolutionistic ideology), when they are not actually studying ‘persons’ or ‘agents’ except in a very loose way. Biology is not as ‘exact’ as chemistry or physics, after all.
OMagain,
Damn! This post originally said almost exactly that!
Have you ever heard of “one black swan”?
My arguments aren’t sophisticated. The beauty of Darwin’s mechanism is that it’s so simple. Once you see it, you go: d’oh, why didn’t I think of that?
Because that’s what NS is – a bias on drift. It’s not “obfuscation” – it’s what it means.
In this case it looks like simple failure to read properly.
Have you actually read On the Origin of Species by Means of Natural Selection Gregory?
Elizabeth,
Then why the fuck did you call it drift instead of natural selection!
About that code…
I called it bias on drift.
It’s the same thing. Natural selection can be thought of as a bias on drift – on the “random walk” that will happen anyway.
I wouldn’t sweat the code. If phoodoo managed to write an EA that didn’t work then that doesn’t prove that EAs don’t work, it just demonstrates that phoodoo isn’t very good at writing them.
As Robin pointed out to you, William, this is dead wrong. To illustrate this, I ran a toy model in which there is ZERO mutation (RM has stopped), just a varied starting population.
(Pop’n size 300, “brood” size 5, 27^23).
Absent selection, we would expect this population to drift; the probability that our population ever contains our “feature” is of the order of 1 in 10^28. Eventually, drift will eliminate all variation. The probability that the end population contains our desired “feature” is below 1 in 10^32.
However, if we allow selection, then :
Our feature is produced in 17 generations.
William, you have failed to consider the importance of sex.
[inset joke here]
Natural selection is perfectly capable of “building” features on its own (in a sexual population).
Cornelius is confused.
Confession time,
I’ve always been a bit dubious about the idea of drift. Though I found OM’s simulation very impressive.
And this analogy of Allan’s is really helpful (to me, at least).
…I agree with the approach of defining evolution as a stochastic sampling process, with a biased and an unbiased element.
Imagine you have a vibrating table covered in sand. It’s level. The sand just bobbles about. Now tilt it, just a teensy fraction. The sand tends to move downslope. But it still jiggles about as well. The jiggling, at any angle of tilt, is drift. The tilt is NS.
Although there is one point – level – when NS (bias) is absent, a fractional increase in tilt does not immediately destroy the jiggling pattern. Now, things are moving due to NS and drift. The more or less you tilt, the more one dominates. But it’s really only one process, with varying degrees of bias.
So that selection always happens, just that drift occurs when selection is unbiased.
Yet I see Allan says:
Now I’m confused!
Elizabeth,
Oh, so, it’s a reading contest now, Lizzie? Have you ever read Étienne Gilson’s “From Aristotle to Darwin and Back Again”?
I’ve read most of OoS and even ‘taught’ (intensive reading) it – specifically, the “On Natural Selection” part!! – with students, studying its rhetoric (not just its ‘science’). Those who promote ideological scientism (guess: more than 60-70% of TSZers) often grow quite uncomfortable when their rhetoric is analysed and exposed.
What’s your point – that Darwin didn’t actually mean what he wrote to Lyell? (He expressed regret about the phrase ‘natural selection’ elsewhere too.) Are you insisting that ‘nature’ actually IS an ‘agent’ just like in SSH? Here on such questions is where your woolly ‘philosophy’ reveals gaps in your thinking & feeling.
That Dennett-Darwin acid sure must feel stinging to you nowadays, Lizzie. Or have you simply become numb to that pain?
Alan Fox,
No. Maybe it was a bad analogy. Drift always occurs to some degree – you can’t eliminate random factors. Selection is always biased (that’s what it is: bias). Selection varies in degree, all the way from zero upwards. As you turn it up, drift goes down in proportion to the amount.
If two alleles have the same selection coefficient, they can only drift. Change one selection coefficient by a tiny amount (above the threshold of effective neutrality for this population size), and drift does not suddenly stop happening, just because selection has begun to be in effect. Drift is strongest at neutrality.
Allan Miller,
You have one big problem-the only definition for something having a selective advantage is if it drifts through the population. Trying to claim one trait is neutral, while another is advantageous is a pointless distinction.
Still, neither helps explain how you build complexity through uselessness.
Not at all. But if we are going to critique Darwin’s metaphor, it would be useful to know that you’d actually read his book.
Perhaps it’s best to rebuild concepts from the ground up. That was the intention with the “M&Ms “. This was a simplified version of an unbiased sampling process – the neutral case. The only thing at work there is drift. All alleles have a chance of being fixed which is directly in proportion to their starting frequency.
Now change it so that, every 100th go, the operator peeks at the colour. If it’s red, they put it back and pick again (or throw it away and pick another). Any other colour, they do the ‘blind’ operation. That’s selection.
Drift is still happening – on 99 turns, the colour has no effect on its own preservation, but every 100th, it does. This makes red more likely to be the final colour, but not guaranteed.
But what is important to note is the fact that the sampling process itself is the thing that is causing elimination of variation, irrespective of selection. It happens faster or slower depending on the degree and direction of bias (NS). But it always happens.
Bear with me while I ask stupid questions. (I think your observation that 90% of arguments are over semantics is pessimistic. 90% is far too low an estimate 🙂 )
Can unbiased selection not happen? I thought you were suggesting a continuum of bias from zero (drift) to what, strong bias, max bias? And even if there is no selective bias, cannot alleles either fix or disappear as in OM’s simulation?
phoodoo,
No. We’ve been through this, and I hold out no hope you’ll get it this time. If it has a selective advantage, the mean fitness of carriers has a higher value than the mean fitness of non-carriers. But on any given run of 2 such alleles through a population, either may be lost. You can’t determine fitness by ultimate success in one run. A biased roulette wheel can throw up any number on any given spin.
Still, I didn’t say it did. How many times are you planning to repeat this mantra? You want the “AVIDA” thread, third door on your left.
Elizabeth,
Well, so now you know something you didn’t before. Shall we actually critique Darwin’s metaphor then or just on the surface play a game where in your interpretation atheism wins?
Alan Fox,
I suppose you could get a net lack of bias from two opposing selectors!
The advantage of an allele relates to its net offspring output (fitness). If you put two alleles together in a population, their relative rates of increase will act as exponents, compounding selective gains and changing frequencies in finite populations. The continuum of bias relates to the continuum of possible differentials that any 2 alleles can experience. But because the process is stochastic, fitness is an expected value – a probability distribution – not a ‘fixed interest investment’.
Not just ‘they can’, but they will! The uncertainty as to which allele will be fixed is replaced by a higher-level certainty that one of them will. Variation is (almost) always eliminated, regardless of selective differentials, and absent the input of novelty. That’s why drift is important.
Sorry for the delay in answering keiths — I have a lot of other things going on.
I am glad to hear that we all knew all along that a population growing without limit could see one allele taking over because of that-bias-that-isn’t-drift-or-mutation.
Mutational biases can change gene frequencies systematically. So can migration. I use “natural selection” as designating a situation where different genotypes, or different phenotypes, have different fitnesses (at least some of them).
My point is that you can define “natural selection” as you wish, but that is not binding on evolutionary theorists, who use it my way.
The usage of evolutionary biology agrees with that.
Selectionist theory is next to worthless in study of long term evolution. What works well for describing truncation selection in anti-biotic and pesticide resistance isn’t applicable to changes over deep time primarily because what defines “fit” is equivocated, redefined, and auto-renormalized over many generations.
In species that have insulin receptors and use insulin to regulate their metabolism individuals unable to synthesize insulin will be selected against (since it’s pretty much lethal) and hence we say insulin is selectively favored.
Evolutionists falsely presume this is evidence insulin came about through natural selection since it is life critical. The problem however is we have no data on the selective pressure on insulin when the machinery for an insulin metabolism isn’t present. As St. Mivart pointed out, selection can’t select for non-existent traits! We can’t use the fact a trait is selectively favored today as evidence it was selectively favored in the deep past.
Lewontin points out fitness is an ill-defined concept in his Santa Fe 2003 papers, and that paper pretty much unDarwinized Stanley Salthe.
Suppose random drift causes a population of cave fish to be all blind or any given population for that matter to lose some functional trait. At that point, what defines a eugenically ideal phenotype is a functionally defective creature relative to its ancestors. An entire population can continue to be reduced functionally by reductive evolution while all the while getting more eugenically “fit” in the sense of population genetics. Absurdity.