94 Life and Letters of Francis Galton
breeding, that shall afford a secondary focus of regression, and become the dominant one, if the ancestral qualities that interfere with it be eliminated by sustained isolation and selection. Then a new variety would, as I conceive, arise; but into this disputable topic there is no need to enter now." [See, however, my footnote, p. 79.]
We now know that on the theory of multiple regression, this indefinite regression has no existence; there is a slight regression in the first generation of breeding from the selected stock, but it ceases with this generation. We have again in the cited passage evidence that Galton was obliged to appeal to "sports" to account for evolutionary progress, because he had misinterpreted the theory of regression. If w be the regression of the offspring of the first generation of selected midparentage, the regression of the offspring of parents of the first generation, who have also selected midgrandparents, is not to be taken w again. Thus the formulae, the numerical table and the conclusions drawn from it in this paper are I think in error. But the idea at the back of it that the more intense the selection, the more rapid is the relative progress, is true; as also the idea that there may in each case be a limiting value. Probably no such continued selection is really feasible; too many characters in the organism are highly correlated, so that if it were possible to carry under conditions of viability an individual character to a height much above the population mean, some one or other of the correlated characters would be almost certain to be incompatible with the continued efficiency of the organism in relation to its environment or its functions*.
I have not recalculated Galton's table, because with the data at present available, I am inclined to believe that selection for two or three generations and then inbreeding would be followed, at any rate in some characters, by a progression rather than a regression. In other words the strength of inheritance is such that with a very brief period of selection followed by isolation a continuous differentiation will proceed-so far as it is not checked by a counter natural selection. This suggests that we may have to seek in heredity itself for the basis of progressive evolution; a variation maintained for a couple of generations, followed by an isolation of the offspring, will continue to progress. If this be true we surmount the difficulty of why variations to which the environment is not hostile, or indeed may be favourable when they are sufficiently developed, can reach the stage of development at which they become important as a new factor of efficiency in the individual. We see that it is not natural selection, but the mere force of heredity, which leads in isolated groups to the genesis of variations of sufficient importance to have survival value to the individual. We may term this theory of the genesis of remunerative variations the "Heredity Theory of Progressive Evolution." It seems at first sight in flat contradiction to Galton's views on continuous regression when selection ceases, unless the selection has led to the creation of a sport. Yet it really flows
* Nor is this confined only to the functions of the individual, but may concern the functions of other members of its race. Thus breeders of bull-dogs have gone on continuously selecting the size of the head until the mortality of puppies and bitches at littering has, become so serious as to threaten even the survival of the breed.