298 



Miss M. Beeton and Prof. Karl Pearson. 



times the magnitudes of their probable errors, and they agree within the 

 probable error of their differences. The only significant difference 

 is the mean age of deaths of sons in the Landed Gentry, which is some 

 two years higher than in the Peerage. This is the more noteworthy 

 in that we have begun our peerage record at 25 and not 20. Clearly 

 the sons of the Landed Gentry are longer lived. We have undoubtedly 

 correlation, say somewhere about 0*12, sensible and definite in 

 amount, but clearly considerably below the 0*3 required by the law of 

 inheritance. 



(b) A second point may be noticed by looking at the diagrams (1) 

 and (2), namely, that from about the age of 32*5 to 52*5 the regres- 

 sion line is sensibly vertical, or when the son dies in middle life, the 

 mean age of death of the father is sensibly uncorrelated with it. In 

 other words, we have the remarkable result that the mortality which 

 in a paper on skew variation by one of us,* has been termed that of 

 middle life is largely uninherited. It is during this period of life that 

 the non-selective death-rate is chiefly predominant. After this period 

 the regression curve becomes sensibly steeper, although not fully up 

 to the steepness of the line given by Galton's Law. This is more 

 properly the inheritance of longevity. The inheritance of duration of 

 life may not be continuous. 



If we seek the best fitting straight line for the regression polygon 

 from 50 years onward we find : — 



First Series. 



Second Series. 



' Peerage,' 52 "5 years of son and on. 



, 



' Landed G-entry,' 50 years of son 

 and on. 



66-680 years 

 69 -686 „ 

 14 -6734 „ 

 9 -6148 „ 

 0-1156 ±0-0232 

 0-1764± 0-0380 



Mf 

 Ms 



<T F 

 <*S 



Rfs 



66-878 years 

 68-960 " „ 

 14-3273 „ 

 10 -4055 „ 

 0-1125 ±0-0243 

 0-1549 ±0-0333 



Results such as these are as close as we could expect, and they mark 

 an increase in the steepness of the regression line from about 0*11 to 

 0-17, an undoubtedly substantial increase of the selective death-rate as 

 we approach old age. The regression line for this old age mortality 

 is marked as jk in diagrams (1) and (2), and we see the advance 

 towards the Galtonian value. 



(c) Below 32*5 years the regression line in figs. 1 and 2, especially 

 the former, seems to indicate increased correlation again, but unfortu- 



* ' Phil. Trans.,' A, vol. 186, p. 408, and Plate XVI. 



