Weismann was led to perceive the principle in question by the consideration that new specific characters, when they first appear, do not all appear together in the same individuals: they appear one in one individual, another in another, a third in a third, &c.; and it is only in the course of successive generations that they all become blended in the same individuals by free intercrossing. Hence, the eventually emerging constant or specific type is the resultant of all the transitory or varietal types, when these have been fused together by intercrossing. From which Weismann deduces what he considers a general law--namely, that "the constancy of a specific type does not arise suddenly, but gradually; and it is established by the promiscuous crossing of all individuals[45]." From which again it follows, that this constancy must cease so soon as the condition which maintains it ceases--i. e. so soon as free intercrossing is prevented by the geographical isolation of a portion of the species from its parent stock. [45] _Loc. cit._, p. 43. Now, to begin with, this statement of the principle in question is not a good statement of it. There was no need while stating the doctrine that separation induces differentiation, to found the doctrine on any such highly speculative basis. In point of fact, there is no real evidence that specific types do attain their constancy in the way supposed; nor, for the purposes of the doctrine in question, is it necessary that there should be. For this doctrine does not need to show how the constancy has been _attained_; it only has to show that the constancy is _maintained_ by free intercrossing, with the result that when free intercrossing is _by any means_ prevented, divergence of character ensues. In short, the correct way of stating the principle is that which has been adopted by Delbœuf and Gulick--namely, the average characters of a separated portion of a species are not likely to be the same as those of the whole species; with the result that divergence of type will be set up in the separated portion by intercrossing within that portion. Or the principle may be presented as I presented it under the designation of "Independent Variability"--namely, "a specific type may be regarded as the average mean of all individual variations, any considerable departure from this average mean being, however, checked by intercrossing," with the result that when intercrossing is prevented between a portion of a species and the rest of the species, "this population is permitted to develop an independent history of its own, shielded from intercrossing with its parent form[46]." [46] _Physiological Selection_, pp. 348, 389.