107

CHAPTER 17

POPULATION GENETICS AND

EVOLUTION
The mechanism of heredity and variation is basic to the study of
evolution and was therefore a major concern of Darwin and many of his
followers, including Galton and Weismann. It was an interest in evolu-
tion that led several of the earlier Mendelians, such as de Vries and
Bateson, to the study of heredity. But, with the discovery of Mendels
work in 1900, the development of the new methods caused a temporary
lack of interest in the evolutionary implications. As Bateson put it in
1909:
It is as directly contributing to the advancement of pure physio-
logical science that genetics can present the strongest claim. We have
an eye always on the evolution-problem. We know that the facts we
are collecting will help in its solution; but for a period we shall per-
haps do well to direct our search more especially to the immediate
problems of genetic physiology . . . willing to postpone the applica-
tion of the results to wider problems as a task more suited to a
maturer age.
Evolution is concerned with changes in populations, rather than in
individuals, and what was needed was an analysis of the effects of the
Mendelian scheme on populations of interbreeding individuals. The be-
ginning of this analysis is a paper by Yule (1902) in which he pointed out
that, if the members of an F
2
population, segregating for a single pair of
genes (A and a), interbreed at random, the three types of individuals (AA,
Aa, aa) will be represented in the same proportions in the following gen-
erations. He also raised the question: What will happen if all the aa indi-
viduals are removed? His analysis here was in error, but it was corrected
by Castle in 1903. Castle pointed out also that, if such selection ceases in
any generation, the newly established proportions will then be stable.
108 A HI STORY OF GENETICS
Here was the essence of the basic formula of population genetics, though
it was derived by a longhand method and was not stated in simple al-
gebraic form. The result, without selection, was also derived by Pearson
(1904), that is, for the case where p = q, where p = the frequency of A
genes, q = the frequency of a, and p + q = 1.
The generalization that the stable frequency of genotypes is
p
2
AA : 2pqAa : q
2
aa was made by Hardy and by Weinberg, independ-
ently, in 1908. Both knew of Pearsons result. To Hardy, who was a
mathematician, the generalized result seemed so self-evident that he
commented: . . . I should have expected the very simple point which I
wish to make to have been familiar to biologists. That it was not famil-
iar is shown by the fact that it had been seriously suggested that domi-
nant genes would automatically increase in frequency in mixed
populations.
The Hardy-Weinberg formula is strictly valid only if several condi-
tions are fulfilled:
1. The population must be large enough so that sampling errors can be
disregarded. As Hardy pointed out, any chance deviations in the val-
ues of p and q will be as stable in succeeding generations as were
those of the preceding one.
2. There must be no mutation, since change of A to a or of a to A will
alter the values of p and q.
3. There must be no selective mating.
4. There must be no selection, that is, A and a must have no differential
effect on the reproductive capacity of individuals bearing them.
(In the requirements 2 and 4, the wording given is not strictly cor-
rect, since it is possible that balanced mutation and selection rates may
existin which case there will be no net changes in the frequencies of A
and a.)
These are rather stringent requirements, and it may be doubted if
they are ever fully met; nevertheless, they are often approximated closely
enough to make the formula useful in analysis of populations.
The further developments in this field have depended on the alge-
braic analysis of the effects of deviations from the four requirements.
This development dates from Haldanes (1924 and later) analysis of the
effects of selection. He determined the number of generations required to
alter gene frequencies, as related to the intensity of selection. This was
worked out for dominant and for recessive genes, both in haploid and in
diploid organisms, and for sex-linked as well as for autosomal genes.
This analysis has been followed by detailed studies on the algebraic
POPULATION GENETICS AND EVOLUTION 109
consequences of variations in each of the four requirements by Haldane,
Fisher, Wright, and others. Perhaps the most useful general summary is
that of Li (1955).
One of the early developments was a theory of the origin of domi-
nance (Fisher, 1928). Fisher suggested that mutant genes are inher-
ently neither strictly dominant nor strictly recessive but produce more
or less intermediate heterozygotes. There are, however, numerous
modifying genes that affect their dominance. Since most mutant genes
have unfavorable effects on their bearers, an individual heterozygous
for a mutant gene will leave more offspring if the modifiers it carries
happen to make the gene more nearly recessive. This effect will be
slight, since an unfavorable semidominant mutant will not persist long
in the population. However, since mutants continually recur with a
low frequency, the effects will be cumulative over very long periods.
This hypothesis has played a large part in discussions of population
genetics but has been criticized by Wright and others on several
grounds. The postulated effect of modifiers is a second-order effect,
and it seems likely that their frequencies in the population will usu-
ally be determined by more direct effects. Also, an alternative inter-
pretation is that a favorable gene will undergo selectionprobably
largely among alleles of the gene itselffor a factor of safety, so
that it will be capable of producing an excess of its useful product in
times of stress; and such an excess may be supposed to result directly
in dominance.
The first attempt to assemble a coherent general account of the alge-
braic analysis of Mendelian population behavior was Fishers book in
1930. Widely read and discussed, it certainly strongly influenced further
developments. One of the elements it minimized was that of the effects
of population size. This was analyzed by Wright in several papers, first
summarized in 1932.
Wright pointed out that in small populations there is a possibility of
chance shifts in gene frequencies which are not controlled by selection
and that this may lead to the production of combinations of genes
sometimes favorablethat would have almost no chance of being pro-
duced in large populations. He suggested that the most favorable condi-
tion for rapid evolution is that of a large population that is split
(geographically or otherwise) into a series of relatively small subpopu-
lations, with gene flow between these possible, but strongly restricted.
Under these conditions, what has come to be called random drift may
produce more favorable combinations of genes in some subpopulations,
110 A HI STORY OF GENETICS
and these will gradually spread to neighboring populations. More often,
probably, the result of random drift will be a less favorable combination
of genes; when this happens, the subpopulation will diminish and proba-
bly be replaced by migrants from neighboring areas. Selection still re-
mains the determining element, but emphasis is placed on selection
among subpopulations rather than solely among individuals.
These analyses are largely theoretical, based on laboratory experi-
mentsnot on the actual properties of naturally occurring wild popula-
tions. It has turned out, not unexpectedly, that such populations are
difficult to study. The situations encountered are so complex that it be-
comes difficult to evaluate separate variables. Further, when a quantita-
tive analysis is made, it is specific for the population studied and cannot
safely be applied to other populations, even of the same species. In other
words, generalizations are difficult and dangerous. One result often ob-
tained is that new kinds of problems have been suggested for further ex-
act algebraic analyses.
In spite of these difficulties, much progress has been made. The first
attempt to coordinate the existing quantitative data on natural popula-
tions and interpret them in terms of algebraic studies, was the book by
Dobzhansky (1937).
It had, of course, long been evident that inherited diversities do occur
in natural populations. Such examples as separate sexes, or heterostyly in
plants, are obviously special cases. But the occurrence of sports and of
inherited slight differences in such things as size, cold-resistance and
color, among other things, was also familiar. It was not so clear that there
is a store of recessive genes suitable for exact study by Mendelian meth-
ods. Examples of these had also been found by many observers, but the
first attempts at a quantitative determination of their frequency in wild
populations seem to have been made by a series of Russian investigators,
using Drosophila. This work was initiated by Chetverikov (sometimes
transliterated as Tschetwerikoff) in 1926 and culminated in the work of
Dubinin and collaborators (1934, 1936), who studied a large series of
wild populations of Drosophila melanogaster collected in the Caucasus.
They found that up to 16 percent of the second chromosomes in these
populations contained recessive lethals. It had not been supposed that
wild populations contained such high frequencies of unfavorable reces-
sives, but further work by Timofeff-Ressovsky, Sturtevant, Dobzhansky
and co-workers, and others has confirmed the result for several species
and also has, in some cases, shown even higher frequencies of lethals and
other unfavorable recessives.
Out of this work grew the realization that natural populations are full
POPULATION GENETICS AND EVOLUTION 111
of hidden genetic variability, most of it potentially unfavorable. The
study of this genetic load is now being actively prosecuted, in part be-
cause of its importance in the practical breeding of animals and plants,
and because of the increase in frequency of mutations in human popula-
tions that must be supposed to have followed increases in irradiation
caused by medical and dental uses of X-rays and by exposure to radioac-
tive fallout from atomic bombs.
Another approach to the study of the relation between genetics and
evolution is through the use of species hybrids. This, as pointed out in
Chapter 1, has a long history; but only with the development of Men-
delian and cytological methods did it begin to yield really helpful results.
It was soon apparent that different species usually differ in many
pairs of genes and accordingly give numerous recombination types in F
2
;
Mendel himself indicated this in his discussion of Grtners work. With
the development of the multiple-gene interpretation of quantitative in-
heritance (Chapter 9), it became possible to give a more precise inter-
pretation of the results.
One of the awkward circumstances about discussions in this field
is the ambiguity in the use of the word species. There is no generally
accepted definition, but most discussions have centered on the extent
to which cross-sterility between populations is to be taken into ac-
count in deciding whether or not two groups are to be considered as
separate species. The nature and origin of interspecific sterility has
been recognized since Darwin as one of the major problems of evolu-
tion.
The genetic study of the nature of interspecific sterility is difficult. If
sterility is complete, this very fact makes it impossible to study by con-
ventional genetic methods; if it is partial, there usually arises a possibil-
ity (or certainty in some cases) that it involves so great a distortion of
segregation and of the relative viabilities of the various recombination
products as to render analysis difficult or impossible. When these com-
plications are absent or unimportant in any given case it becomes ques-
tionable whether the data obtained are relevant to the general problem.
The outcome of crosses between distinct species varies widely
from one case to the next. The eggs and sperm may never come to-
gether (through lack of mating or failure of pollen tube growth); they
may fail to fuse even if they do come together; if fertilization is ef-
fected, some or all of the chromosomes derived from the sperm may
be eliminated at cleavage in the foreign cytoplasm (Baltzer, 1909);
cleavage and mitosis may be normal, with the development of the go-
112 A HI STORY OF GENETICS
nads of the hybrid blocked at almost any pointmost often just be-
fore meiosis; meiosis may be abnormal or essentially normal; the
gametes may be normal in structure and be functional but may give
rise to some or many abnormal or sterile individuals in the next gen-
eration. In other words, there are many different kinds of mechanisms
that prevent or hinder interspecific exchange of genes. The problem
is, how do these arise?
There is probably no single general answer to this question, but
there is an answer applicable to many cases, namely, polyploidy.
The study of polyploidy may be dated from the work of Boveri
and others in the 1880s and 1890s on the two races of Ascaris
megalocephalaunivalens, with one pair of chromosomes in the
germ line, and bivalens, with two pairs. There is some question about
this being a simple case of polyploidy. The first unambiguous cases
seem to be those reported by the Marchals (1906, 1907) in mosses
(described in Chapter 13) and since studied in great detail by F. von
Wettstein, and in Oenothera by Lutz, also in 1907 (see Chapter 10).
Other more or less probable examples followed (Strasburger, Tisch-
ler, and others) together with cytological studies of meiosis in the
triploids produced by crosses between diploids and tetraploids. The
results were confusing and contradictory, until the analysis by Winge
(1917), which began to clarify the situation.
Winge made a detailed study of the available data on chromosome
numbers in plants and found that in many groups there was a basic
number, with various multiples of this number represented in different
species. He pointed out that, in a hybrid between two species, it
sometimes happens that some (or all) chromosomes are sufficiently
different so that they do not pair at meiosis, so that the resulting gam-
etes do not all contain a single complete set of chromosomes, and
partial, or essentially complete, sterility results. Now, if the chromo-
somes of such a hybrid are doubled (by chromosome division without
an accompanying cell division), each chromosome will now have an
exact mate, and meiosis can be expected to be normalwith a resto-
ration of fertility, as had already been shown by Federley (1913). This
process was demonstrated by Clausen and Goodspeed (1925) and
Clausen (1928), in Nicotiana. N. tabacum (24 pairs of chromosomes)
was crossed to N. glutinosa (12 pairs). The hybrid, with 36 chromo-
somes, was sterile and at meiosis showed only a few paired chromo-
somes. One hybrid individual was fertile, however, and was found to
be a tetraploid, with 72 chromosomes that formed 36 bivalents at
POPULATION GENETICS AND EVOLUTION 113
meiosis; this plant was fully fertile and bred true to type.
The terminology which was suggested by Kihara and Ono (1926) has
been generally accepted and has helped to clarify the relations. They
suggested that:
Under polyploidy we must distinguish two phenomena, namely,
autopolyploidy and allopolyploidy. Autopolyploidy signifies the dou-
bling of the same chromosome set; allopolyploidy the multiplication
of different chromosome sets brought together by hybridization.
This distinction has proved to be convenient and usefulthough
it is not of an all-or-none type, since intermediate conditions occur.
Allopolyploidy has been found to be widespread in the higher plants,
but is rare in animals. Muller (1925) suggested that this is because it
leads to difficulties in sex determination in species with separate
sexes. It now seems more probable that the difficulty lies usually with
the crossing of new tetraploids to diploids and the production of rela-
tively sterile triploids; self-fertilizing hermaphrodites can avoid this
difficulty.
The relative (or complete) sterility of the triploids is, however, of
importance in that it leads to an immediate effective sexual isolation of
an allopolyploid from both its parental forms. In fact, this sterility often
operates to make difficult the production of any backcross offspring at all
(Karpechenko and others), presumably because of an interaction between
the tissues of the style and of the pollen tube, since it is known that, in
some autopolyploid series, haploid pollen functions best in diploid styles,
and diploid pollen in tetraploid styles.
Polyploidy, then, does form one method of bringing about interspeci-
fic sterility, but even in the higher plants where it is relatively frequent, it
must be considered a rather unusual cause that has little bearing on the
general question.
It should be added that the study of polyploidy has been of impor-
tance in other directions in genetics. Its use by Bridges and others in the
analysis of sex determination has been discussed in Chapter 13, and it
has also contributed largely to our understanding of chromosome me-
chanics, as developed by many authors. There have also been extensive
applications in the breeding of cultivated plants. One of the important
events in this field was the discovery (Blakeslee and A. G. Avery, 1937)
that doubling of the chromosome set in plants may be induced by the use
of colchicine.
Winges interpretation of polyploidy grew out of a comparison of the
114 A HI STORY OF GENETICS
chromosomes of related species. Another type of conclusion was based
on such comparative studies by Metz (1914 and later) with species of
Drosophila and by Robertson (1916) with a series of genera and species
of grasshoppers. In both of these groups there are rod-shaped chromo-
somes, and also (in other species) V-shaped ones. Both authors found
that, if each V was counted as two rods, the haploid number of elements
was constant within a given group: 5 in Drosophila, if the small dot (of-
ten difficult to see) is neglected; 7 in the grouse-locusts, and 12 in ordi-
nary grasshoppers. The conclusion was drawn that, in general, the
elements maintain their individuality (to a large extent at least) within
such groups, and are separated, or united, in various ways in different
species.
This conclusion was questioned by R. C. Lancefield and Metz (1921)
as a result of studies on Drosophila willistoni. This species, like D.
melanogaster, has a pair of rods and two pairs of Vs, but in melanogas-
ter the rod is the X chromosome, whereas in willistoni, Lancefield and
Metz showed that one of the Vs is the X.
This anomalous result was explained later as a result of studies on
the mutant genes found in various species of Drosophila. Such studies,
by Metz, Sturtevant, D. E. Lancefield, Weinstein, Chino, Moriwaki, and
others, showed that mutants with phenotypes closely resembling those of
melanogaster could often be found in other species, and in the case of
simulans, which can be crossed to melanogaster (although the hybrids are
all completely sterile), it was possible to show that many of these resem-
blances are in fact due to mutations of the same wild-type genes (Stur-
tevant, 1921). This conclusion rests on more indirect evidence for the
other species, but, as more examples accumulated, one rule appeared:
mutants in other species that closely resemble sex-linked mutants in
melanogaster are also sex-linkedthough the reverse relation does not
hold so consistently. In retrospect, it is obvious that most of the clear
exceptions to the reverse rule concerned mutants in D. willistoni and D.
pseudoobscura that were sex-linked in those species but resembled auto-
somal mutants in melanogaster; in both of these species the X was known
to be a V.
The obvious conclusion was implied by D. E. Lancefield (1922) but
was not elaborated or made more specific until Crew and Lamy (1935)
carried out a detailed comparison of the known mutants of pseudo-
obscura with the apparently similar types of melanogaster. This paper
was not very critical and used terminology that made it difficult to under-
stand, but the conclusions were fully confirmed and extended by the
more substantial accounts of Donald (1936) and of Sturtevant and Tan
POPULATION GENETICS AND EVOLUTION 115
(1937). These results confirmed the conclusion of Metz and Robertson:
the six elements of pseudoobscura do in fact correspond to those of
melanogaster. The X of pseudoobscura is V-shaped, and one arm con-
tains the material of the melanogaster X, the other that of the left limb of
the melanogaster III. The remaining four elements of melanogaster are
all intact but are separate. That is, both Vs of melanogaster have their
two arms separated, and one of these (III L) is now the right arm of the
X.
This type of comparison has been extended to several other species
of Drosophila that have haploid chromosome numbers from 3 to 6 (sum-
mary and analysis by Sturtevant and Novitski, 1941). It appears that the
6 elements have retained their composition, with relatively few excep-
tions, but that within each element the sequences of loci are little more
alike than would result from chance alone. In other words, inversions
within an element have been frequent; translocations, or inversions in-
cluding centromeres, have been rare, in the evolution of the genus, as
they are within existing species. There are a few examples of persistent
associations of closely linked genes, which may be due to chance or to
the existence of favorable position effects, but such persistent sections
are not common.
The studies just discussed lead to the conclusion that there is a long-
time stability in the genetic basis of particular characters, but such a sta-
bility has often been questioned. Perhaps the most extreme formulation
of this point of view is that by Harland (1936):
The genes, as a manifestation of which the character develops,
must be continually changing . . . we are able to see how organs such
as the eye, which are common to all vertebrate animals, preserve their
essential similarity in structure or function, though the genes respon-
sible for the organ must have become wholly altered during the evo-
lutionary process, since there is now no reason to suppose that
homologous organs have anything genetically in common.
This conclusion was based on solid facts derived from extensive spe-
cies crosses in the genus Gossypium (cotton). I have elsewhere (Stur-
tevant, 1948) given my reasons for an alternative interpretation of these
facts, based on the polyploid nature of cotton. The more recent compara-
tive biochemical data also favor the idea of the great stability of genetic
systems, since they show essential identity of some of the gene-con-
trolled basic biochemical pathways in bacteria, fungi, and vertebrates.
It is true, however, that in many, probably most, loci there exist se-
ries of isoalleles (Stern and Schaeffer, 1943), which carry on the function
116 A HI STORY OF GENETICS
characteristic of the locus, but with different efficiencies and different
temperature characteristics and reactions to the presence of gene differ-
ences in other loci. These often lead to no phenotypic differences under
normal conditions, and can only be studied by special methods. It seems
probable that the efficiencies of those concerned with any given devel-
opmental system need to be properly adjusted among themselves to give
a harmonious system, as postulated by Goldschmidt for strong and
weak races of Lymantria (Chapter 13). But other equally effective
systems are also possible, and may come to exist in related species. Dis-
harmonies will then arise on species crossing, and there are examples
that suggest that this is often the case (Sturtevant, 1948).