Breeding experiments which show that hybridisation and mutation etc. 240 



growth (see fig. 15). Wliile the main stem was still short a circle of side 

 branches was formed in all cases, the branches growing directly up- 

 wards until they reach nearly the height of the main stem. The latter 

 then shoots up again, so that the largest dwarfs are sometimes nearly 

 as high as some of the tails. But the internodes are always short. 

 Fig. 16 is from a photograph taken September 26, when in full bloom. 

 The dwarfs have a type of foliage of their own, quite different from 

 any of the tails, so it is quite impossible to say what relation it bears 

 to either parent type of foliage, except that it is very different from 

 both, though in many features intermediate. The plants are wholly 

 different from pure 0. nanella grown under the same conditions. The 

 latter is very small and stunted, usually unbranched, and more or less 

 obviously diseased. Intermediate between the two as regards size 

 and branching are the dwarfs derived in Fo from rubrinervis x nanella. 



This culture, — (see pedigree i), yielded 10 of these dwarfs to 32 rubri- 

 nervis. It seems probable that the better health of the extracted dwarfs 

 is due to the failure of disease-producing bacteria to be transmitted 

 through the cross. The production of sliort internodes is a unit-diffe- 

 rence, quite independent of disease-inheritance. The dwarfs of culture 

 176 also differed from those in the grandiflora-rubricaly x crosses, in 

 having more crinkled leaves and less brittle stems. The stems of the 

 latter were so remarkably brittle that when struck a blow they would 

 break squarely across almost as though cut with a knife. 



Thus it appears that the dwarf character is the same whether 

 derived from tlae rubricalyx or the grandi flora grandparent, but is 

 inherited in different ratios, which in tlie latter case agree fairly well 

 with that from selfed offspring from the grand-parental plant. The 

 essential feature of the dwarf is the development of short internodes, 

 but along with this goes a peculiar habit of branching, extremely 

 brittle stems, and other changes. 



G. Other bud characters. 



The main morphological differences between the buds of grandi- 

 flora and rubricalyx have already been pointed out (vide supra, p. 227) 

 and in section E 2 the inheritance of the pigment difference, R, was 

 described. It now remains to consider the inheritance of the morpho- 

 logical bud differences, which are th.e same as those distinguishing 

 grandi flora and rubrinervis, for in the absence of the pigmentation- 

 character R, rubricalyx becomes rubrinervis. 



