Data for the Problem of Evolution in Man. 



ment of brothers during their lives being as a rule much more alike 

 than that of father and son. It must be noted that the predominance 

 of the non-selective death-rate in middle life, so marked in the latter 

 case, no longer appears in the case of brothers. This would suggest 

 that the environments of father and son differ most in middle life and 

 are then much more unlike than those of brothers. 



8. We conclude this first stud}- by putting on record formulae for 

 estimating the age at death of a man, using the theory of multiple 

 correlation as developed in a memoir* by one of the present writers, 

 and taking as basis the second and third series, which seem to us to 

 present the best results. 



Let P be the probable age in years at death of a man, F be the age 

 at death of his father, Si of his first son, S 2 of his second son, B x of his 

 first brother, Bo of his second brother. Then we have the following- 

 cases : — 



Prediction of Age at Death. All Deaths after 20 Years. 



(a) From age of father at death — 



P = 49-8201 +0-1682 F, 2 = 16-9259. 



(b) From age of brother at death — 



P = 45 1063 + 0-2602 B l5 ^ = 16 2555. 



(c) From age of son at death — 



P = 58-6771 +0-1196 S x , 2 = 14-2850. 

 ((7) From ages of father and brother at death — 



P = 37-6647 + 0-12685 F + 0-24502 B T , 2 = 16-4099. 

 (e) From ages of father and son at death — 



P = 48-7991 +0-15706 F + 0-11168 S, 2 = 14-1573. 

 (J) From ages of two brothers at death — 



P = 35-7930 + 0-206475 (B x + B 2 ), 2 = 15-9052. 



(g) From ages of two sons at death — 



P = 54-3928 + 0-09497 (Si + S 2 ), 2 = 14-1987. 



(h) From ages of brother and son at death — 



P = 44-2601+0-1046 S + 0-2514 B, 2 = 13-8508. 



Here 2 is the standard deviation of the array of men for each group. 

 Such formulae f seem to us to give a quantitative accuracy to much 



* " Contributions to the Mathematical Theory of Evolution. III. Kegression, 

 Heredity, and Panmixia," ' Phil. Trans.,' A. vol. 187, pp. 253 — 318. 



f In obtaining the formulae for prediction from the age at death of Uvo relatives, 

 certain assumptions have had to be made. Thus the correlation of ages of a man 

 and his grandfather and of a man and his uncle at death, being at present unknown, 

 were taken to be half the correlation of father and son. This cannot be far wrong, 

 but the actual values ought to be found. We did not feel justified in assuming the 



