W. Batkson and Carounk Pkli.ew 31 



Since llu! ciirvrd pods inuy be combined with foliar parte aj)proxi- 

 niatuiy like those of the types, it seemed at first sight possible that 

 these two characters might be governed by separate factors of the 

 familiar kind, but the genetic evidence at once disproves this suggestion. 

 The comiwsition of the families in which rogues appear from types 

 makes any scheme of this kind inapplicable, and the fact that the 

 families derived from intermediates consist chiefly of rogues is of course 

 quite at variance with such a possibility. 



As regards the appearance of rogues from types we plainly have to 

 deal with an irregular phenomenon. The gametes capable of producing 

 rogues are given off sporadically and not in accordance with any system 

 that we can perceive. This is abundantly proved by much evidence 

 and especially by the case of the strain of Duke of Albany quoted at 

 the beginning of this paper, in which a strain after breeding true on a 

 large scale for two generations gave rise to six thorough rogues. In 

 view of this irregularity it may be supposed that some phenomenon of 

 mosaicism may be involved. We have much inclined to this hypothesis. 

 Can the type plants or some of them contain " islands " of rogue 

 tissue ? When characters are distributed in a plant mosaically, that 

 is, to speak strictly, not according to any geometrical system, families of 

 irregular composition are to be expected. The formation of rogue- 

 gametes b}' the types must be describable in these terms, but we can 

 get no evidence that the somatic tissue of the types is thus actually 

 mosaic. We have often sowed seeds from the most curved pods on 

 plants otherwise type, but rogues did not come with increased frequency 

 from such pods. In one case a whole branch had curved pods while 

 another branch was typical, but all the offspring were typical. The 

 evidence suggests that there is a gradation in genetic proportion from 

 the types which breed approximately true to the intermediates which 

 throw a large majority of rogues, the group of three plants which gave 

 only 6'5 types to one rogue forming a connecting link between them. 

 Moreover the fact that these three j)lants were reckoned as of the inter- 

 mediate class when judged by their somatic characters, proves that the 

 somatic gradation imperfectly corresponds with the genetic. But there 

 is still a wide gap between 5'6 types to one rogue as found in family 145 

 (p. 23) and one type to two rogues as found in family 112 (p. 24). 



No hypothesis of mosaicism, even if otherwise probable, will repre- 

 sent the results of crossing types with rogues, which are, so far as we 

 know, without any close jjarallel among plants or animals. We must 

 suppose that the gametes concerned in the production of the cross-bred 



