NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Griffiths AJF, Miller JH, Suzuki DT, et al. An Introduction to Genetic Analysis. 7th edition. New York: W. H. Freeman; 2000.

  • By agreement with the publisher, this book is accessible by the search feature, but cannot be browsed.
Cover of An Introduction to Genetic Analysis

An Introduction to Genetic Analysis. 7th edition.

Show details


So far in this chapter, we have considered changes in a population arising from forces of mutation, migration, recombination, and breeding structure. But these changes are random with respect to the way in which organisms make a living in the environments in which they live. Changes in a species in response to a changing environment occur because the different genotypes produced by mutation and recombination have different abilities to survive and reproduce. The differential rates of survival and reproduction are what is meant by selection, and the process of selection results in changes in the frequencies of the various genotypes in the population. Darwin called the process of differential survival and reproduction of different types natural selection by analogy with the artificial selection carried out by animal and plant breeders when they deliberately select some individuals of a preferred type.

The relative probability of survival and rate of reproduction of a phenotype or genotype is now called its Darwinian fitness. Although geneticists sometimes speak loosely of the fitness of an individual, the concept of fitness really applies to the average survival and reproduction of individuals in a phenotypic or genotypic class. Because of chance events in the life histories of individuals, even two organisms with identical genotypes and identical environments will differ in their survival and reproduction rates. It is the fitness of a genotype on average over all its possessors that matters.

Fitness is a consequence of the relation between the phenotype of the organism and the environment in which the organism lives, so the same genotype will have different fitnesses in different environments. In part, this difference is because exposure to different environments during development will result in different phenotypes for the same genotypes. But, even if the phenotype is the same, the success of the organism depends on the environment. Having webbed feet is fine for paddling in water but a positive disadvantage for walking on land, as a few moments spent observing a duck walk will reveal. No genotype is unconditionally superior in fitness to all others in all environments.

Furthermore, the environment is not a fixed situation that is experienced passively by an organism. The environment of an organism is defined by the activities of the organism itself. For example, dry grass is part of the environment of a junco, so juncos that are most efficient at gathering it may waste less energy in nest building and thus have a higher reproductive fitness. But dry grass is part of a junco’s environment because juncos gather it to make nests. The rocks among which the grass grows are not part of the junco’s environment, although the rocks are physically present there. But the rocks are part of the environment of thrushes; these birds use the rocks to break open snails. Moreover, the environment that is defined by the life activities of an organism evolves as a result of those activities. The structure of the soil that is in part determinative of the kinds of plants that will grow is altered by the growth of those very plants. Environment is both the cause and the result of the evolution of organisms. As primitive plants evolved photosynthesis, they changed the earth’s atmosphere from one that had had essentially no free oxygen and a high concentration of carbon dioxide to the atmosphere that we know today, which contains 21 percent oxygen and only 0.03 percent carbon dioxide. Plants that evolve today must do so in an environment created by the evolution of their own ancestors.

Darwinian, or reproductive, fitness is not to be confused with “physical fitness” in the everyday sense of the term, although they may be related. No matter how strong, healthy, and mentally alert the possessor of a genotype may be, that genotype has a fitness of zero if for some reason its possessors leave no offspring. Thus such statements as the “unfit are outreproducing the fit, so the species may become extinct” are meaningless. The fitness of a genotype is a consequence of all the phenotypic effects of the genes involved. Thus, an allele that doubles the fecundity of its carriers but at the same time reduces the average lifetime of its possessors by 10 percent will be more fit than its alternatives, despite its life-shortening property. The most common example is parental care. An adult bird that expends a great deal of its energy gathering food for its young will have a lower probability of survival than one that keeps all the food for itself. But a totally selfish bird will leave no offspring, because its young cannot fend for themselves. As a consequence, parental care is favored by natural selection.

Two forms of selection

Because the differences in reproduction and survival between genotypes depend on the environment in which the genotypes live and develop and because organisms may alter their own environments, there are two fundamentally different forms of selection. In the simple case, the fitness of an individual does not depend on the composition of the population; rather it is a fixed property of the individual’s phenotype and the external physical environment. For example, the relative ability of two plants that live at the edge of the desert to get sufficient water will depend on how deep their roots grow and how much water they lose through their leaf surfaces. These characteristics are a consequence of their developmental patterns and are not sensitive to the composition of the population in which they live. The fitness of a genotype in such a case does not depend on how rare or how frequent it is in the population. Fitness is then frequency independent.

In contrast, consider organisms that are competing to catch prey or to avoid being captured by a predator. Then the relative abundances of two different genotypes will affect their relative fitnesses. An example is Mullerian mimicry in butterflies. Some species of brightly colored butterflies (such as monarchs and viceroys) are distasteful to birds, which learn, after a few trials, to avoid attacking butterflies with that pattern. If two species differ in pattern, there will be selection to make them more similar because both will be protected and they share the burden of the birds’ initial learning period. The less frequent pattern will be at a disadvantage with respect to the more frequent one, because birds will less often learn to avoid them. Within a species, rarer patterns will be selected against for the same reason. The rarer the pattern, the greater is the selective disadvantage, because birds will be unlikely to have had a prior experience of a low-frequency pattern and therefore will not avoid it. This selection to blend in with the crowd is an example of frequency-dependent fitness.

For reasons of mathematical convenience, most models of natural selection are based on frequency-independent fitness. In fact, however, a very large number of selective processes (perhaps most) are frequency dependent. The kinetics of the evolutionary process depend on the exact form of frequency dependence, and, for that reason alone, it is difficult to make any generalizations. The result of positive frequency dependence (such as competing predators, where fitness increases with increasing frequency) is quite different from that of negative frequency dependence (where fitness of a genotype declines with increasing frequency). For the sake of simplicity and as an illustration of the main qualitative features of selection, we deal only with models of frequency-independent selection in this chapter, but convenience should not be confused with reality.

Measuring fitness differences

For the most part, the differential fitness of different genotypes can be most easily measured when the genotypes differ at many loci. In very few cases (except for laboratory mutants, horticultural varieties, and major metabolic disorders) does the effect of an allelic substitution at a single locus make enough difference to the phenotype to be reflected in measurable fitness differences. Figure 24-8 shows the probability of survival from egg to adult—that is, the viability—of a number of second-chromosome homozygotes of D. pseudoobscura at three different temperatures. As is generally the case, the fitness (in this case, a component of the total fitness, viability) is different in different environments. A few homozygotes are lethal or nearly so at all three temperatures, whereas a few have consistently high viability. Most genotypes, however, are not consistent in viability between temperatures, and no genotype is unconditionally the most fit at all temperatures. The fitness of these chromosomal homozygotes was not measured in competition with each other; all are measured against a common standard, so we do not know whether they are frequency dependent. An example of frequency-dependent fitness is shown in the estimates for inversion homozygotes and heterozygotes of D. pseudoobscura in Table 24-10.

Figure 24-8. Viabilities of various chromosomal homozygotes of Drosophila pseudoobscura at three different temperatures.

Figure 24-8

Viabilities of various chromosomal homozygotes of Drosophila pseudoobscura at three different temperatures.

Table 24-10. Comparison of Fitnesses for Inversion Homozygotes and Heterozygotes in Laboratory Populations of Drosophila pseudoobscura When Measured in Different Competitive Combinations.

Table 24-10

Comparison of Fitnesses for Inversion Homozygotes and Heterozygotes in Laboratory Populations of Drosophila pseudoobscura When Measured in Different Competitive Combinations.

Examples of clear-cut fitness differences associated with single-gene substitutions are the many “inborn errors of metabolism,” where a recessive allele interferes with a metabolic pathway and causes lethality of the homozygotes. An example in humans is phenylketonuria, where tissue degeneration is the result of the accumulation of a toxic intermediate in the pathway of tyrosine metabolism. A case that illustrates the relation of fitness to environment is sickle-cell anemia. An allelic substitution at the structural-gene locus for the β chain of hemoglobin results in substitution of valine for the normal glutamic acid at chain position 6. The abnormal hemoglobin crystallizes at low oxygen pressure, and the red cells deform and hemolyze. Homozygotes Hb S/Hb S have a severe anemia, and survivorship is low. Heterozygotes have a mild anemia and under ordinary circumstances exhibit the same or only slightly lower fitness than normal homozygotes Hb A/Hb A. However, in regions of Africa with a high incidence of falciparum malaria, heterozygotes (Hb A/Hb S) have a higher fitness than normal homozygotes because the presence of some sickling hemoglobin apparently protects them from the malaria. Where malaria is absent, as in North America, the fitness advantage of heterozygosity is lost.

It has not been possible to measure fitness differences for most single-locus polymorphisms. The evidence for differential net fitness for different ABO or MN blood types is shaky at best. The extensive enzyme polymorphism present in all sexually reproducing species has for the most part not been connected with measurable fitness differences, although, in Drosophila, clear-cut differences in the fitness of different genotypes have been demonstrated in the laboratory for a few loci such as those encoding α-amylase and alcohol dehydrogenase.

How selection works

The simplest way to see the effect of selection is to consider an allele, a, that is completely lethal before reproductive age in homozygous condition, such as the allele that leads to Tay-Sachs disease. Suppose that, in some generation, the allele frequency of this gene is 0.10. Then, in a randommating population, the proportions of the three genotypes after fertilization are

Image ch24e19.jpg

At reproductive age, however, the homozygotes a/a will have already died, leaving the genotypes at this stage as

Image ch24e20.jpg

But these proportions add up to only 0.99 because only 99 percent of the population is still surviving. Among the actual surviving reproducing population, the proportions must be recalculated by dividing by 0.99 so that the total proportions add up to 1.00. After this readjustment, we have

Image ch24e21.jpg

The frequency of the lethal a allele among the gametes produced by these survivors is then

Image ch24e22.jpg

and the change in allelic frequency in one generation, expressed as the new value minus the old one, has been 0.091 − 0.100 = −0.019. We can repeat this calculation in each successive generation to obtain the predicted frequencies of the lethal and normal alleles in a succession of future generations.

The same kind of calculation can be carried out if genotypes are not simply lethal or normal, but if each genotype has some relative probability of survival. This general calculation is shown in Box 24-6. After one generation of selection, the new value of the frequency of A is equal to the old value (p) multiplied by the ratio of the average fitness of A alleles to the fitness of the whole population. If the fitness of A alleles is greater than the average fitness of all alleles, then Image Wbar.jpg A /Image Wbar.jpg is greater than unity and p′ is larger than p. Thus, the allele A increases in the population. Conversely, if Image Wbar.jpg A /Image Wbar.jpg is less than unity, A decreases. But the mean fitness of the population (Image Wbar.jpg) is the average fitness of the A alleles and of the a alleles. So if Image Wbar.jpgA is greater than the mean fitness of the population, it must be greater than Image Wbar.jpga, the mean fitness of a alleles.

Box Icon

Box 24-6

Effect of Selection on Allele Frequency. Suppose that a population is mating at random with respect to a given locus with two alleles and that the population is so large that (for the moment) we can ignore inbreeding. Just after the eggs have been fertilized, (more...)


The allele with the higher average fitness increases in the population.

It should be noted that the fitnesses WA / A , WA / a , and Wa / a may be expressed as absolute probabilities of survival and absolute reproduction rates or they may all be rescaled relative to one of the fitnesses, which is given the standard value of 1.0. This rescaling has absolutely no effect on the formula for p′, because it cancels out in the numerator and denominator.


The course of selection depends only on relative fitnesses.

An increase in the allele with the higher fitness means that the average fitness of the population as a whole increases, so selection can also be described as a process that increases mean fitness. This rule is strictly true only for frequency-independent genotypic fitnesses, but it is close enough to a general rule to be used as a fruitful generalization. This maximization of fitness does not necessarily lead to any optimal property for the species as a whole, because fitnesses are only defined relative to each other within a population. It is relative (not absolute) fitness that is increased by selection. The population does not necessarily become larger or grow faster, nor is it less likely to become extinct.

Rate of change in gene frequency

The general expression for the change in gene frequency, derived in Box 24-6, is particularly illuminating. It says that Δp will be positive (A will increase) if the mean fitness of A alleles is greater than the mean fitness of a alleles, as we saw before. But it also shows that the speed of the change depends not only on the difference in fitness between the alleles, but also on the factor pq, which is proportional to the frequency of heterozygotes (2pq). For a given difference in fitness of alleles, gene frequency will change most rapidly when the alleles A and a are in intermediate frequency, so pq is large. If p is near 0 or 1 (that is, if A or a is nearly fixed), then pq is nearly 0 and selection will proceed very slowly. Figure 24-9 shows the S-shaped curve that represents the course of selection of a new favorable allele A that has recently entered a population of homozygotes a/a. At first, the change in frequency is very small because p is still close to 0. Then it accelerates as A becomes more frequent, but it slows down again as A takes over and a becomes very rare. This is precisely what is expected from a selection process. When most of the population is of one type, there is nothing to select. For evolution by natural selection to occur, there must be genetic variance; the more variance, the faster the process.

Figure 24-9. The time pattern of increasing frequency of a new favorable allele A that has entered a population of a/a homozygotes.

Figure 24-9

The time pattern of increasing frequency of a new favorable allele A that has entered a population of a/a homozygotes.

One consequence of the dynamics shown in Figure 24-9 is that it is extremely difficult to significantly reduce the frequency of an allele that is already rare in a population. Thus, eugenic programs designed to eliminate deleterious recessive genes from human populations by preventing the reproduction of affected persons do not work. Of course, if all heterozygotes could be prevented from reproducing, the gene could be eliminated (except for new mutations) in a single generation. Because every human being is heterozygous for a number of different deleterious genes, however, no one would be allowed to reproduce.

When alternative alleles are not rare, selection can cause quite rapid changes in allelic frequency. Figure 24-10 shows the course of elimination of a malic dehydrogenase allele in a laboratory population of D. melanogaster. The fitnesses in this case are:

Image ch24e32.jpg
The frequency of a is not reduced to 0, and further reduction in frequency will require longer and longer times, as shown in the negative eugenics case.

Figure 24-10. The loss of an allele of the malic dehydrogenase locus MDH F due to selection in a laboratory population of Drosophila melanogaster.

Figure 24-10

The loss of an allele of the malic dehydrogenase locus MDH F due to selection in a laboratory population of Drosophila melanogaster. The red dashed line shows the theoretical curve of change computed for fitnesses WA / A  = 1.0, WA / (more...)


Unless alternative alleles are present in intermediate frequencies, selection (especially against recessives) is quite slow. Selection is dependent on genetic variation.

By agreement with the publisher, this book is accessible by the search feature, but cannot be browsed.

Copyright © 2000, W. H. Freeman and Company.
Bookshelf ID: NBK22009


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...