— 391 — 



Fg of this cross agrees with Fi of tricolor X arvensis, as the eli- 

 mination of chromosomes in Solanum does not take place before 

 in Fg. In one Fg-individual Rose Stoppel foiind 45, 41, 40, 38, 

 37, 36 and in another 54 and 55 chromosomes and so on. Both 

 in Viola and Solanum some of the bivalents must be formed by 

 chromosomes from the same parent. In Solanum Fi is regular, 

 but in Viola the irregularities begin already in Fj. 



As usually only the two chromosome numbers 13 and 17 

 occur in nature, the chromosome-sets of the hybrids must for 

 some reason or another be unstable, or the piants with these 

 combinations of chromosomes can not be fit to the strong com- 

 petition in nature. It is likely, that the ofTspring of a hybrid 

 sooner or later ends in either a tricolor or an arvensis, i. e. in a 

 13- or in a 17-chromosome plant. But the chromosome-set of a 

 tricolor that has come into existence in this way need not be 

 identical with the parental tricolor. Likely a chromosome-ex- 

 change has taken place between the two chromosome-sets, so 

 that the new^ tricolor has chromosomes both from the parental 

 tricolor and the parental arvensis and in the same manner the 

 new arvensis has chromosomes from the parental tricolor too. 

 Consequently the factors carried by these chromosomes must be 

 transferred together with these, so that we must expect to fmd 

 tricolors that show some of the characters of arvensis and vice 

 versa. In many piaces it is very difficult indeed to fmd a Viola 

 typical in all its characters. 



Even in the stabilized types, the reduction division can be 

 irregular in some cases: In my cultures I have a curious little 

 dwarftype. In 1919 I found in Sophienholm Hills, Sjælland, 

 (Clausen 1921, table I and II, population G, p. 220—221) a 

 plant which I determined as parv, pall. It had very narrow petals 

 and, according to its type, I expected it to approach arvensis. 

 But, under cultivation, I observed that the petals or the sepals 

 might vary slightly, so that the petals usually were a little longer 

 than the sepals. It had a small labellum. I expected a segregation 

 similarto sowing V. 246 and V. 247 p. 382, but in two generations 

 it has bred true to type. .411 the offspring were /M«a-individuals 

 with narrow petals slightly longer than the sepals, but the absolute 

 size of them w^as less than that of many arvensis. The time of devel- 

 opment to the flowering stage is the same as arvensis\ only about 

 IV2 month; considerably shorter than tricolor's (about 2^/2 months). 



