An actual proof of the theory of sexual selection is out of the question, if only because we cannot tell when a variation attains to selection-value. It is certain that a delicate sense of smell is of value to the male moth in his search for the female, but whether the possession of one additional olfactory hair, or of ten, or of twenty additional hairs leads to the success of its possessor we are unable to tell. And we are groping even more in the dark when we discuss the excitement caused in the female by agreeable perfumes, or by striking and beautiful colours. That these do make an impression is beyond doubt; but we can only assume that slight intensifications of them give any advantage, and we _must_ assume this _since otherwise secondary sexual characters remain inexplicable_. The same thing is true in regard to natural selection. It is not possible to bring forward any actual proof of the selection-value of the initial stages, and the stages in the increase of variations, as has been already shown. But the selection-value of a finished adaptation can in many cases be statistically determined. Cesnola and Poulton have made valuable experiments in this direction. The former attached forty-five individuals of the green, and sixty-five of the brown variety of the praying mantis (_Mantis religiosa_), by a silk thread to plants, and watched them for seven days. The insects which were on a surface of a colour Similar to their own remained uneaten, while twenty-five green insects on brown parts of plants had all disappeared in eleven days. The experiments of Poulton and Sanders[45] were made with 600 pupae of _Vanessa urticae_, the "tortoise-shell butterfly." The pupae were artificially attached to nettles, tree-trunks, fences, walls, and to the ground, some at Oxford, some at St. Helens in the Isle of Wight. In the course of a month 93% of the pupae at Oxford were killed, chiefly by small birds, while at St. Helens 68% perished. The experiments showed very clearly that the colour and character of the surface on which the pupa rests--and thus its own conspicuousness--are of the greatest importance. At Oxford only the four pupae which were fastened to nettles emerged; all the rest--on bark, stones and the like--perished. At St. Helens the elimination was as follows: on fences where the pupae were conspicuous, 92%; on bark, 66%; on walls, 54%; and among nettles, 57%. These interesting experiments confirm our views as to protective coloration, and show further, _that the ratio of elimination in the species is a very high one, and that therefore selection must be very keen_.