the third and fourth digits are completely united
by skin; "so that in feather-footed pigeons, not only does the exterior
surface support a row of long feathers, like wing-feathers [which, as
just stated, may in some cases be obviously differentiated into
primaries, secondaries and tertiaries], but the very same digits which
in the wing are completely united by skin become partially united by
skin in the feet; and thus by the law of correlated variation of
homologous parts, we can understand the curious connexion of feathered
legs and membrane between the two outer toes[46]." The illustration is
drawn from the specimen to which I have referred.

[46] _Variation of Plants and Animals_, vol. ii. p. 315.

[Illustration: FIG. 117.--Feather-footed pigeon. Drawn from nature.]

Many similar instances of the same law are to be met with throughout
organic nature; and it is evident that in this principle we find a
conceivable explanation of the origin of such adaptive structures as
could not have been originated by natural selection acting directly upon
themselves: they may have been originated by natural selection
developing other adaptive structures elsewhere in the organism, the
gradual evolution of which has entailed the production of these by
correlation of growth. And, if so, when once started in this way, these
structures, because thus accidentally useful, will now themselves come
under the _direct_ action of natural selection, and so have their
further evolution determined with or without the correlated association
which first led to their inception.

Of course it must be understood that in thus applying the principle of
correlated growth, to explain the origin of adaptive structures where it
is impossible to explain such origin by natural selection having from
the first acted directly upon these structures themselves, Darwinists
do not suppose that in all--or even in most--cases of correlated growth
the correlated structures are of use. On the contrary, it is well known
that structures due to correlated growth are, as a rule, useless. Being
only the by-products of adaptive changes going on elsewhere, in any
given case the chances are against these correlated effects being
themselves of any utilitarian significance; and, therefore, as a matter
of fact, correlated growths appear to be usually meaningless from the
point of view of adaptation. Still, on the doctrine of chances, it is
to be expected that sometimes