Evolution of Hawaiian Animals — Gosline 
is on the adjacent Cornish mainland. On the 
small islands of the group, however, the num- 
ber of spots on the wing of the females not 
only varies from island to island, but increases 
on some and decreases on others. Second, Mer- 
tens (1934:116) has pointed out that the same 
island may contain a dwarf form of one reptile 
and a giant form of another. It would be dif- 
ficult to postulate environmental factors that 
would select animals in these ways. 
If small-population forces are to be postu- 
lated for such differences, three possibilities 
must be considered. The first is the random loss 
of genes which may occur in small populations 
by genetic drift (Wright, 1931, etc.). Such a 
factor would presumably be operative in all 
small populations. A special case of genetic 
drift is the phenomenon often called founder 
effect. This merely expresses the fact that the 
original immigrants to an island are frequently 
few in number, and, whether or not they con- 
stitute a representative sample, they can bring 
with them only a small proportion of the alleles 
present in the parental population (see, for 
example, Zimmerman, 1948:122, 123). The 
third possible small-population factor is what 
Mayr (1954) calls internal selection. In large 
populations where each gene often has many 
alleles those which work best as heterozygotes 
will tend to be selected; on the other hand, in 
small populations there will be a larger propor- 
tion of homozygotes, and alleles which work 
best in the homozygotic condition will tend to 
be selected. Thus some shift in internal selec- 
tion pressure between large and small popula- 
tions would be expected. 
These small-population factors, acting per 
se, should affect insular immigrants during 
those initial stages when the population is still 
small (Fig. 1). But there appears to be no 
known instance in which a change at this stage 
has been recorded (cf. Mayr, 1954). Further- 
more, it is a generally accepted dictum that, 
other things being equal, the older the island 
the greater will be the differentation in its 
biota; this implies continuous, not just initial, 
change. 
Then how is the differentation that occurs on 
islands, and more rapidly on small islands, to 
be explained? King (1955) conducted selec- 
tion experiments for DDT resistance on two 
269 
cultures of Drosophila melanogaster. After a 
dozen or more generations some degree of 
DDT resistance began to be built up in both 
lines. But, as judged from crossbreeding ex- 
periments, the resistance had been built up 
differently in the two lines. King (1955:314) 
states: "The manner in which a line could re- 
spond to selection was to some extent deter- 
mined by the genetic nature of the sample from 
which it started, and having started along one 
certain road, it kept on. The inevitable sam- 
pling error which occurs when a line is taken 
from a larger population is very likely the an- 
lage of the genetic individuality of the line. 
This is, of course, an example of the principle 
of genetic drift. . . .” The second example is 
that reported on by Dobzhansky and Pavlov- 
sky (1957). In this experiment ten cultures 
from a specially developed laboratory stock of 
Drosophila pseudo-obscura were started with 
20 flies each and compared after 17 months 
with ten other cultures that did not begin with 
a reduced number of individuals. Those stocks 
which had started with 20 flies showed more 
variation than the controls. Again, Dobzhansky 
and Pavlovsky conclude (1957:316): "Al- 
though the trait studied (gene arrangement in 
the third chromosome) is subject to powerful 
selection pressure, the outcome of the selection 
in the experimental populations is conditioned 
by random genetic drift." 
One aspect of these experiments by King 
and by Dobzhansky and Pavlovsky may well be 
of importance for insular evolution. In both 
instances not only were the original samples 
small, but the selection that was exerted upon 
them was far different from the selection of the 
natural environment from which the flies came. 
It is as though the samples in the experiments 
were subjected to an intense selection pressure 
at right angles to the pressures to which the 
ancestral "wild" forms had presumably adapted 
themselves. Possibly some of the alleles in- 
tensely selected under the laboratory conditions 
had been of only peripheral significance to the 
wild stocks and hence variably represented in 
them. Such alleles would be more subject to 
sampling error among small founder popula- 
tions drawn from the parental stock than those 
previously under intense positive selection pres- 
sure. 
