The second doctrine of dogmatic Darwinism states that all the given diversities among the organisms that natural selection has to work upon are offered to natural selection by so-called fluctuating variation; that is, by variation as studied by means of statistics. This sort of variation, indeed, is maintained to be indefinite in direction and amount, at least by the most conservative Darwinians; it has occasionally been called a real differential; in any case it is looked upon as being throughout contingent with regard to some unity or totality; which, of course, is not to mean that it has not had a sufficient reason for occurring. It could hardly be said to be beyond the realm of possibility that such differences among organic species as only relate to degree or quantity and perhaps to numerical conditions also, might have been “selected” out of given contingent variations, if but one postulate could be regarded as fulfilled. This postulate may appropriately be stated as the fixation of new averages of variation by inheritance. Let the average value of a variation, with regard to a given property of a given species be *n* and let the value *n* + *m*--*m* being variable--which is represented in fewer individuals of course than is *n*, be such as to offer advantages in the struggle for existence; then the individuals marked by *n* + *m* will have the greater chance of surviving. Our postulate now states that, in order that a permanent increase of the average value of the variation in question may be reached, *n* + *m* in any of its variable forms must be able to become the average value of the second generation, as *n* was the average value of the first. Out of the second generation again it would be the few individuals marked by *n* + *m* + *o*, which would be selected; *n* + *m* + *o* would be the new average; afterwards *n* + *m* + *o* + *p* would be selected, would become the new average, and so on. A black variety for instance might be selected by such a series of processes out of a grey-coloured one without difficulty. But our postulate is not beyond all doubt: certain experiments, at least, which have been carried out about the summation of variations of the true fluctuating type by any kind of selection seem to show that there may be a real progress for a few generations, but that this progress is always followed by a reversion. Of course our experience is by no means complete on this subject, and, indeed, it may be shown in the future that positive transforming effects of fluctuating variability, in connection with selective principles, are possible in the case of new quantitative differences (in the widest sense), but we are not entitled to say so at present. And this is the only condition on which we can give credit to the second doctrine of dogmatic Darwinism. Its second principle, indeed, proves to be absolutely inadequate to explain the origin of any other kind of specific properties whatever.