Examples of Constant Races. 77 
viously made under conditions which excluded the effects 
of crossing. 
In spite of the existence of these experimental data, 
it is still the general view that varieties are inconstant 
forms. That which distinguishes them from true spe- 
cies is supposed to be their faculty of giving rise to 
occasional and not even rare reversions to the type of 
the species. This, moreover, is supposed to be a proof 
of their assumed relation to their species. 
Every one of course is free to choose his own defi- 
nition of a variety. But he who makes inconstancy an 
essential part of the definition will have to exclude a very 
large number and perhaps the most important of our 
horticultural varieties, and regard them as elementary 
species. 
I have spent much time in the endeavor to test the 
constancy of horticultural and also of wild varieties with 
a view partly of directly satisfying myself as to their 
purity and partly of finding inconstant forms for sub- 
sequent experiments. I have usually started from seed 
but sometimes, in the case of perennial varieties, from 
bought plants. Whenever possible the visits of insects 
were excluded and the plants artificially pollinated. But 
in the great majority of cases pollination has to be left 
to bumblebees and moths, and we must be content in 
providing complete isolation. 
The most important point is the extent of the experi- 
ment. Absolute constancy can obviously never be di- 
rectly demonstrated. The space and time needed for 
other experiments seldom permit the bringing to flower 
of more than a few thousands of plants of one sort. 
And even if this is continued for several years the possi- 
bility of the occurrence of rare cases of atavism (e. g., 
