CYRTANDREÆ (AUCT. C. B. CLARKE). 7 
but is treated as a matter of convenience. And a character which is 
treated in one place as of generic importance is in a nearly allied 
genus considered scarcely as of specific. 
Thus the genera Rhyncotechum and Isanthera only differ in that 
the latter has all the leaves alternate, the former has (at least) the 
upper opposite. But in the closely allied genus Cyrtandra we have 
some species with opposite leaves, some with alternate, many with 
opposite leaves but one of each pair very unequal or rudimentary. 
It may be doubted whether Cyrtandra rostrata, Blume, with quasi- 
alternate leaves differs specifically from C. nemorosa, Blume, with 
opposite subequal leaves. Nevertheless, the character of opposite sub- 
similar leaves not merely is good for numerous species of Cyrtandra 
but is of great value in demarcating natural groups of this large 
genus; and we are not precluded from using it because it breaks 
down at particular points or proves illusory in nearly-allied genera. 
It is somewhat different with the character of confluent anther- 
cells. The anther-cells in this Tribe are sometimes parallel, the slits 
parallel their whole length, and distinet through the whole life of 
the anther ; more often the young anther-slits are parallel below but 
curve inwards so as nearly to meet upwards; in age (sooner or 
later) the slits then often become confluent, and the anther appears 
L-celled. It is never from the beginning 1-celled in the Tribe Cyr- 
tandreæ. 
Benth. and Hook. f. (in Gen. PI.) have employed this character 
of confluent anther-cells to define the sub-tribes under which they ` 
have grouped the genera ; I have preferred to revert to the characters 
from the fruit employed by R. Brown and Aug. Pyr. DC. Not 
merely is the character of confluent anther-cells a question of degree 
of eonfluence; in the great mass of species now accumulated in 
Cyrtandra, many have the anther-cells distinctly confluent. 
The analogy of the neighbouring Orders of didynamons mono- 
petals would not lead us to expect that the difference between 4 per- 
fect stamens, and 2 perfect stamens 2 rudimentary, would be a very 
sharply defined one or of high classificatory value. But in the Cyr- 
