Ewens is not alone in his opinion that the substitution cost and other genetic load arguments do not place the kind of limits on the substitution rate that Haldane and Kimura imagined. In the chapter on genetic loads in "Evolution and the Genetics of Populations: Vol 3. Experimental results and Evolutionary Deductions" (published in 1977), Sewall Wright makes several statements that agree substantially with Ewens, including his conclusion that the load arguments are nonsense. Wright, along with Haldane and R.A. Fisher is one of the founders of modern evolutionary theory.
In the quotes that follow, I have made a few notes that are delineated with square brackets.
On page 480, Wright points out that there have been several attempts
at defining genetic loads:
"Morton, Crow, and Muler (1956) proposed a definition of the genetic load designed to include both of these types [segregation and heterotic loads]. In a later paper (1970) Crow gives three definitions, the broadest of which was as follows:""The genetic load is the fraction in which the population mean differs from a reference genotype. The trait measured is usually fitness or some component thereof and the reference genotype is usually the maximum among the actual or theoretically available types."
[Wright continues]
"Thus he had in mind a family of parameters rather than one narrowly defined one. Care must obviously be taken in applying the concept to specify the precise sense in which it is being used."
[On page 481, Wright concludes the section:]
"In general, the operation of selection is more complex than in these simple cases. There seems to be no general biological meaning of load as defined above other than the percentage difference of the mean selective values of the population in question from that of some reference genotype."
[In the next section on Multiple Loci, Wright concludes:]
"These formulas [for the calculation of loads due to multiple loci] must be used with caution since interactions with respect to selective value are usually present. If many loci are involved and loads per locus are not very small, the genotype that combines the reference genotypes of all loci is in general so rare theoretically that neither it nor anything approaching it exists in a finite population. In this case, Lt [total load summed over all loci] becomes practically meaningless, even if there are no interaction effects."
[Note that this conclusion exactly mirrors Ewens' more quantitative
arguments.]
[Skipping to page 487:]
"As pointed out by Sved, Reed, and Bodmer (1967) and by Milkman (1967), the load relative to the best genotype, where many loci are heterotic, may give
a grossly exaggerated impression on the strain on the reproductive capacity. This is because the best genotype is practically never present, or at all closely approached, in a finite population.""Consider the case of equal selective disadvantage of both of two homozygotes, s = t = 0.10, L = 0.05 at each locus. With 100 loci, selected independently, Lt [total load summed over all loci] = 1 = e-5 = 0.9993, a load so close to 1 that it would seem to be inevitably fatal to the species. With a gene frequency array (0.5A + 0.5A'), the distribution of heterozygotes (Het) and homozygotes (Hom) would be (0.5Het + 0.5Hom)100 for 100 loci, with mean 50 Het, standard deviations 5 Het, and thus would range from 35 to 65 heterozygous loci. The genotype with heterozygosis at all loci is so far outside this range that neither it nor anything approaching it ever occurs in the population. It is absurd to suppose that all genotypes produce as many offspring as would be produced by this 100-fold heterozygote, if it existed, but that these are reduced to infinitesimal numbers by the cumulative selective coefficients where there is homozygosis at 35 or more loci. Thus the load relative to the 100-fold heterozygote is meaningless. It would be more appropriate to take as a reference genotype one that is above the average by only two or three standard deviations."
[End of quotes from Sewall Wright]
Notice that the final quotation from Wright is exactly the same
argument Ewens made against the substitution cost that Wright has applied
to the heterozygosity load.
Further validation that Ewens has solved Haldane's Dilemma is found
from Austin Burt in a 1997 review of Kimura's collected works (Quarterly
Review
Biology 72, pg. 73). The quote is as follows:
"...Intermingled with these results is an insistence that most nucleotide and amino acid substitutions observed within and between species are selectively neutral - much the weakest part of the book. The original argument (Chapter 48, 1968) is based on Haldane's cost of natural selection, and was soon shown to be nonsense by Ewens (1970. Theoretical Population Biology 1:129-139). Subsequently, Kimura does not mention the cost of selection (nor Ewens), but instead argues..."
[End of quote from Burt]
Note however that Burt has missed the fact that Ewens did in fact address
the cost of selection again (Ewens, 1993). At any rate, Ewens arguments
were
so successful that Kimura never really countered them and subsequently
abandoned load theory as support for the neutral theory.
A much more recent reference to Ewens' work is found in "The Mathematical Theory of Selection, Recombination, and Mutation" , pg. 110 - 111 (Burger, R. 2000 ):
"Closely related to the genetic load is another concept introduced by Haldane (1957), namely 'the cost of natural selection' which is now called the substitutional load. Haldane considered an initially rare but favorable gene, and compared the amount of selective death, or cost, needed to carry out such a substitution. He defined the cost as the sum of all expressions of the form (3.1), but with Wmax replaced by the average fitness of the favorable allele, and the summation being over all generations required for the substitution. The surprising result is that this substitutional load is roughly independent of the selection coefficient and depends only on the initial frequency of the favorable mutant. The substitutional load has played a central role in the development of the neutral theory of evolution (cf. Kimura 1983). A comprehensive treatment of the theory of genetic loads may be found in Crow (1970, 1992). For well taken criticisms, in particular, concerning the substitutional load, see Ewens (1979)."