As regards other forms of natural selection, we know that races are kept pure by the much more frequent destruction of those individuals who depart the more widely from the typical centre. But natural selection was shown to be inoperative in respect to individual varieties of patterns, and unable to exercise the slightest check upon their vagaries. Yet, for all that, the loops and other classes of patterns are isolated from one another just as thoroughly and just in the same way as are the genera or species of plants and animals. There is no statistical difference between the form of the law of distribution of individual Loops about their respective typical centres, and that of the law by which, say, the Shrimps described in Mr. Weldon's recent memoirs (_Proc. Roy. Soc._, 1891 and 1892) are distributed about theirs. In both cases the distribution is in quasi-accordance with the theoretical law of Frequency of Error, this form of distribution being entirely caused in the patterns, by _internal_ conditions, and in no way by natural selection in the ordinary sense of that term. It is impossible not to recognise the fact so clearly illustrated by these patterns in the thumbs, that natural selection has no monopoly of influence in the construction of genera, but that it could be wholly dispensed with, the internal conditions acting by themselves being sufficient. When the internal conditions are in harmony with the external ones, as they appear to be in all long-established races, their joint effects will curb individual variability more tightly than either could do by itself. The normal character of the distribution about the typical centre will not be thereby interfered with. The probable divergence (= probable error) of an individual taken at random, will be lessened, and that is all. Not only is it impossible to substantiate a claim for natural selection, that it is the sole agent in forming genera, but it seems, from the experience of artificial selection, that it is scarcely competent to do so by favouring mere _varieties_, in the sense in which I understand the term. My contention is that it acts by favouring small _sports_. Mere varieties from a common typical centre blend freely in the offspring, and the offspring of every race whose _statistical_ characters are constant, necessarily tend, as I have often shown, to regress towards their common typical centre. Sports, on the other hand, do not blend freely; they are fresh typical centres or sub-species, which suddenly arise we do not yet know precisely through what uncommon concurrence of circumstance, and which observations show to be strongly transmissible by inheritance.