Inconstancy of Improved Races 789 



rows. Choosing always the principal ear, 

 the figures may be taken to indicate the de- 

 gree of personal deviation from the average of 

 the race. But whenever we make a mistake, and 

 perchance sow from an ear, the deviation of 

 which was largely due to partial variation, the 

 regression should be expected to become con- 

 siderably larger. Hence it must be conceded 

 that exact calculations of the phenomena of in- 

 heritance are subject to much uncertainty, 

 resulting from our very imperfect knowledge 

 concerning the real proportion of the contribut- 

 ing factors, and the difficulty of ascertaining 

 their influence in any given case. Here also we 

 encounter more doubts than real facts, and much 

 remains to be done before exact calculations 

 may become of real scientific value. 



Keturning to the question of the effects of se- 

 lection in the long run, two essentially differ- 

 ent cases are to be considered. Extremes may 

 be selected from among the variants of ordinary 

 fluctuating variability, or from ever-sporting 

 varieties. These last we have shown to be 

 double races. Their peculiar and wide range of 

 variability is due to the substitution of two 

 characters, which exclude one another, or if 

 combined, are diminished in various degrees. 

 Striped flowers and stocks, "five-leaved" clover, 

 pistilloid opium-poppies and numerous other 



