A. H. Trow 243 



wino-ed form of which doubtless carries pollen freely, and (4) various 

 other insects. 



Selfing is therefore necessary and has been carried out in various 

 ways; viz. (1) under a framework covered with muslin, (2) inside 

 muslin bags, the mouths of which have been carefully closed with 

 cotton wool — the most effective open-air method, (3) by isolation in 

 the greenhouse — the most effective method of all, and perhaps abso- 

 lutely safe when muslin bags are used as well. Before making crosses, 

 it is of course necessary to self both parents— a precaution neglected in 

 some of the earlier experiments. 



Constant handling of groundsels soon trained the eye to the per- 

 ception. of comparatively slight differences of form, and at an early stage 

 in the investigation (1904) distinct strains were collected and separated 

 which can be regarded provisionally as elementary species. Three of 

 these, named lyraecox, latifoliits and multicatdis, were non-radiate, and 

 were soon subjected to the test of experimental culture. 



Praecox was first noted in 1904, growing in the Cathays Park, 

 Canliff, in a colony of about a hundred plants, associated with an 

 erectes-like and much larger form. In 1905 a single head of seeds 

 was collected, and the type has since then been cultivated and main- 

 tained as a pure strain for six generations (1906 to 1911). The plants 

 are early and dwarf, with dark-green slightly cut leaves and reddish 

 glabrous stems with long internodes and few nodes. (PI. XV, fig. 1 and 

 PL XVII, fig. 13.) 



Latifoliiis is of garden origin and has been maintained pure in type 

 for five generations. It has broad incurved leaves, glabrous and shining 

 above, and is a tall erect plant like erectus. 



Multicaulis was first observed near Ba,rry, and has been cultivated 

 for five generations. It is rather late, generally produces many strong 

 branches from the axils of the basal rosette of leaves, has glabrous green 

 stems, dark-green leaves, yellowish-cream coloured flowers, and large 

 capitula. (PI. XVII, fig. 12.) 



Genevensis is the name given to a non-radiate type which occurs 

 abundantly, apparently unmixed with other forms, in the vineyards 

 about Montreux, and which resembles praecox. It has small capitula. 

 (PI. XVII, figs. 15 and 16.) ,,: ," ' . - \ . ,•• ■. : •. 



The full pedigrees of these types and a more detailed consideration 

 of their characters are not required for our immediate purposes. 



But it seemed to me of some importance to test whether the rayed 

 character RR was transferable to these lour non-radiate types.^ If one 



