or to selective death-rate, or to secular evolution—diminishes with age.
(4) The following are the coefficients of correlation (r) and the coefficients of regression (R) for parents and sons:
| r. | R. | |
| Father and elder sons | 0·4120 ± 0·0264 | 0·4281 |
| Father and younger sons | 0·4170 ± 0·0262 | 0·4427 |
| Mother and elder sons | 0·4094 ± 0·0265 | 0·4374 |
| Mother and younger sons | 0·4111 ± 0·0264 | 0·4488 |
If we measure, as seems reasonable, the hereditary influence of parentage by the magnitude of the coefficient of correlation between parent and offspring, then several important conclusions may be drawn from this table.
(i) There is no sensible difference between the influences of the father on younger and on elder sons, and no sensible difference between the influences of the mother on younger and on elder sons.
If we pay attention to such slight differences as exist, there would appear, not to be an increase of paternal and a decrease of maternal influence on younger children, but an extremely slight increase of both. In other words, so far as stature in sons is concerned, judged by correlation: No steady telegonic influence exists.
(ii) There is a very slight prepotency of the father over the mother in the case of both younger and elder sons; a prepotency which will be slightly magnified when account is taken of the absolute stature of the two parents.
But the great prepotency of paternal inheritance noticed in the former memoir is not confirmed. The co-efficients of maternal inheritance have been increased by more than 30 per cent. (from 0·293 to 0·410), while those of paternal inheritance (0·396 as compared with 0·414) have remained almost stationary. This result seems to show the want of constancy of the Galton's functions for heredity within the same race. An explanation on the ground that the present statistics embrace a wider range of the community than the earlier, and possibly a more closely correlated class,[1] fails, at any rate in part, owing to the sensible constancy of the paternal correlation. The main difference of course between the present and the former statistics is the exclusion of the influence of reproductive
- ↑ I have pointed out (loc. cit., p. 284) that working and lower middle class families appear to be more closely correlated than those of the upper middle class.
