Duration of the Process of Selection. 89 



effected, is attained as a rule after 2 or 3 generations or 

 sometimes after 4 or 5 or perhaps a few more. We are 

 of course speaking of the improvement of a single char- 

 acter. In practice where one is concerned with several 

 or more characters the process of selection may of course 

 last much longer. And this is true even when improved 

 methods render possible a more rigid selection from vast 

 numbers of individuals, as in the case of beet culture 

 where however one is usually dealing with fractional 

 percentages. 



In scientific experiments, with a single character in 

 view, the duration of the period of selection is in my 

 opinion to be placed as a rule at from 2 to 4 generations.^ 

 There is no point in continuing the experiment further 

 unless one wishes to arrive at some decision on the ques- 

 tion of regression which we have just been discussing.^ 



I am convinced that great harm has been done by 

 this exaggeration of the length of the process, since it 

 must have deterred many investigators from instituting 

 experiments of this kind. 



I shall now briefly recapitulate some such experi- 

 ments gathered from the early literature on this subject. 

 The first I refer to is the well-known essay of P. P. A. 

 Leveoue de Vilmorin on the culture of the wild carrot 

 {Dmicus Carota).^ He succeeded in less than 5 genera- 

 tions in improving the wild form until the roots were 



' Cf. Fritz Muller's breeding experiment with maize : Kosmos, 

 loc. cit. 



~ A well-known practical difficulty is presented by the fact that 

 it is only after cultivating a plant, which is new to one, for a con- 

 siderable number of generations, that one gets to know its needs in 

 the matter of cultivation, manure, artificial fertilization, selection, etc. 



^L. De Vilmorin, Notices sur I'amcUoration des pJantcs par le 

 semis, i886; cf. pp. 10-12. Also Carriere, Gardeners Chronicle, 1865, 

 P- 1 1 54- 



