Appendix A: Directional Selection


Eugene M. McCarthy, PhD Genetics

Directional selection is selection resulting in a directional shift in the population mean along a continuum, for example, the continuum height. In the absence of mutation, this sort of selection, in itself, can have only limited effect. That limit is reached when an optimum allele is present at all loci affecting a quantitative trait. The following reasoning explains why:

Table of Contents
Works Cited
If there were n loci for height and two extreme alleles, sᵢ and tᵢ, at any given locus i (where individuals with the allele tᵢ at locus i are taller than those having any other allele at locus i, and those with the allele tᵢ are shorter), then the shortest possible person would have the genetic constitution

s1s1, s2s2, s3s3, ... , snsn

(two alleles at each locus since humans are diploid), and the tallest would have

t1t1, t2t2, t3t3, ... tntn,

whereas, someone of intermediate height might have the constitution

s1t1, s2t2, s3t3, ... , sntn.

The variation of height in the population would be nearly continuous due to the many different possible combinations of alleles that a person might have at these loci. In a real situation there would also be an environmental component causing further variation. For example, difference in nutrition during growth would cause variation in height. However, under such circumstances, the amount that meiotic recombination could increase or decrease height would be limited. The shortest person that could be produced by directional selection would have the genetic constitution

s1s1, s2s2, s3s3, ... , snsn

and the tallest possible individual that could be produced by directional selection would have the alleles

t1t1, t2t2, t3t3, ... tntn.

Once all individuals in the population had two tall alleles at all n loci, further directional selection among variants produced by meiotic recombination alone would never produce taller progeny. In general, an optimal pair of alleles will exist for each locus affecting any quantitative trait. When the optimal pair for each such locus is actually present at each locus, then the limit of variation will have been reached. Directional selection will not be able to bring about a progression outside those limits unless mutation creates some new allele with more potent effect than any preexisting allele (or hybridization introduces such an allele).

The conclusions of the preceding paragraph can be reached without experiment or observation. They are a priori given the known nature of meiotic recombination. It would therefore be pointless to carry out experimentation to verify them. Nevertheless, in the first decades of the twentieth century before the nature of meiotic recombination was well understood, researchers actually did carry out experimentation and found that such limits can quickly be reached. These were the so-called "pure line" studies.

The first researcher to demonstrate means of experimentation this limitation of directional selection was the Danish botanist Wilhelm Johanssen (1903, 1915), who studied size change in an initially variable population of beans. In each generation he selected the largest and the smallest individuals and self-fertilized them. After only a few generations he had two stabilized populations, one small, the other large, and no amount of subsequent directional selection made beans in the larger line any larger. Nor did it make the small beans any smaller. The two lines had both become pure (genetically invariant) so that the small line was homozygous for small alleles at all size-governing loci, and the large line was homozygous for large alleles at all such loci. Johansen obtained similar results using barley.

Likewise, Pearl (1915) subjected chickens to long-term selection for increased egg production. Again, there were definite limits to the amount of increase that could be achieved. William Ernest Castle (1915), Sewell Wright's professor at the Bussey Institution, argued that Pearl's study was flawed. Castle asserted that his own experiments with hooded rats were much more reliable and that they had shown selection could go on changing a character indefinitely.

Pearl (1916) adequately defended himself against Castle's attack. Indeed, Castle believed in error that selection actually changed alleles, what were then called "Mendelian factors" (Provine 1986: 38-41). For example, he thought selection for darker rats would actually create alleles for darker coat color. This notion is false. In fact, experiments carried out in his own laboratory later convinced him of his error and he formally retracted his claim that selection could change "Mendelian factors" (Castle 1919a, 1919b). Since Castle's notion of creating new alleles through selection was erroneous, his belief that selection would continue to change coat color indefinitely in a certain direction was mistaken too. As soon as rats in Castle's study became homozygous at all loci affecting coat color, further selection would have been to no avail.

Studies like Castle's aimed at "taking variability beyond the range of variability in the original control stock." Castle's methods did in fact produce individuals with coat colors outside the range of variation seen in the control population. But such findings are spurious. They do not demonstrate selection can go on altering a trait indefinitely. Extreme individuals homozygous for all extreme alleles affecting a quantitative trait would occur with such low frequency that one would not expect to observe them in a small unselected laboratory population. But in a natural population, in the long run, they would in fact occur. So, due to small sample sizes, they would not be observed in any control stock, even though the alleles to produce them through meiotic recombination did exist in the control population. And yet they could be produced by continued selection.

Various other early workers carried out experiments similar to Johanssen's. In particular, Jennings' (1909) work with Paramecium and Hanel's (1907) with Hydra yielded similar results. For example, in selecting for size in Paramecium, Jennings (1909: 331) says that in initial generations great progress is made in increasing the mean size of the population.

But finally we reach a stage in which all but the largest race have been excluded. Thereafter we can make no further progress. In vain we choose for breeding the largest specimens of the lot; all belong to the same races so that all produce the same progeny. Selection has come to the end of its action.

Neo-Darwinians later viewed pure line research as an attack on the validity of the notion of evolution via natural selection (when it in reality it merely showed that the effects of selection are limited in the absence of mutation and hybridization). For this reason pure line research has often been misrepresented. For example, Mayr (1982: 585) says Johanssen claimed "selection cannot produce a deviation from the mean in a self-fertilizing species." What Johanssen actually said was that directional selection can produce a deviation from the mean, but that the deviation cannot be indefinitely increased. This is in fact true if the only source of variation is meiotic recombination (as has already been shown by a priori reasoning).

Later studies supposedly refuting these early findings, such as the much heralded work of Kettlewell (1955) on industrial melanism in moths, focused on the fact that steady changes in the mean measures of traits can be accomplished by directional selection in initially variable populations. But all such studies gloss over the fact that there is always a limit to such change when the source of variation is meiotic recombination alone.

In general, the karyotype defining a chromoset also defines the limits of variation that can be observed in the corresponding somaset since:

  1. A finite set of loci in that karyotype will have an effect on the expression of any given trait; and
  2. Each such locus will have a finite set of alleles. Some allele for each locus must exist that maximizes (or minimizes) expression of that trait.

Once an organism is homozygous at each locus for the allele having the extreme effect at that locus, directional selection can do no more (in the absence of mutation and hybridization).

Table of Contents >>

Appendix A: Directional Selection - ©