Rose Forms as determined by their Cytological Behaviour . 183 
rose to take place, situated as it would be amidst an abundance of R. pimpi- 
nellifolia , other than by the agency of that rose appears a very remote 
contingency. Thus an egg-cell nucleus with fourteen chromosomes would 
fuse with that of a pollen grain with fourteen, giving us a zygote with 
14+14 (= 28) such as we have. 
In this manner, our hybrid reveals itself as a back cross between 
R. pimpinellifolia and a hybrid pimp inellifolia x coriifolia. Strangely 
enough it betrays its hybrid nature in the same characters as the coriifolia- 
lutetiana cross, i. e. in styles, stigmas, sepals, and prickles. 
From this we gather that the chances are great that reciprocal rose 
crosses should differ as stated above, and that they should be prepon- 
deratingly matroclinous. 
As we have demonstrated experimentally (Harrison, 1920 ), certain 
of the Villosae, Afzelianae, Rubiginosae, and Agrestes set seed by some 
process of apomixis. Moreover, we have likewise proved this to be of the 
facultative order, since seedlings showing hybrid characters have originated 
from flowers, pollinated with foreign pollen, growing on the same bush 
as those castrated, and producing seeds apomictically. Despite this, in the 
sections just named, our experiments lead us to the opinion that in their case 
apomixis is the rule and sexual reproduction the exception. On the con- 
trary, with the lutetiana and dumetorum allies, as well as with certain 
Villosae, our work indicates with some certainty that, whilst they are too 
facultatively apomictical, they favour pollination — a view not incompatible 
with occasional apomixis, as the work of Pace has proved. However, in any 
case, the difference is merely one of degree, and the situation is best summed 
up by stating that except for Rosa arvensis , R. rugosa , and R. pimpinelli- 
folia all the roses brought under examination exhibit apomixis. 
This must not be assumed to be simple parthenogenesis, because, 
as we have explained, the egg-cell in the Villosae has only twenty-one 
chromosomes and in the rest of the superse'ction Caninae twenty-eight. 
Should these cells develop without fertilization the somatic numbers in the 
embryos thus generated must necessarily be twenty-one and twenty-eight 
respectively, which is emphatically not the case. Judging from the 
uniform chromosome number of twenty-eight in the Villosae and of thirty- 
live in the Afzelianae, &c., we must conclude that, whatever cell produces 
the new organism, it must be somatic in origin and therefore possess the 
chromosome number proper to such cells. 
Now let us assemble our facts. We have decided that, whilst the 
sexual roses are genetically pure, the non-sexual (or occasionally sexual) 
microgenes are hybrid in nature. Apparently, therefore, some intimate 
connexion exists^between hybridity and apomixis. Many authors following 
Rosenberg and Strasburger have seen in the mere size of the chromosome 
numbers in plants like the Alchemillae and Antennariae a cause ofapogamy, 
