Inconstancy of Improved Races 785 



is exactly what is gained by repeated selections. 

 To my mind this reduction of the size of the cul- 

 tures is probably the sole effect of the repeti- 

 tion. But experience is lacking on this point, 

 and exact comparisons should be made when- 

 ever possible, between the descendants of a 

 unique but extreme choice, and a repeated but 

 smaller selection. The effect of the repetition 

 on the nourishment of the chosen representa- 

 tives should be studied, for it is clear that a 

 plant with 22 rows, the parents and grandpar- 

 ents of which had the same number, indicates a 

 better condition of internal qualities than one 

 w)<;li the same number of rows, produced acci- 

 dentally from the common race. In this way it 

 may perhaps be possible to explain, why in my 

 experiment an ear with 22 rows gave an average 

 offspring with 20, while the calculation, found- 

 ed on the regression alone would require a 

 parental ear with 32 rows. 



However, as already stated, this discussion 

 is only intended to convey some general idea 

 as to the reduction of the cultures by means of 

 repeated selections, as the material at hand is 

 wholly inadequate for any closer calculation. 

 This im.portant point of the reduction may be 

 illustrnted in still another manner. 



The sowing of very large numbers is only re- 

 quired because it is impossible to tell from the 



