ceives further support from
the fact that, in many such trees, the leaves are drawn out into a
beak-like prolongation (Stahl and Haberlandt) which facilitates the
rapid falling off of the rain water, and also from the fact that the
leaves, while they are still young, hang limply down in bunches which
offer the least possible resistance to the rain. Thus there are here
three adaptations which can only be interpreted as due to selection.
The initial stages of these adaptations must undoubtedly have had
selection-value.

But even in regard to this case we are reasoning in a circle, not
giving "proofs," and no one who does not wish to believe in the
selection-value of the initial stages can be forced to do so. Among
the many pieces of presumptive evidence a particularly weighty one
seems to me to be _the smallness of the steps of progress_ which we
can observe in certain cases, as for instance in leaf-imitation among
butterflies, and in mimicry generally. The resemblance to a leaf, for
instance of a particular Kallima, seems to us so close as to be
deceptive, and yet we find in another individual, or it may be in many
others, a spot added which increases the resemblance, and which could
not have become fixed unless the increased deceptiveness so produced
had frequently led to the overlooking of its much persecuted
possessor. But if we take the selection-value of the initial stages
for granted, we are confronted with the further question which I
myself formulated many years ago: How does it happen _that the
necessary beginnings of a useful variation are always present_? How
could insects which live upon or among green leaves become all green,
while those that live on bark become brown? How have the desert
animals become yellow and the Arctic animals white? Why were the
necessary variations always present? How could the green locust lay
brown eggs, or the privet caterpillar develop white and lilac-coloured
lines on its green skin?

It is of no use answering to this that the question is wrongly
formulated[37] and that it is the converse that is true; that the
process of selection takes place in accordance with the variations
that present themselves. This proposition is undeniably true, but so
also is another, which apparently negatives it: the variation required
has in the majority of cases actually presented itself. Selection
cannot solve this contradiction; it does not call forth the useful
variation, but simply works upo