on).]

The essential difference between true variation and fluctuation or
variability of an extreme nature, is with reference to the inheritance
of such divergence. In the second generation the offspring of extreme
variates or fluctuations have not the same average as their own
parents but an average much nearer that of the whole group to which
their parents belonged; the average stature of the children of
unusually short or tall parents is respectively greater or less than
that of their own parents--that is, is nearer the average of the whole
group of parents, provided the shortness or tallness of the parents is
a fluctuation. When the shortness or tallness is a true variation or
mutational character, offspring have approximately the same average
stature as their immediate parents, although the children of course
show fluctuation in height so that some are slightly above and others
slightly below the parental height.

Mutations may occur through the addition or the subtraction of single
characters of the simple or unit type. Such are the variations from
brown or blue eyes to albino, five fingers to six, and the like. These
are the familiar "sports" of the horticulturalist and breeder. They
are of the greatest value in evolution, for it seems quite likely that
it is only through the permanent racial fixation of these mutations
that permanent changes in the characters of a breed may be effected,
i. e., evolution occurs primarily through mutation.

In connection with the general subject of variation we should mention
briefly certain aspects of the recent work of Johannsen and Jennings,
showing that many organic specific groups or "species," whose
characters, when measured accurately give what is called a normal
variability curve similar to that of stature illustrated in Fig. 3,
are not really homogeneous groups of fluctuating individuals as the
curves would indicate superficially, but that each gross group or
species is actually composed of a blend of a number of smaller groups,
each with its own average and fluctuating variability. It is only when
these are taken all together as a lump that they fuse into a single
and apparently simple curve.

For example, the curve shown in Fig. 6, A, which is approximately that
of a normal distribution, in some cases might be shown by
experimentation to consist in reality of several truly distinct
elements, say three for purposes of illustration, as shown in Fig. 6,
B. Each of the