382 On the Laim of Inheritance in Man 



B. Span. 

 Für Son : 



(7) Probable Spau = 3S"-4G + -40» (Fatlier's Span) + l"-89, 



(8) Probable Span = 38"-38 + -öl 1 (Mother's Span) + l"-88, 



(9) Probable Span = 1S"04. + STö (Father's Span) 



+ -423 (Mother's Span) + l"-70. 

 For Danghter: 



(10) Probable Span = 34"-20 + •425 (Father's Span) + l"-77, 



(11) Probable Span = 34"18 + 473 (Mother's Span) ± l"-77, 



(12) Probable Span = 14"-70 + 'Söö (Father's Span) 



+ -394 (Mother's Span) + 1"-61. 



C. Forearm. 

 For Son : 



(13) Probable Forearm = 10"-65 + -430 (Father's Forearm) + "W, 



(14) Probable Forearm = 10"-88 + -463 (Mother's Forearm) ± "-60, 



(15) Probable Forearm = 5"-58 + -362 (Father's Forearm) 



+ -383 (Mother's Forearm) + "-56. 

 For Daughter: 



(16) Probable Forearm = 9"-43 + -400 (Father's Forearm) + "-56, 



(17) Probable Forearm = 9"-40 + -445 (Mother's Forearm) ±"'56, 



(18) Probable Forearm = 4"-50 + -334 (Father's Forearm) 



+ -371 (Mother's Forearm) + "öl. 



On the right is given in each c;ise the probable error of the prediction*. We 

 see froni these formulae that with the selection of one parent only, the otispring 

 rise to within 40 to 50 per cent. of the selected value ; with the selection of both 

 parcnts to within 70 to 80 per cent. of it. The diniinution of the variability of the 

 array due to two selected parents, is however ouly sligiitly less than that due to 

 the selection of one parent only. 



If we selected for two generations we should have offspring tlie same very 

 nearly as the selected ancestryt". With our values for parental correlatiou, it is 

 obvious that two or three generations of selection will snttice to bring the average 

 of the ofispring sen-sibly ujJ to the selected ancestry, and the regi'ession after this, 



* lu using these formulae for prediction, tbosc not fuUy conversant with Statistical thcory, must 

 bcar in niind thiit tliey H'^e only "le niean or most probable reaults of a whdle array of offspring due to 

 all parcnts of deiinite cliaracters. The validity of the formulae canuot be tested on merely iudividual 

 cascs. This warning is necessary because I hare so often had iudividual cases in man or dogs cited as 

 upsetting the whole of the ancestral law I 



t "The Law of Ancestral Heredity," Biometriku, Vol. ii. pp. 221 — ü. 



