264 



Gates. 



grandparents and also why complete reversion occurs in crossing back 

 with one parent (i). The absence of mcrogony from the grandiflora-ruhri- 

 calyx crosses would also account for the differences in behaviour from 

 the hiennis-muricata series, for in the former the double-reciprocal 

 crosses exhibit no marked tendency to revert to either grandparent, 

 and in the back-crosses a new secondary-intermediate condition arises. 

 The double-reciprocal crosses of de Vries are therefore not of general 

 application, and in the back-crosses of the grandiflora-ruhricalyx series, 

 fractionation of characters rather than reversion occurs. The Mendelian 

 conception of fixed units is here again totally inapplicable, as we have 

 already seen in the case of the foliage, physiological and bud characters. 



K. The occurrence of Mutations. 



The previous sections of this paper have been devoted to a des- 

 cription of the hybrids between grandiflora and ntbricalyx, and a com- 

 parison of their characters with those of the original parents. In the 

 introduction I mentioned that these hybrid families also show that 

 mutation is a process independent of mere hybrid combinations, and 

 a portion of the evidence for that statement will be found in this section. 

 The following table (IX) gives a list of the mutants in my 1912 cultures 

 of these hybrids: 



Table IX. 

 Mutations occurring in the 1912 hybrid families. 



(1) Renner (Ber. d. deutsch. Bot. Gesellsch. 31 1913, S. 334 — 335) has since 

 repeated the work however, and finds no evidence of merogony. 



