Thus we have here the origin of a new character, arising from small beginnings, at least in _S. tiliae_, in which species the coloured stripes are a normal specific character. In the other species, _S. populi_ and _S. ocellata_, we find the beginnings of the same variation, in one more rarely than in the other, and we can imagine that, in the course of time, in these two species, coloured lines over the oblique stripes will arise. In any case these spots are the elements of variation, out of which coloured lines _may_ be evolved, if they are combined in this direction through the agency of natural selection. In _S. populi_ the spots are often small, but sometimes it seems as though several had united to form large spots. Whether a process of selection in this direction will arise in _S. populi_ and _S. ocellata_, or whether it is now going on cannot be determined, since we cannot tell in advance what biological value the marking might have for these two species. It is conceivable that the spots may have no selection-value as far as these species are concerned, and may therefore disappear again in the course of phylogeny, or, on the other hand, that they may be changed in another direction, for instance towards imitation of the rust-red fungoid patches on poplar and willow leaves. In any case we may regard the smallest spots as the initial stages of variation, the larger as a cumulative summation of these. Therefore either these initial stages must already possess selection-value, or, as I said before: _There must be some other reason for their cumulative summation_. I should like to give one more example, in which we can infer, though we cannot directly observe, the initial stages.