a change of structure which has thus been
indirectly produced by correlation of growth might happen to prove
useful for some purpose or another; and in as many cases as such
indirectly produced structures do prove useful, they will straightway
begin to be improved by the direct action of natural selection. In all
such cases, therefore, we should have an explanation of the _origin_ of
such a structure, which is the only point that we are now considering.

I think, then, that all this effectually disposes of the doctrine of
"prophetic germs." But, before leaving the subject, I should like to
make one further statement of greater generality than any which I have
hitherto advanced. This statement is, that we must remember how large a
stock of meaningless structures are always being produced in the course
of specific transmutations, not only by correlation of growth, which we
have just been considering, but also by the direct action of external
conditions, together with the constant play of all the many and complex
forces internal to organisms themselves. In other words, important as
the principle of correlation undoubtedly is, we must remember that even
this is very far from being the only principle which is concerned in the
origination of structures that may or may not chance to be useful.
Therefore, it is not only natural selection when operating indirectly
through the correlation of growth that is competent to produce new
structures without reference to utility. In all the complex action and
reaction of internal and external forces, new variations are perpetually
arising without any reference to utility, either present or future.
Among all this multitude of promiscuous variations, the chances must be
that some percentage will prove of some service, either from the first
moment of their appearance, or else after they have undergone some
amount of development. Such development prior to utility may be due,
either to correlation of growth, to the structure having previously
performed some other function, as already explained, or else to a
continued operation of the causes which were concerned in the first
appearance of originally useless characters. In a series of chapters
which will be devoted to the whole question of utility in the next
volume, I shall hope to give very good reasons for concluding that
useless characters are not only of highly frequent occurrence, but are
due to a variety of other causes besides correlatio