Analogous illustrations to any amount may be drawn from Mammals. For instance, I have worked through the Marsupials with the aid of Mr. Oldfield Thomas' diagnostic description of their numerous species. Now, let us take any one of the genera, such as the kangaroos. This comprises 23 species living on an island continent of high antiquity, and not exposed to the depredations of any existing carnivorous enemies; so that there is here no present need to vary colour for purposes of protection. Moreover, in all cases the diagnostic distinctions of colour are so exceedingly trivial, that even if large carnivora were recently abundant in Australia, no one could reasonably suggest that the differences in question would then have been protective. On an average, each of the 23 species presents rather more than 20 peculiarities of shading, which are quoted as specifically diagnostic. Altogether there are 474 of these peculiarities distributed pretty evenly among the 23 species; and in no case can I conceive that utility can be suggested. * * * * * Hitherto we have been considering the question of fact, as to whether "a large proportional number of specific characters" do or do not admit of having their utility demonstrated, or even so much as plausibly suggested. In the result, I can only conclude that this question of fact is really not an open one, seeing that it admits of an abundantly conclusive answer by any naturalist who will take the trouble to work through the species of any considerable number of genera in the way above indicated. But although the question of fact is thus really closed, there remains a more ultimate question as to its theoretical interpretation. For, as already pointed out, no matter how great an accumulation of such facts may be collected, our opponents are always able to brush them aside by their _a priori_ appeal to the argument from ignorance. In effect they say--We do not care for any number of thousands of such facts; it makes no difference to us what "proportional number" of specific characters fail to show evidence of utility; you are merely beating the air by adducing them, for we are already persuaded, on antecedent grounds, that _all_ specific characters _must_ be either themselves useful, or correlated with others that are, whether or not we can perceive the utility, or suggest the correlation.