70. 
genera of Berberidaceae, but the genus Berberis may perhaps 
similarly be traced through Mahonia, and. some lost’ type of 
plant related to Hpimedium, which is itself a highly evolved 
member of the family. And similar reasoning may be put forward 
regarding the origin of the Menispermaceae and Aristolochiaceae, 
where in the case of the latter family complete reduction of the 
inner perianth has been carried out, as in Clematis. It is 
significant that the production of peculiar woody structure in 
these groups should be accompanied by reduction and degradation 
of the flowers. 
CONSIDERATIONS FOR THE DELIMITATION OF GROUPS 
iy 
‘With the exception of the primary division into Dicotyledons 
and Monocotyledons, large groups are usually artificial, especially 
if the characteristic fixed upon to distinguish them is a general 
tendency and founded on a single character. Examples are the 
De Candollean artificial groups Thalamiflorae, Disciflorae, Calyci- 
florae, Inferae and to a less extent the Parietales, Centrospermae 
and Amentiferae. The special characteristics indicated by the 
names of some of these groups are general tendencies in many 
families of flowering plants. Although it is still very convenient 
to recognise two main groups of Dicotyledons as Archi- 
chlamydeae and Metachlamydeae (Polypetalae and Gamopetalae), 
a true phylogenetic classification would be better attained in - 
light, 7.e., as a general tendency, for gamopetaly is quite a 
common feature in many so-called polypetalous families.* It is 
perhaps too early to suggest so revolutionary a change. The - 
result, however, would be the closer approximation of such. 
clearly related families as the Anonaceae and Sapotaceae, Caryo- 
phyllaceae, Primulaceae and Gentianaceae, the Rhamnaceae and 
Myrsinaceae, ete. In a new classification, therefore, smaller 
groups must be recognised, which are bound together by a 
combination of characters, and which allow the introduction of - 
the general tendencies of floral development such, at any rate, 
as perigyny, epigyny and apetaly. For this reason I have- 
proposed (Kew Bull. 1921, 185-191) the establishment of the~ 
order Magnoliales as distinct from Ranales, because I consider’ 
that they may have been evolved from separate primitive stocks 
developed on parallel lines, é.e., with strobilus-like hypogynous: 
polycarpellary flowers. The ultimate development of the more. 
primitive families of these two groups, the Magnoliaceae and° 
Ranunculaceae respectively, is entirely dissimilar, the one. 
remaining entirely arboreal, the other mainly herbaceous. Tam 
not sure that a third primitive group should not be recognised, 
say the Dilleniales, whence might be traced the Ternstroemiales 
and the vast domain of Rosales and “‘ Amentiferae.’ 
* For example Anonaceae, Meliaceae, ete. 
