be explained later, are considered as originating through the loss of 

 characters and hence would all come under De Vries' category of 

 retrogressive mutants. Even in the case of 0. rubrinervis, which De 

 Vries cites as a progressive mutant, while there is no apparent loss of 

 characters, experimental studies which I can not undertake to ex- 

 plain here have also shown that there is no reason for assuming that 

 anything new or additional has been acquired which was not present 

 in 0. LamarcJciana. 



In the case of 0. gigas, another "progressive mutant" of 

 De Vries, it is also believed to be a false conception to say that addi- 

 tional characters have suddenly appeared in its germ plasm, for this 

 mutant differs from all the others I have studied in that it has 

 double the number of chromosomes present in the parent form, 

 0. Lamarckiana. This means that when 0. gigas originated the^e 

 was a sudden doubling in its number of chromosomes, but this does 

 not necessarily imply anything more than a duplication of the set 

 of chromosomes already present in the parent. It will be seen, how- 

 ever that on account of this change in the number of chromosomes, 

 0. gigas comes in a different category from all the rest. It has lost 

 no characters, but seems to have merely had a duplication of the 

 chromosome set present in 0. Lamarckiana. The assumption that 

 new unit characters suddenly appeared in one of the germ cells giving 

 rise to the 0. gigas mutant is just as vmcalled for as in the case of 

 the other mutants. 



Several series of facts which I can not take up here support my 

 view that the mutants, with the exception of 0. gigas, have arisen 

 through a process of analysis, different sets of characters being lost 

 from the germ cells giving rise to each mutant. The peculiar phe- 

 nomena of hybridization among these forms are simply explained on 

 this basis. In general, when two mutants are crossed, the parent 

 form, 0. Lamojrckiana, appears in the first generation of hybrid off- 

 spring, in addition to both the mutants. Other complexities may 

 enter in some cases, but these are the only facts that it is necessary 

 to consider here. These two results, first that the mutants them- 

 selves breed true when self -pollinated and never revert; second that 

 when crossed they produce the parental type, 0. Lamarckiana, at 

 once, but never any type showing new synthetic characters, are both 

 easily explained if we suppose that each of the mutants was lacking 



