. 82). Another factor "whiting" which
produces no effect on red makes eosin entirely white. Since cream or
whiting may be carried by red eyed flies without their presence being seen
until eosin is used, the experimenter must be continually on the lookout
for such factors which may lead to erroneous conclusions unless detected.
As yet breeders have not realized the important role that modifiers have
played in their results, but there are indications at least that the
heaping up of modifying factors has been one of the ways in which highly
specialized domesticated animals have been produced. Selection has
accomplished this result not by changing factors, but by picking up
modifying factors. The demonstration of the presence of these factors has
already been made in some cases. Their study promises to be one of the most
instructive fields for further work bearing on the selection hypothesis.

In addition to these well recognized methods by which artificial selection
has produced new things we come now to a question that is the very crux of
the selection theory today. Our whole conception of selection turns on the
answer that we give to this matter and if I appear insistent and go into
some detail it is because I think that the matter is worth very careful
consideration.

ARE FACTORS CHANGED THROUGH SELECTION?

As we have seen, the variation that we find from individual to individual
is due in part to the environment; this can generally be demonstrated.
Other differences in an ordinary population are recognized as due to
different genetic (hereditary) combinations. No one will dispute this
statement. But is all the variability accounted for in these two ways? May
not a factor itself fluctuate? Is it not _a priori_ probable that factors
do fluctuate? Why, in a word, should we regard factors as inviolate when we
see that everything else in organisms is more or less in amount? I do not
know of any _a priori_ reason why a factor may not fluctuate, unless it is,
as I like to think, a chemical molecule. We are, however, dealing here not
with generalities but with evidence, and there are three known methods by
means of which it has been shown that variability, other than environmental
or recombinational, is not due to variability in a factor, nor to various
"potencies" possessed by the same factors.

(1) By making the stock uniform for all of its factors--chief factors and
modifiers alike. Any change in such a stock produced by