This last case shows that it is not merely a lesser power of growth that accounts for the difference between the regenerated part and the original, for here more is regenerated than was previously present. There remains nothing for it but the assumption that the regeneration-determinants have remained at a lower phyletic level, while the determinants which direct embryogenesis have varied, and either developed further or retrogressed. It is easy to understand that the regeneration-rudiment must vary phyletically much more slowly than the parts which evolved in the ordinary way and much more slowly than the determinants of these parts, for natural selection means a selection of the fittest, and the speed with which the establishment of a variation is attained depends, _ceteris paribus_, on the number of individuals that are exposed to selection with respect to the varying part. If in a species of a million living at the same time nine-tenths perish by accident, there will remain only 100,000 from which to select the 1,000 which we will assume constitute the normal number of the species. The more of these 100,000 which possess the useful variation the higher will be the percentage of the normally surviving 1,000 possessing it, and the more rapidly will the useful variation increase. But when it is a question of the variation of the regeneration-primordium, the selection will take place not among all the 100,000 individuals which chance has spared, but only among those of them which have lost a limb by accident, and thus are in a position to regenerate it more or less completely. If we assume that this takes place in 10 per cent. of cases, then selection for the improvement of the regeneration-apparatus will only take place among 1,000 individuals, and thus the process of modification of the regeneration-primordium must go on very much more slowly than that of the limb itself.