SELECTION IN NATURAL POPULATIONS. III. MEASUREMENT AND ESTIMATION

Abstract

Five related terms are in current use to describe the effects of selection on a population: the load, cost, intensity, pressure, and coefficient of selection. The cost of selection (Haldane, 1957, 1960; MacArthur, 1962) is an aspect of the genetic load, and has been called the substitutional by Kimura (1960). Cost and represent (with the qualification on density-dependence below) the reduction in fitness of a population below the level it would attain if composed entirely of the optimum genotype or combination of genotypes in it at the time measured. This statistic, like most ranges, can be investigated experimentally, and estimates of it on the phenotypic level are given below for several characters of a horse. The load is the price a population pays for at least much of its genetic variance, and is presumably balanced on the average, through interspecific selection, by a greater probability of extinction (cf. Lewontin, 1961) with a smaller genetic variance. The complete redefinition of load by Dobzhansky and Spassky (1963) is unnecessary. To the extent that population size is determined by density-dependent factors (as by the carrying capacity of the environment in relation to the present population genome), some fertility reduction or pre-reproductive mortality is however necessary. Mortality therefore need not be considered a burden on a population of stable size if it does not exceed the proportion of deaths caused by density-dependent factors (and similarly for fertility reduction). The occurrence within a population of genotypes with differential sensitivity to density makes the estimation of this proportion more difficult but does not affect the general conclusion on the effect of density-dependence. I no longer believe, however (contrary to Van Valen, 1963d; Mayr, 1963; and Brues, 1964), that this kind of argument allows a large increase in the rate of genotypic evolution. It is perfectly true, as these authors have noted in somewhat different ways, that selection for generally favorable alleles will improve population fitness over what it would have been without such alleles. But it does not follow from the fact that a change will be good for a population, that it will happen. This statement is obviously true, as I realized earlier (Van Valen, 1963d), for selection produced by a deteriorating environment. If in such an environment 10 independent allelic changes will be necessary for survival in a few generations and all the now favorable alleles have an initial frequency of 1 0-3, even with no dominance an enormous initial population size will be necessary for a reasonable probability of any survival through the crisis for a species with relatively little density-dependent wasting of immatures. Otherwise the population will become extinct before it can adapt itself. For general adaptations, not related to a deteriorating environment, the result is similar although less obvious. Any density-dependent wastage will permit some selection that is not a burden, the amount varying with the species or population. When this is used up (even ignoring random deaths and fertility differentials), simultaneous selection at two or more recombining loci produces a mutual interference, this interference increasing at a greater than linear rate with the number of loci and the mean selection intensity and resulting from biological limits on increase of population size. This interference (between loci, not with species survival) exists even though all replaceIFor J. B. S. Haldane (1892-1964), one of the greatest geneticists of this century.