376 castle: role of selection in evolution 



and students of geographical distribution in general favor it, 

 nor because DeVries and Johannsen have attacked it, but be- 

 cause the facts of experimental breeding, as I understand them, 

 prove it. 



For DeVries may be claimed the merit of having first syste- 

 matically set about testing the effects of selection by actual experi- 

 ment. His methodical selections for many years in succession 

 of maize, buttercups, striped flowers, and four-leaved clover 

 will long be remembered, but they fall far short of conclusiveness 

 because they were not continued long enough to show whether 

 selection had attained all that was attainable under existing 

 variability or whether further variation in the direction of selec- 

 tion would occur, and because DeVries' cultures were not suffi- 

 ciently guarded from hybridization which might conceivably 

 influence the result. These necessary precautions were fully 

 met by Johannsen, who in the case of beans, which are self 

 fertilizing but show fluctuating variation in the size of the seed, 

 proved that selection generation after generation in a particular 

 direction may be without result, so far as any change in average 

 seed size is concerned. Cases of this sort involve "pure lines," 

 those which are devoid of genetic variation to any appreciable 

 extent in the character studied, size of seed. But in other cases, 

 as where Johannsen made his size selections from a field crop 

 harvested from many different plants, he found that average size 

 was influenced by selection, which he reasonably explains on the 

 ground that the material from which selection was made con- 

 sisted of a mixture of pure lines genetically distinct. The 

 correctness of Johannsen's conclusion has been repeatedly veri- 

 fied in the case of other self-fertilizing plants such as wheat and 

 oats. Attempts were at once made to generalize Johannsen's 

 brilliant demonstration of the principle of pure lines in the 

 following ways : 



1. Since a line of beans long self -fertilized is devoid of genetic 

 variation in seed size, self-fertilization, if long enough continued, 

 will produce lines genetically pure as regards all characters. 

 Selection can not bring about modification of such pure lines. 

 In respect to this generalization it may be said that it remains to 



