The Powers of Natural Selection

Natural selection ‘acts directly on the phenotype, and on the genotype only in so far as this expresses itself in the phenotype’ (Mather, 1953), for it is of course the individual animal, and not its genotype, that actually survives and reproduces, or fails to do so. So even indirectly it acts on the whole interacting gene complex, and not on single genes. However, it is possible, and indeed usual, to abstract and measure selection on single characters, genotypes of single genes, and even single genes themselves.

The starting-point for these measurements is the law ‘known after its discoverers as the Hardy-Weinberg equilibrium … Suppose we consider two alleles of a gene, A and A´ … Suppose the relative frequencies of the two alleles are p and q, where p + q = 1. In the next generation, there are three possible genotypes with respect to this gene: AA, A A´ and A´ A´. The law states that, if no selection is operating between the genotypes, the frequencies of the three genotypes will be p²:2pq:q²; moreover, the ratio will remain the same through all subsequent generations. All measurements of selection pressure are based on the observation of deviations from this ratio, persisting systematically through the generations.’ (Russell, 1959; for a thorough discussion of the equilibrium, see Wallace, 1968.)

There are various measures of selection. If one genotype is at a disadvantage, and its observed frequency is (1 - s) times its expected Hardy-Weinberg frequency, s is called the selection coefficient. It can then be inferred what is the selective advantage or disadvantage of the alleles, expressed as fractions of 1, or more usually as percentages. (Lerner, 1958, Wallace, 1968).

There are three types of natural selection: stabilising, directional and disruptive. The first two were defined by Schmalhausen in 1949 (Lerner, 1954), the third by Mather (1953) ,who gives the following description of the three types. ‘It is characteristic of continuous variation, as we see it in living organisms, that the extreme expressions of the character are rarer than those expressions nearer to the average. The frequency distribution approximates more or less to the normal: in Galton’s words, the majority are mediocre. Now the action of selection may be classified into three basic types. It may favour one extreme phenotype at the expense of all others, as is commonly the case with artificial selection in domesticated plants and animals (directional selection). It may favour the average expression at the expense of both extremes (stabilising selection). Or, finally, it may favour both extremes simultaneously, though not necessarily to the same extent, at the expense of the average (disruptive selection).’ The three types are illustrated in Figure 1, modified from Mather.

Since artificial selection has now been mentioned, I may note a different kind of selection measure, commonly used in artificial selection, namely selection intensity, which means the percentage of the population who reproduce at all (which of course the artificial breeder can determine). If this percentage is chosen as (say) those having the highest value of a character, it is possible to measure the selection differential, the deviation of the mean of the expected offspring of the chosen breeders from that of the original population. But allowance may have to be made for some of the breeders having more offspring than others. (Lerner, 1958). We shall see that this value may be greatly affected by natural selection opposing the artificial breeder.

We may finally note the rather obvious fact that the effect of selection on a character must depend on its heritability. Obviously if this is low the effect of selection will be much less than if it is high. Lerner (1958) gives an example of this in artificial selection of hens for number of eggs laid in November (low heritability, little effect) and weight of these eggs (high heritability, large effect). Of course this applies equally to natural selection.