If, however, the coloured edges certainly depend on natural selection, this agency causing the scattered spots to coalesce and become linear, we have here the proof that such spots first arose in a precisely similar manner in several species, quite independently of one another--that, in fact, a “fixed direction of variation” in a certain sense exists. In three species of _Smerinthus_-larvæ, red spots appear towards the end of the ontogeny; in _S. Populi_ and _Ocellatus_ in only a minority of individuals, and always separate (not coalescent), and in _S. Tiliæ_ in a majority of specimens, the spots frequently becoming fused into one large, single, longish marking. These three species cannot have inherited the spots from a common ancestor, since they are absent in the younger ontogenetic stages, or occur only exceptionally, becoming larger and more numerous in the last stage; they obviously form a character which must be considered as a case of “anticipated development.” How is it then that three species vary independently of each other in an analogous manner? I know of no other answer to this question than that similar variations must necessarily arise from similar physical constitutions--or, otherwise expressed, the three species have inherited from an unknown parent species, devoid of spots, not this last character itself, but a physical constitution, having a tendency to the formation of red spots on the skin.[157] The case offers many analogies to that of the colour varieties of _Lacerta Muralis_, to which Eimer[158] briefly calls attention in his interesting communications on the blue lizard of the Faraglioni Rocks at Capri. The South Italian lizards, although having differently formed skulls, show the same brilliantly coloured varieties as those of North Italy; and Eimer believes that these parallel variations in widely separated localities, some of which have long been isolated, must be referred to a tendency towards fixed directions of variation innate in the constitution of the species. I long ago insisted[159] that it should not be forgotten that natural selection is, in the first place, dependent upon the variations which an organism offers to this agency, and that, although the number of possible variations may be very great for each species, yet this number is by no means to be considered as literally infinite. For every species there may be _impossible_ variations. For this reason I am of opinion that the physical nature of each species is of no less importance in the production of new characters than natural selection, which must always, in the first place, operate upon the results of this physical nature, _i.e._ upon the variations presented, and can thus call new ones into existence.