1 7< I LOCK : STUDIES EN PLANT BREEDING 



Tin- above curve was thus obtained by plotting the actual 

 -inn- oi deviations from the value 50 per cent, shown by the 

 percentage numbers of heterozygotes in 247 samples, each of 

 which contained a known number of individuals (average 

 number of individuals 450). The "mode" of the curve lies 

 v.i v nearly at 50 per cent., and the mid departure found gra- 

 phically from the area of the curve is T9. 



We have thus arrived by experiment at a value for the 

 distance from the mode within which half the observed devia- 

 tions lie. Now the theoretical probable error for samples con- 

 taining 450 individuals is 1*6,* so that the experimental value 

 is somewhat too high.f Certain possible sources of minor 

 error, in addition to those pointed out in the footnote, have 

 been alluded to in the case of the particular series in which 

 they occurred (see Expts. 30, 54). 



On the whole we may say that the experimental curve 

 agrees with the theoretical one within reasonable limits of 

 error, and the result is therefore the same as would be expected 

 supposing the deviations from 50 per cent, to have no special 

 -iiiiiificance, but to depend simply upon the size of the 

 samples. 



The above facts lead to the conclusion that in the case of 

 any particular grain the chances of one or the other character 

 1 »< i i i-_ r present arc equal. 



* An allowance for tlio fact that the samples arc not all of the- sam< 

 • i/.. would presumably somewhat raise ihis. value. 



f The normal fonta of the unsmoothed .nine in fi^. I is somewhat 

 upoilt by 5 extreme variants, which shoM deviations from the mode 

 i. |.. -.•lively of lo in i* cases and of 9 in 3 case?. It will ho well 

 therefore to look a little more closelj at these variants : 1 of 40 per 

 cent., 3 oi 59 por cent., and 1 of 60 per cent. The sample showing 

 W per cent. net. included only 218 individuals ; the 1st sample showing 

 .".'.i percent, het. included only i:is individuals (Table 4); the 2nd 

 sample showing 59 i» i cent. het. included only -\u\ individuals and 

 Found among an unusually variable set of values (Table 10). 



fed lampll showing 59 per cent. het. unhid.'. I tSS individuals 

 le -'II) The fie-. ..t CD p.-r <-.iil. ln-1. oe, uired in Tahlo . r », in 



which a pen v I error wan noied. and the sample included 



only 305 ni.ii\ idus 



