with small differences rather than with
the larger and more striking ones that the breeder is mainly concerned. It
does not matter much to him whether the colour of a pea flower is purple or
pink or white. But it does matter whether the plant bears rather larger
seeds than usual, or rather more of them. Even a small difference when
multiplied by the {160} size of the crop will effect a considerable
difference in the profit. It is the general experience of seedsmen and
others that differences of this nature are often capable of being developed
up to a certain point by a process of careful selection each generation. At
first sight this appears to be something very like the gradual accumulation
of minute variations through the continuous application of a selective
process. Some recent experiments by Professor Johannsen of Copenhagen set
the matter in a different light. One of his investigations deals with the
inheritance of the weight of beans, but as an account of these experiments
would involve us in the consideration of a large amount of detail we may
take a simple imaginary case to illustrate the nature of the conclusions at
which he arrived. If we weigh a number of seeds collected from a patch of
plants such as Johannsen's beans we should find that they varied
considerably in size. The majority would probably not diverge very greatly
from the general average, and as we approached the high or low extreme we
should find a constantly decreasing number of individuals with these
weights. Let us suppose that the weight of our seed varied between 4 and 20
grains, that the greatest number of seeds were of the mean weight, viz. 12
grains, and that as we passed to either extreme at 4 and 20 the number
became regularly less. The weight relation of such a collection of seeds
can be expressed by the accompanying curve (Fig. 30). Now if we select for
{161} sowing only that seed which weighs over 12 grains, we shall find that
in the next generation the average weight of the seed is raised and the
curve becomes somewhat shifted to the right as in the dotted line of Fig.
30. By continually selecting we can shift our curve a little more to the
right, _i.e._ we can increase the average weight of the seeds until at last
we come to a limit beyond which further selection has no effect. This
phenomenon has been long known, and it was customary to regard these
variations as of a continuous nature, _i.e._ as all chance fluctuations in
a homogene