600 THE AMERICAN NATURALIST [Vol. XLV 



be similar to the behavior just described, and increases 

 the probability that 0. Lamarchiana may have had a 

 similar origin in nature and perhaps also in gardens. 

 As already pointed out in this paper, it must be assumed 

 that 0. Lamarchiana lias undergone crossing which has 

 determined its present characters, and that its ancestry 

 is therefore "impure." Whether it has resulted from 

 any particular cross is of less interest from the evolu- 

 tionary standpoint. As I have tried to show in the in- 

 troduction to this paper, it probably must be conceded 

 that the mutations of 0. Lamarchiana are connected with 

 previous crossing in the ancestry. This crossing was ac- 

 companied by a disturbance of the germ plasm, such 

 disturbance manifesting itself in the occasional pro- 

 duction of various aberrant types displaying whole 

 series of new characters. But there are certain reasons 

 why the resulting "disturbance" appears to be some- 

 thing more than a case of hybrid splitting. I shall not 

 go into the details of this question in the present paper, 

 but I may point out one or two facts. Even if (as is 

 quite possible) certain mutants, such as 0. lata and 0. 

 nanella, which are evidently retrogressive forms, could 

 be so accounted for; there are at least two forms which 

 could not be brought into this category. The first of 

 these is 0. gigas with its tetraploid number of chromo- 

 somes and a new series of characters, in part, no doubt, 

 the result of the chromosome doubling. I ('09/?) have 

 suggested that this origin is probably similar to that of 

 many tetraploid species in nature. Ordinary hybrid 

 splitting will not account for it. 



Another mutant which is clearly positive or progres- 

 sive, and which can not be explained as a result of hy- 

 brid splitting, is 0. rubricalyx, which appeared in my 

 cultures in 1907 as a mutant from 0. rubrinervis, show- 

 ing a great increase in anthocyanin production (see 

 Gates, 'lib, for an account of its behavior). These two 

 cases alone make it necessary to assume that "muta- 



