More on the "costs" of evolution

Walter ReMine (an anti-evolutionist who ardently believes that "Haldane's Dilemma" is a real problem for evolution) recently updated the entry for "Haldane's Dilemma" at the CreationWiki. The update does not directly refer to my recent posts on the topic, but does address the points that I made. Actually, "address" is probably the wrong word - he provides a hand-waving dismissal without actually responding to any of the specific points I raised. Ordinarily, a hand-waving response isn't worth the effort needed to write a reply, but in this case the errors that ReMine makes are worth discussing simply because they provide a convenient jumping-off point for a discussion of the way evolution actually works.

Remine's latest update to the Haldane article reads:

Recently, evolutionists are promoting another version of soft selection. In reality, it is not a type of selection, but an illegitimate accounting trick, similar to the double-booking used in financial scams. It operates like this. In effect, the various other costs of evolution (such as the cost of continuity, the cost of mutation, the cost of segregation, the cost of random loss, and a several other costs) are collectively -- and misleadingly -- called "background mortality". Then this is claimed as an additional source to pay the cost of substitution. That is, the accounting that should go toward paying the "background mortality" is diverted to paying the cost of substitution. That leaves the other costs of evolution unpaid, which is not a plausible scenario. In effect, evolutionists are robbing Peter to pay Paul. This approach does not reduce the cost of substitution one iota, instead it attempts to increase the payment by means of misplaced accounting practices. [Note: This version of soft selection is promoted only when evolutionists are trying to solve cost-related problems, otherwise it plays no role in their evolutionary presentations.]

Apparently, ReMine thinks I'm playing shell games, and that I've selected the examples I have to try and "solve cost-related problems." Actually, neither could be further from the truth. In reality, what I did was to demonstrate exactly why cost isn't a problem - at least it's not the way that ReMine defines it.

Here's what ReMine has to say about evolutionary costs in the CreationWiki article:

A cost argument operates as follows. Every evolutionary scenario requires a certain level of reproduction rate - called a "cost." A cost is a required reproduction rate for a given evolutionary scenario. If the species cannot actually produce the required reproduction rate, then the scenario is not plausible. In other words, each scenario has a requirement, and if the requirement is not met, then the scenario is not plausible. That requirement is reproduction rate.

There are various types of cost, each for a specific purpose. Such as:

1. The cost of continuity - a reproduction rate of 1, which is required merely for continuing into the next generation.

2. The cost of mutation - the extra reproduction rate required for preventing genetic deterioration due to harmful mutations.

3. The cost of segregation - the extra reproduction rate required for maintaining polymorphisms in a population.

4. The cost of random loss - the extra reproduction rate required for coping with random losses to the population, (such as fire, flood, famine, selection for non-heritable traits, and many other factors).

5. The cost of substitution - the extra reproduction rate required specifically for making substitutions - that is, required for increasing the number of copies of the substituting mutations.

[Haldane's version of the cost concept was far more indirect and confusing, but (when correctly reinterpreted) it was effectively identical in how it operated, and in the results it produced.]

The sum of all those costs (plus a few other minor costs) is the cost of evolution: the total reproduction rate required by the scenario. Thus:

Cost of evolution = Cost of continuity + Cost of mutation + Cost of segregation + Cost of random loss + Cost of substitution + ...

If the species cannot produce that required reproduction rate - in other words, if the species cannot "pay the cost" - then the scenario is not plausible. Haldane's Dilemma merely applies that logic each generation.

Strangely enough, this is a case where ReMine has the basics more or less right - he just managed to forget a really, really important cost - the cost of not-entirely-random losses.

The thing about life is that not all death is entirely random. In fact, quite a bit of death isn't all that random. The organisms that are lost to predation are (usually) those that are easiest for the predators to catch. The males that are the worst match to the female ideal for that species are the ones that are least likely to find a mate. The bird with the strongest beak is the most likely to crack the harder shells, and therefore less likely to die from starvation.

That's not to say that everything is non-random, of course. Random chance can always come in and screw things up. The better-swimming deer might get hit by a floating log in the flood and die anyway. But the deer that can't swim at all will die if it gets stuck in the flood. When you get right down to it, it's a lot like that old joke about the two guys getting chased by the bear. One stops to tie his shoe, and the other says to him, "Are you nuts? That's not going to help you outrun the bear!" The first guy says, "That's OK. I don't need to outrun the bear. I just need to outrun you."

The point here is that survival doesn't need to be totally non-random. As long as one variety of the species is less likely to suffer from a particular type of death, there is something there for natural selection to work on. Non-random survival is what makes natural selection possible.

Non-random survival, strangely enough, is missing from ReMine's list of (unless it's one of his "other minor costs"). Natural selection can (and has, and does) act to reduce the rate of non-random death. Reductions in the death rate for a particular genotype can (and do) result in that genotype spreading through the population. There's absolutely no "misplaced accounting" involved here. If a mutation results in fewer individuals dying from a particular cause before reproducing, then more individuals with that mutation will reproduce.

All of the talk about costs and accounting misses one simple, basic point. If a mutation is spreading through the population, whatever "costs" might exist are being paid. Evolution doesn't get credit - any and all costs must be paid immediately and without fail. If the costs are not being paid, the trait doesn't spread. In fact, viewing things in term of "costs" and "payments" can obscure an important point: the reduction in mortality from another source does not "pay" the "cost" of spreading the new trait through the population. If a new, heritable trait is the cause of a decrease in mortality, the decrease in mortality is the cause of the trait's spread.

This is not a novel concept, of course, and it hasn't been a novel concept for almost 150 years. This is just natural selection. Not only is it possible for a decrease in mortailty to result in the spread of a trait through the population, it is almost inevitable. The only way this wouldn't work is if there is no such thing as non-random mortality, or if it is not possible for any mutation to make an organism less likely to die before reproducing.

More like this

Intentionally ignoring Natural Selection in a model can be valid, when honest. Such is the case for Hardy-Weinberg equilibrium. I agree with you completely, Mike Dunford, that Walter ReMine is making a dishonest argument that, de factor, ignores Natural Selection. Whether he himself is honest, or dishonest, I do not know. Your analysis is clear, and your examples, in "man in the street" nontechnical terms, should be compelling to anyone who thinks things through. I've emailed you, offline, a length draft paper that (in two small parts) makes a careful attempt to explain Hardy-Weinberg equilibrium, and explain that "noise" in the model is precisely those things that are neither natural selection, mutation, but otherwise excluded from the assumptions of Hardy-Weinberg equilibrium.

Couching an argument in terms of "costs" is, I suspect, a red herring associated with Actuarial math, as the basis for the Insurance industry.

"Apparently, ReMine thinks I'm playing shell games, and that I've selected the examples I have to try and 'solve cost-related problems.'"

It may be that ReMine is reacting to my rants, because I have been talking about "background mortality" lately both at PT and UD.

ReMine is correct that the concept "background mortality" does not ease the cost of selection as such. It is merely an additional cost, and it doesn't help to argue that selection can grow "for free" by reducing a cost that was probably not even included in ReMine's scenarios. However, ReMine believes that Haldane's limit 0.1 for intesity of selection applies universally and unavoidably. It is this belief that the concepts "background mortality" and "soft selection" address. Although Haldane(1957) is unclear about why Haldane's estimate for average intensity of selection is so low, many believe that it is (implicitly) based on the existence of "background mortality" and the belief that selection can only act on what is left by large amounts of randomly applied mortality (as suggested by Haldane's treatment of moths and their larvae). Thus, soft selection does relax Haldane's limit and allows for a higher rate of substitution. If ReMine insists that "background mortality" is an unnecessary confusion factor, I wonder how he is going to defend the limit 0.1 for all lineages and time intervals. I think he should be insisting that "background mortality" is a remarkable factor and that it can't be evolved against nor reduced by non-random mortality.

I have already suggested other possible reasons for why Haldane's limit is so low. One suggestion is that Haldane's limit is based on estimates for the average rate of evolution, over a large amount of lineages. This estimate would incorporate periods of stasis and rapid evolution. Note that ReMine suggests at CreationWiki that Haldane did not take into account factors such as 1. potential stasis and 2. potential evolutionary dead ends. I wonder how ReMine can be so sure. It is quite possible that Haldane's limit is so low because he did account for the relative slowness of evolution when averaged over many lineages and long periods of time. (Besides, I don't think "dead ends" and "stasis" are separate factors; rather, stasis is likely the result of a local optimum, i.e. at least temporary dead ends in the evolution of many traits.) But most importantly, ReMine is not satisfied if a large collection of data shows significant amount of stasis. ReMine additionally wants to choose 1. where stasis occurs and 2. where rapid evolution occurs (nowhere, of course!), and to do so independently of data. That's exactly what ReMine is doing if he insists that the average rate of evolution has to apply to each lineage and each time interval, such as the evolution of Homo sapiens during the past few million years. (And he seems to insist this when erroneously suggesting that his "magic number" should be reduced to 16 if we were to believe in punctualism.)

Half-seriously, I've also suggested that Haldane favored the limit 0.1 because of its mathematical convenience. Most importantly, it seems to remove the complications of concurrency. (I'd like to receive criticism for this paragraph if someone disagrees with me!) Note that five beneficial alleles with coefficient s_i=0.02 (i=1..5) result in about the same cost as a single coefficient s=0.1. With multiplicative fitness, the mean fitness is reduced to (1.0 - 0.02)^5 ~ 0.90 = 1 - s, when beneficial alleles still have tiny frequencies. I.e. additive fitness and multiplicative fitness are approximately the same. However, what if there were twenty beneficial alleles s_i=0.02, whould they behave the same as one allele with s=0.4? (1.0 - 0.02)^20 ~ 0.67 > 1 - s. No, you'd need about 25 alleles here. You'd need 34 alleles to recude the mean fitness as much as one allele with s=0.5, as opposed to the 25 alleles suggested by additive fitness. In other words, it seems to me that if selection is more intense than 0.1, concurrency does speed up the rate of substitution. ReMine seems to partially admit this at CreationWiki, when claiming that Haldane assumed thousands of concurrent loci and that he exploited the benefits of concurrency. Well, I don't think Haldane did anything of the sort. Did Haldane say exactly how many coefficients d_i he assumes, and did he really use the benefits of multiplicative fitness -- or did he instead just approximate multiplicative fitness with additive fitness? I invite comments on this.

Despite his partial admission to the contrary in CreationWiki, I think ReMine also believes that Haldane somehow showed that two alleles s1=s2=0.1 fix roughly as slowly as one s=0.1 allele. But this is not what Haldane showed; Haldane merely showed that two concurrent alleles s1=s2=0.05 fix about as slowly as one s=0.1 allele, and that two concurrent alleles s1=s2=0.1 are in his eyes unlikely in practice due to the intensity of selection. ReMine reacted very negatively at ARN when I told him about my simulations on "pure intraspecific competition", demonstrating remarkable rates of substitution. He relied on rhetorics like the fact that I had not read much of anything on the subject before writing my simulation. While his criticism was largely justified -- I made many basic errors with e.g. terminology -- I think he would have spent his time better by simply pointing out what is wrong in my simulation. It honestly seems to me that ReMine was simply unable to do the latter. Soon after our public exchange, and during our private exchange, Walter apparently became increasingly interested in population genetics simulations:

http://groups.google.com/group/sci.bio.evolution/browse_frm/thread/f44f…

I wonder if Walter has been able to justify Haldane's limit 0.1 in his simulations, and whether he has simulated "pure intraspecific competution", i.e the renormalization of fitnesses each generation in such a way that the mean fitness remains constant.

In other words, it seems to me that if selection is more intense than 0.1, concurrency does speed up the rate of substitution.

On one of these threads someone who had read Haldane's original paper with this model suggested (from memory IIRC) that Haldane tried to explain the lowest rates of evolution that was observed, with the mechanism known then. Given what caligula says that could explain his choice of parameters.

It would be interesting to see the paper.

By Torbjörn Larsson (not verified) on 31 Jan 2007 #permalink

"with the mechanism known then" - with the mechanisms known then

By Torbjörn Larsson (not verified) on 31 Jan 2007 #permalink

> caligula :
> I wonder if Walter has been able to justify Haldane's limit 0.1 in his simulations, and whether he has simulated "pure intraspecific competution", i.e the renormalization of fitnesses each generation in such a way that the mean fitness remains constant.

Speaking of which, I have been looking into Haldane's Dilemma recently. I just wrote-up an article to explain the problems I saw with Haldane's Dilemma. I also found that intraspecific competition produced extremely fast fixation rates. Anyway, you can read the whole thing at the link below. I thought I'd toss it over to TalkOrigins eventually (i.e. once I get some constructive criticism to polish it up a bit).
http://www.empiresofsteel.com/haldane/

BC:
There are a couple of things I'd like to point out about your article.

First, I think it is wrong to say that Haldane applied his model, or his "magic number", to human evolution, let alone that Haldane estimated whether the number of possible adaptations given by his scenario is sufficient to account for the human evolution during the past few million years. The claims related to human evolution are made by Walter ReMine, not Haldane. Also, I would make it very clear that it is Walter ReMine who thinks that 1,667 for the number of adaptations would be "far too low to explain humanity's capabilities". And Walter doesn't come up with any better reason than, "Who are they kidding?"

http://www1.minn.net/~science/Haldane.htm#a_problem

Other popular reasons for requiring a much higher number of adaptations are, "You just don't turn an ape into an opera-admirer", "Apes generally don't design space shuttles", etc. I.e. attributing the complex products of our culture evolution directly to our biological evolution, which clearly is illogical. Natives of the jungle don't design space shuttles, either. "Wolf-children" aren't exactly opera admirers -- in fact they aren't even self-conscious, demonstrating that the symbol "I" is largely the product of language rather than a property of the soul or of a genetically pre-programmed brain. Yet another, and much better, approach is to list traits where humans are apparently unique compared to other apes. But the common denominator for all of these arguments is that they aren't able to provide an explicit number for the required adaptations. As long as argumentation relies on mere rhetorics and intuition, ReMine's thesis is quite shaky. ReMine's number already allows for more than a hundred different traits more than a dozen subsequent adaptations each, on average. It is well worth bypassing the "Who are you kidding?" and ask: do you happen to have some concrete figure?

That said, it is quite possible that 1,000-2,000 adaptations turn out to be too few. (But it is not decided by mere intuition.) It is even more possible that there are instances of rapid evolution in some places of the history of life where Haldane's model fails. Thus, it is important to show that evolution can be more rapid if we break many of Haldane's oversimplifications (e.g. uniform environment, no internal competition, lots of randomly applied mortality(?)).

Second, in response to your arcticle, Walter would list his various cost classes and ask: where is e.g. the cost of mutation accounted for? I think you should predict this tactic and present a rigorous cost scenario for a diploid population, accounting for all relevant cost classes.

I think your examples on selection are quite clear and good, however. (Do you somwehere clarify that the first examples concern a haploid population BTW?) It is also good that you demonstrate how concurrent substitution in sexual species can in principle produce a rate of substitution higher than one per generation. This is a good contrast for Walter's example scenario on extreme serial substitution, producing 500,000 adaptations in 500,000 generations. (Walter himself is probably somewhat confused here, as he probably hasn't paid enough attention to concurrency. He certainly doesn't encourage his readers to "think concurrently".) An example like this also sort of clarifies how purifying selection can prevent harmful mutations from becoming fixed, even when each individual is born with a newly generated mutation, on average.

Someone certainly has already remarked that in a stable population of 10,000 people, 5000 couples, 2.5 child per couple on average, implies 12,500 born, so 2500 die from whatever causes. These deaths can absorb much selection. I do not understand the idea of an innate genotypically determined fitness value independent of ecological circumstances.

Thanks for the feedback, caligula.

First, I think it is wrong to say that Haldane applied his model, or his "magic number", to human evolution, let alone that Haldane estimated whether the number of possible adaptations given by his scenario is sufficient to account for the human evolution during the past few million years. The claims related to human evolution are made by Walter ReMine, not Haldane. Also, I would make it very clear that it is Walter ReMine who thinks that 1,667 for the number of adaptations would be "far too low to explain humanity's capabilities".

Oh, ok. I was under the impression that Haldane thought this as well. The wikipedia article (which could very well be wrong) quotes Haldane (citing page number 521) as saying, "'Good' species, even when closely related, may differ at several thousand loci, even if the differences at most of them are very slight. But it takes as many deaths, or their equivalents, to replace a gene by one producing a barely distinguishable phenotype as by one producing a very different one. If two species differ at 1000 loci, and the mean rate of gene substitution, as has been suggested, is one per 300 generations, it will take 300,000 generations to generate an interspecific difference. It may take a good deal more, for if an allele a1 is replaced by a10, the population may pass through stages where the commonest genotype is a1a1, a2a2, a3a3, and so on, successively, the various alleles in turn giving maximal fitness in the existing environment and the residual environment."

He never mentions humans specifically in this passage, but it seems like the passage would apply to the human lineage.

That said, it is quite possible that 1,000-2,000 adaptations turn out to be too few. (But it is not decided by mere intuition.)

I agree with you on that point, although even I have an intuitive feeling that 1000-2000 changes is too few. For a creationist who is hoping to find reasons to dismiss evolution, I'm sure this reasoning sounds throughly convincing. It even has the air of "truthiness" and mathematical rigor. I thought it would be better to concentrate on why this number is not some sort of upper limit on the number of beneficial alleles that could appear in the human lineage.

Do you somwehere clarify that the first examples concern a haploid population BTW?

I probably need to go back and make sure I'm using diploids uniformly. I originally wrote the software to assume haploid organisms, then I went back and changed it to diploids. I believe all my examples involve diploid organisms (with the possible exception of the first sexual-selection calculation, the one that involves one beneficial allele). I wanted to switch to diploids to shut-down any creationist claim that "diploids fix alleles slower than haploids, therefore, I'm fudging numbers by using haploids". (In reality, there isn't a big difference between the two, but I thought it would be good to shut-down that avenue of argument.)

This is a good contrast for Walter's example scenario on extreme serial substitution, producing 500,000 adaptations in 500,000 generations.

I should probably make ReMine's claim explicit at the end of the article. I was thinking of exactly that claim when I wrote it.

Heleen:

These deaths can absorb much selection.

Yes, they can. However, that only gives you 2,500 deaths, and Haldane's model would require something like 95,000 deaths for the fixation of an allele.

I do not understand the idea of an innate genotypically determined fitness value independent of ecological circumstances.

A lot of animals compete directly with each other. For example, bucks fight with each other for mates. If the only selection pressure on deer is their competition for mates, then the best (strongest, fastest) bucks will have much higher fertility than the weak, sickly ones. This fertility rate may average 2 children per deer across the entire species (no matter how powerful the competition between bucks is), but higher levels of competition will skew the population towards the alleles of the biggest, strongest bucks.

You write:
> All of the talk about costs and accounting misses one
> simple, basic point. If a mutation is spreading through
> the population, whatever "costs" might exist are being
> paid.

This misses Remine's point. For this argument at least, evolution could be capable of the full molecules-to-man development... starting from now. But there hasn't been enough time for it to happen yet.

BTW, I believe in ReMine's table "cost of substitution" is the "the cost of not-entirely-random losses". Reading ReMine's description, I wouldn't have picked it out, but Haldane talks about "substitution cost" as the mechanism to drive beneficial alleles through the population.

Whether the specific numbers that ReMine concocts by directly and strictly applying Haldane's model to human evolution is, in terms of his overall argument (i.e., human evolution from an ape-like ancestor cannot possibly be accounted for by only 1667 fixed beneficial mutations) irrelevant.

An anlogy to ReMine's arguemnt:

You have a destination.

Say, Los Angeles.

You have a speed limit.

Say, 55 mph.

You have duistance travelled.

Say, 1667 miles.

You have a conclusion:

You could not possiobly have gotten to Los Angeles drivng 1667 miles at 55 mph. Therefore, you did not travel as claimed.

What is missing?

I gave a similar scenario to one of my upper-level biology classes a few years ago and they all got it right away.

But I have presented the conundrum to ReMnie himslef and several of his internet followers, and they all just dismiss it and claim it is 'posturing' and 'misrepresenting the numbers' and the like:

If you do not know where you started form, it is IMPOSSIBLE to claim that you could not reach your destination.

ReMine does not know what the ancestor was 10 million years ago, thus he cannot know what traits it did not have that must be accounted for by the 1667 fixed beneficial mutations.
He assumes that ALL changes from the ancestor were beneficial.
He assumes that all changes required a large number of beneficial mutations to arise (indicating his lack of in-depth knowledge on genetics and development). In fact, he writes in his book and elsewhere the rhetorical question 'is even 500,000 such changes enogh tio produce a sapien from a simian.'
His argument, even if the numbers are right, is literally without fouondation, but he has so muchinvested in it that he simply will not even consider the possibility that his argument is nto what he presents it as.