Recognized HTML document


descent from one generation of a people to the next, showed that the generations would be identical if this kind of Regression was allowed for.

In 1886 I contributed two papers [91, 92] to the Royal Society on Family Likeness, having by that time got my methods for measuring heredity into satisfactory shape. I had given much time and thought to Tables of Correlations, to display the frequency of cases in which the various deviations say in stature, of an adult person, measured along the top, were associated with the various deviations of stature in his mid-parent, measured along the side. (I had long used the convenient word " mid-parent " to express the average of the two parents, after the stature or other character of the mother had been changed into its male equivalent.) But I could not see my way to express the results of the complete table in a single formula. At length, one morning, while waiting at a roadside station near Ramsgate for a train, and poring over the diagram in my notebook, it struck me that the lines of equal frequency ran in concentric ellipses. The cases were too few for certainty, but my eye, being accustomed to such things, satisfied me that I was approaching the solution. More careful drawing strongly corroborated the first impression.

All the formulae of Conic Sections having long since gone out of my head, I went on my return to London to the Royal Institution to read them up. Professor, now Sir James, Dewar, came in, and probably noticing signs of despair in my face, asked me what I was about ; then said, "Why do you bother over

1 See Pres. Address, Section H, Brit. Assoc. Aberdeen, 1885 [87].