Atavism by Seeds and Buds. 105 
But before I proceed to examine these two forms of 
atavism more closely, I think it desirable to state that I 
here use the word "atavism" in its narrower sense, for 
in its wider sense it embraces so large a group of phe- 
nomena that it would not be possible to deal with them 
all within the limits at my disposal. It seems worth while 
to indicate the more important of these types because they 
are often confused with one another and because results 
obtained with one form are often taken to apply to an- 
other, simply because they both go by the same name. 
We must first of all draw a sharp distinction between 
atavism as applied to variability and as applied to muta- 
bility. In the first case we are dealing with the phenom- 
ena presented by a single heritable character ; in the latter, 
with the conflict of two or more. In the improvement 
of races the offspring do not resemble the selected pa- 
rents, they always revert partly towards the mean of 
their ancestors. We are of course dealing in such cases 
with the phenomenon of regression which was fully dis- 
cussed in the first volume (pp. 82 and 120) ; and it 
would be better to refer to all those individuals of less 
value which are eliminated in selection as regressive and 
those which exceed the level attained by their parents as 
progressive. But it is customary to call the former atav- 
ists ; and, as a matter of fact, they exhibit the degree of 
development of the characters in question as it was mani- 
fested by their grandparents and more remote ancestors, 
and not as in their parents. They could perhaps be 
called "curve atavists," since this term does not suggest 
a reversion beyond the curves in question. 1 
The most fascinating section of the subject of atav- 
ism is that which deals with so-called "youth" forms 
1 See the pedigree of the many-rowed maize, Vol. I, p. 73, Fig. 18. 
