<:iu)ssiN(;s in melanilm-violkts 271 



the large flowering lorms ot V. tricolor (lines 10 and 40) and V. ur- 

 vcnsis (line 25) with the small flowering tyjx's of V. arix'nsis (lines 3 

 and 4) ha\ c: petals of ahout the same length as the latter ones. I 

 have also made iiyhrids hetween these and another spe<-ies of Mcld- 

 iiiuni violae viz. \'. Miinhyaud. tlie flowers of which are of the si/e 

 of a middle sized i)ansy. The petals of the hyhrid were not much 

 larger than those of the small flowering forms of V. arvensis. 



Quite contrary to this is the behaviour of a large flowering type 

 of v. tricolor or V. (irocnsis crossed with another large flowering form. 

 In this case the bastard has at least as large flowers as the largest 

 parent. The same was the case when these types wert; crossed with 

 V. Miinhydno. The flowers of these hybrids were at least as large as 

 those of the latter species. 



The behaviour of Viola is evidently quite analogous to that of 

 Oenothera. Heribert-Nilsson (1912) crossed (). Lamarckiana with 

 O. biennis and found that the small flowers of O. hienni.s were do- 

 minant to the larger flowers of 0. Lamarckiana. O. yiyas has larger 

 flowers than O. Lamarckiana and its size of flowers is dominant to 

 that of the latter species. 



The size of the flowers in Fo was only investigated in the 

 case of cross 2 X 10. In order to obtain the means of the 

 F, 5 flowers were measured on each plant. The uppermost 

 flower of the main axis which often is larger than the other and the 

 flowers on the small side branches, which often are rather small, were 

 not measured: only the uppermost, fully developed flowers on the 

 large basal branches were used for determining the mean. The reason 

 why only these flowers were used lies in the fact that the number 

 of these branches is rather constant: the number of the other branches 

 varies much and so is also the case with the number of their flowers. 

 It would have been better, of course, to measure all the flowers on 

 the plant but this would mean too much work; man}' of the plants 

 had about 50 — 60 flowers. I think. h(nvever, that the means obtained 

 in the above-meiflioned manner give a good and exact standard of the 

 size of the flowers: it is the relative variation that is of interest, not 

 the absolute. 



The tables 10 and 11 show the variation of the length of the 

 petals in the parent-lines and of the F^. The classification is made 

 with the mean of line 10 as starting-j)oint and with 0,o mm. between 

 the classes for the spur-petal and 0,8 for the upper petals. 



