84 Life and Letters of Francis Gatton
rightly interpreted lead to no "universal regression," still less to an argument that "variations proper" cannot be the subject of selection and the formation of new breeds.
This does not prove that "variations proper" have been the basis of evolution, but it removes Galton's chief reason for belief in evolution by discontinuity, that is by sports or mutations. The law of "universal regression "-over which Galton undoubtedly stumbled-is only true when we neglect ancestry beyond the parents and suppose mating at random, but these are not the conditions which exist when intense selection is taking place and the selected interbreed.
Having prefaced Galton's views on Discontinuity with some criticism, which I think is needful, of his theory of regression, we may turn to his paper on "Discontinuity in Evolution," which was published in Mind, Vol. III, N.S. pp. 362-372, July, 1894. Galton begins by saying that students of the laws of variation need not be disheartened by the impossibility of learning what is the cause of variation. Galton, who, as we have seen, believed in individuality in the numerous germ cells of an organism, and that germ cells were subject to selection, found no difficulty- in attributing variation to the effect of interacting germinal elements*. He considered that the actual cause of any particular variation might be put on one side by those who study the degree and character of variation generally.
We are next provided with a definition of race based upon the idea of a typical centre of regression. As I understand him A and B are two different races, if the offspring of the members of A and the offspring of the members of B regress to two different centres of regression. But how can we practically demonstrate this ? If we take the offspring of a pair of individuals of race A, the degree in which they differ from their parental mean will depend upon the long line of ancestry of those parents (to adopt Galton's own views) ; if the parents were relatively small in stature, say, for their ancestry, the offspring average may exceed the parental stature; if they were relatively tall, the offspring average may fall short of the parental. If we choose such a large number of parents of given statures, that we may assume the ancestors of the parents have for average value that of the general population, then the offspring average will regress to the population mean, and should we know the regression coefficient accurately, this will provide the population mean or " typical centre of regression." Similarly we might determine the typical centre of the race B, and ascertain whether the two centres were or were not significantly different. But I cannot see that this is any more than inquiring whether the populations A and B have different
* I must confess to feeling it extremely difficult to accept the view that the population of germ cells belonging to an individual organism are like atoms, identical in character, and have a germinal capacity defined by absolutely the same formula. Such a population of germ cells is, if parasitical, still an organic population, and one continually in a state of reproduction and change. No other organic population that we know of is without variation among its members, and I find it extraordinarily hard to believe that it is a matter of complete indifference which individual spermatogonium of an organism is the ultimate source for fertilisation of an individual ovum of a second organism.