Natural selection does not, of course, induce or evoke variations; these are given to its activity, and they are the material on which it operates. What, then, is the nature of the deviations from the specific types of morphology that are selected or eliminated? Not those induced by the environment, and transmitted in their nature and direction to the progeny of the organisms first displaying them. It is not unproved that such variations do occur, and it is even probable that they do occur. But we may conclude that the frequency of their occurrence is not great enough to afford sufficient material for natural selection. It is also clear that the ordinarily occurring variations that we observe in any large group of organisms collected at random are not alone the material for selection; for we have seen that experimental breeding from such variations does not lead to the establishment of a stable race or “variety.” Nevertheless some effect is produced, and this may be accounted for by supposing that the observed variations are really of two kinds--fluctuating variations, which are not inherited, and mutations, which are inherited. The small observed effect is due to the selection of the mutations alone: it is a real effect of selection, an undoubted transmutation of the specific form, but experimental and statistical investigations seem to show that selection from the variations that we usually observe is too slow a process to account for the existing forms of life. Natural selection acts, therefore, on mutations. Now it seems that we are forced to recognise the existence of two categories of mutations, (1) those stable modifications of an “unit-character” which we term “Mendelian characters,” and (2) those groups of stable modifications to which de Vries applied the term mutations. It seems at first difficult to see how permanent modifications of the specific form can be brought about by the transmission of Mendelian characters, for these characters are always transmitted in pairs. Let us take a concrete case--that of a man who has six fingers on his right hand, and let us suppose that this was a real, spontaneously appearing character or mutation which had not previously occurred in the ancestry of the man. Two contrasting characters would then be transmitted, (1) the normal five-fingered hand, and (2) the six-fingered hand. Both of these characters are supposed to be present at the same time in the organisation of the men and women of the family originating in this individual, but one of them is always latent or recessive. There would, however, be individuals in which only one of the characters would be present--either the normal or abnormal number of digits, but intermarriage with individuals belonging to the other pure strain would immediately lead again to the transmission of the contrasting characters, or allelomorphs, although marriage with an individual belonging to the same pure strain