ferent seeds are described as examples of
individual variability, but those which may be observed on the same
plant, or on cuttings, bulbs or roots derived from one individual
are referred to as cases of partial variability. Partial variability,
therefore, determines the differences among the flowers, fruits, leaves
or branches of one individual: in the main, it follows the same laws
as individual variability, but the position of a branch on a plant also
determines its strength, and the part it may take in the nourishment of
the whole. Composite flowers and umbels therefore have, as a rule,
fewer rays on weak branches than on the strong main ones. The number of
carpels in the fruits of poppies becomes very small on the weak lateral
branches, which are produced towards the autumn, as well as on crowded,
and therefore on weakened individuals. Double flowers follow the same
rule, and numerous other instances could easily be adduced.

Mutating variability occurs along three main lines. Either a character
may disappear, or, as we now say, become latent; or a latent character
may reappear, reproducing thereby a character which was once prominent
in more or less remote ancestors. The third and most interesting case
is that of the production of quite new characters which never existed in
the ancestors. Upon this progressive mutability the main development of
the animal and vegetable kingdom evidently depends. In contrast to this,
the two other cases are called retrogressive and degressive mutability.
In nature retrogressive mutability plays a large part; in agriculture
and in horticulture it gives rise to numerous varieties, which have
in the past been preserved, either on account of their usefulness or
beauty, or simply as fancy-types. In fact the possession of numbers
of varieties may be considered as the main character of domesticated
animals and cultivated plants.

In the case of retrogressive and degressive mutability the internal
cause is at once apparent, for it is this which causes the disappearance
or reappearance of some character. With progressive mutations the case
is not so simple, since the new character must first be produced and
then displayed. These two processes are theoretically different, but
they may occur together or after long intervals. The production of the
new character I call premutation, and the displaying mutation. Both of
course must have their external as well as their internal causes, as
I have