SELECTION 453 



variations are of the nature called Mendelian, which do not 

 blend, but are handed on in intactness to a certain propor- 

 tion of the descendants. 



A third even more serious criticism has arisen out of the 

 recent selection-experiments of the Danish biologist Johann- 

 sen, the Dutch botanist De Vries, the American zoologists 

 Jennings and Pearl, and others, which are to some extent' 

 at variance with the Darwinian view, that the average of a 

 stock can be improved as regards a particular character by 

 always breeding from those that show most of it. If the 

 descendants of an individual high-class bean are kept apart, 

 forming what is called " a pure line ", there are observable 

 fluctuations of characters. Some are tall plants, others are 

 short, and so on. But if the tails are selected out and bred 

 from, or the shorts, there is no establishment of a tall race, 

 getting gradually taller, or of a short race getting gradually 

 shorter, nor is there anything to choose between the descend- 

 ants of the tails and the descendants of the shorts. There 

 is no departure from the average of the original pure line. 

 From a mixed wild stock a selection may be made of par- 

 ticular types which start pure lines or distinct races, but 

 when the pure line has been started there is no further 

 progress, select as one may. There is no getting beyond the 

 mean of the inbred line. The reason for this seems to be 

 that the fluctuations within the pure or inbred line are modi- 

 fications or indents, and not transmissible. 



If selection of the best of a pure line does not improve 

 the stock, how do the breeders succeed? The answer is that 

 their success is due to making a good start with a good line; 

 beyond the level of this they cannot pass without the intro- 

 duction of fresh blood from another line. There are obvious 

 reasons, however, why these facts from artificial selection 



