first place w ith a view to the problem of heredity in the direct line, and with no thought of th eir throw ing any lig h t on th e problem of telegony. That steady telegonic influence m ight be deduced from such family data has only recently occurred to me, and I should now hesitate to publish any conclusions on this subject, based on somew hat mixed and sparse returns, did I not consider th a t it may be a long time before more extensive returns are available, and th a t the publication of this m ethod of dealing with telegony m ay induce others to undertake the collection of a wider range of material.
My own 800 family data cards did not provide a sufficiently large num ber of either brother-brother or sister-sister couples to give a strong hope of a difference between the correlation coefficients sufficiently large as com pared with its probable error to base any legitim ate conclusion upon. I, therefore, again borrowed from Mr. Galton his 200 family data returns, and from these 1,000 families was able to select 385 brother-brother pairs and 450 sister-sister pairs. In these statistics each individual is only included in one pair, and the difference in age between the elder and younger members of each pair differs very widely from pair to pair. In some cases there may be several years between the ages and several intervening ch ild re n ; in others the members of the pair m ay be successive children following each other in successive years. In each case all we can say is, th at if there be a steady telegonic influence, the relation of the elder m em ber to the parent will weigh down the same scale, and in the final result we ought to find a distinctly greater or less correlation, as the case may be. I think a more serious objection to the data thau the variation in the num ber of years between fraternal pairs is the m ixture I have made of data collected at different periods and in somewhat different manners. My own data are drawn, I think, from a w ider class of the com m unity than Mr. Galton’s. They are not exclusive of his class, but, I think, cover his class, and go somewhat further down in the social scale. They suffice to show that the means and variations change considerably from one social stratum to another, and what is still more im portant th a t the Galton-1 unctions or coefficients of correlation for heredity are far from being constant even within the same race, as we pass from one i ank of life to a second. Thus, my means for stature in the case of both fathers and mothers are upwards of ± in. less than Mr. Galton’s, ut my means agree fairly well w ith his results in the case of both sons and daughters. There are also good agreements and somewhat puzzling disagreements not only in the variations, but, above all, in the coefficients of correlation for heredity. I reserve for the present t le full discussion of my heredity data, b u t I wish it to be quite understood th at my conclusions in this paper are based, not upon the best Possible data, e.g., measurements made on one class of the com
