76   Life and Letters of Francis Galton

frequency in the case of the offspring of a parent suffering from this disease, then by a series of approximations we can readily obtain the value of the correlation between offspring and parent, or the intensity of heredity in the case of that disease. Galton himself states

"Too much stress must not be laid on this coincidence*, because many important points had to be slurred over, as already explained. Still, the primd facie result is successful, and enables us to say that so far as this evidence goes, the statistical method we have employed in treating consumptivity seems correct, and that the law of heredity found to govern all the different faculties as yet examined, appears to govern that of consumptivity also, although the constants of the formula differ slightly." (p. 185.)

The penultimate chapter of Natural Inheritance is termed Latent Elements. The main point to which Galton appeals here is that the parents contributing on the average a definite amount to the heritage of a child, according to Galton each 4, the residue of the stock of either parent can on the average only contribute a definite amount, i.e. 4 on this view, to the child, or only I of the characters of the ancestry can lie latent in the parent, and be contributed to the child. Galton argues that "either parent must contribute on the average only one quarter of the Latent Elements, the remainder of them dropping out and their breed becoming absolutely extinguished" (p. 188). He illustrates this by the selection of 13 out of a pack of 52 cards; any card may be chosen but actually 39 are rejected, yet if a great many sets of 13 are chosen, i.e. a great many individuals be taken, every card in the original pack will ultimately appear. " No given pair can possibly transmit the whole of their ancestral qualities; on the other hand there is probably no description of ancestor whose qualities have not been in some cases transmitted to a descendant" (p. 189). The throwing out of half the latent (as well as half the patent) elements at each crossing is really part of Galton's idea of all inheritance being particulate, a mosaic of ancestral characters. Even his idea of a blend is not a summation of continuous contributions, but a summation so to speak of quanta from individual ancestors.

In his next paragraph Galton deals with a Pure Breed, and again his error as to regression appears to come to the surface. He discovered regression simply as a statistical result, i.e. because he took parents of given characters, whose earlier ancestry might be anything whatever, he naturally found the offspring nearer than the parents to mediocrity. But unfortunately this idea of regression fixing itself in his mind became for him a biological fact, and he considered that he had discovered in stability of types, i.e. in groups each with their own focus or centre of regression, the source of evolution, or change from one type to a second. He now tells us that in the case of pure breed in which there has been long selection, the influence of a large quantity of mediocre ancestry would disappear, and so would the tendency to regression, except in so far as it is "connected with the stability of different types" (p. 189). In other words we have now, two sources of regression, while Galton's original deduction of regression as purely statistical and depended

* That of the above percentages.