NOTE 23F: With asexual reproduction, this is true regardless of how the mutations interact with each other (see p. 680).
NOTE 23G: Sniegowski et al. (2000) review the evolution of mutation rates and briefly discuss the issues raised in this problem. Several more recent analyses include André and Godelle (2006).
NOTE 23H: This is known as the twofold cost of meiosis or the cost of genome dilution (see p. 669).
NOTE 23I: There are many cases where parthenogenetic females still need to mate with males in order to reproduce (see, e.g., the Poeciliopsis minnows [p. 666]). If males only need to mate, without providing parental care, then selection to avoid parthenogens may be weak.
NOTE 23J: If there is some reallocation of resources from pollen to seed production, then homozygotes will increase in frequency.
NOTE 23K: The expected frequency of an allele that is destined to fix is higher than this deterministic calculation, by a factor 1/2s (see Fig. P18.4). This does not much alter the answer, however.
NOTE 23L: This argument was first made by Crow and Kimura (1965), although they omitted the factor of 2s that accounts for the probability of fixation of each mutation. Maynard Smith (1968) criticizes their assumption that each mutation is unique and later (1974) reviews the various different arguments on the relative rates of evolution of sexual and asexual populations. See also Crow and Kimura’s (1969) reply.
NOTE 23M: More generally, greater diversity within families increases the selective advantage of alleles that enhance competitive ability; if such alleles also enhance the ability to compete with survivors from other families, then this leads to an advantage for recombination, multiple matings, etc. (see p. 670).
NOTE 23N: This is an example of the Levene model at two loci (see p. 673).
NOTE 23O: In reality, there are a finite number of genes, of finite effect, and so selection will gradually reduce variation. In the long run, continued response must rely on new mutations (pp. 482–484).
NOTE 23P: There is an interesting way to work out the equilibrium number of mutations. When fitness is plotted against number of mutations on a log scale (as in Fig. P23.6), the slope of the fitness curve equals the selection coefficient s (i.e., the increase in relative fitness due to one extra mutation). Because the equilibrium number of mutations is U/s, the vertical distance between the intercept of the tangent, and the point where it touches the curve is just U (Fig. P23.6). When the fitness curves downward, as here, this is much greater than the actual mutation load L.
NOTE 23Q: This can be found more directly from the leading eigenvalue of the recursion; see Chapter 28.