6 Life and Letters of Francis Galton
regression and homoscedasticity where it is quite inappropriate. It is interesting to note how the history of the spread of knowledge follows with halting steps the history of its discovery.
Again, if the reader anticipates that Galton was a faultless genius, who solved his problems straightaway without slip or doubtful procedure, he is bound to be disappointed. Some few creative minds may have done that, or appear to have done it, because, the building erected, they left no signs of the scaffolding; but the majority of able men stumble and grope in the twilight like their smaller brethren, only they have the persistency and insight which carries them on to the dawn.
B. The First Idea of "Regression." I think these conceptions will be well illustrated if we consider Galton's first paper dealing with the subject of regression, namely the lecture entitled : Typical Laws of Heredity, which he gave on February 9, 1877 at the Royal Institution. It is the next forward step he took after the memoir of 1875, in which he had propounded for the first time the continuity of the germ-plasm. See our Vol. ii, pp. 184-8. The paper itself embraces three fundamental sections, which I will take in logical sequence if not that of the paper itself.
First : an account of the experimental data on sweet-peas. Galton assumes here that sweet-peas are invariably self-fertilised, a result which from my own observation I consider only partially true. There is also a further difficulty here : he does not take the average seed of the mother plant as representing the maternal character. He takes seeds of equal weight which may have been the ordinary produce of large-seeded plants, or the exceptional produce of small-seeded plants, and treats these as representing the parental character. This very fact would in itself involve regression in the offspring seeds, and leaves unsettled two important questions (i) whether in the average result from all the seeds of a self-fertilising plant, there would be any regression at all, and (ii) whether there is any difference in the average seed weights of daughter plants grown from light and heavy seeds of the mother plant? Had Galton had these points in mind, he might have thrown light on controversies of a much later date. Again, does the size of the mother seed influence the daughter seed only by way of heredity? Galton's small seeds led to sickly and often sterile plants, and it is quite probable that this might affect the weight of their seeds (see our Vol. ii, p. 181). Be this as it may, Galton found from his data* that there was a linear regression of daughter seed on maternal seed. He does not yet use the term "regression," but speaks of a "reverting" towards "what may be roughly and perhaps fairly described as the average ancestral type." But it is difficult to believe that this reversion was solely due to heredity ; if the original seed had fully represented the maternal plant and that plant had been indefinitely self-fertilised, the Law of Ancestral. Heredity would suggest no regression at
* He issued packets of seven sizes of seeds, each containing ten seeds, and nine friends grew the plants. Two crops failing, he had all the seed offspring of 7 x 7 x 10 = 490 carefully weighed seeds.