IN CEYLON. 
99 
in pairs, each pair alternating with the corolla segments 
and being attached to the corolla as one or two filaments. 
When there are less than ten stamens some are arranged 
in pairs and others singly, but always as a series of five 
alternating with the corolla lobes. It would thus appear 
that if we have to maintain the importance of abortion 
we must have the primitive hermaphrodite flower with 
numerous stamens arranged in pairs or otherwise. It would 
by no means necessitate a complex series of changes to 
derive the five-seried staminal whorl of the female flower 
from that of the male, the barrenness in the anthers of the 
staminodes being followed by a reduction of each pair to a 
single member ; the inner stamen is always small. 
Now, let us take the case of D. Ebenum, the flowers of 
which may be polygamous, dioecious, or monoecious. We 
have seen that though the male inflorescence and female 
flowers usually occur on separate trees, examples have been 
found where a true female flower replaced a member of a 
male inflorescence. The staminal whorl of the female 
flowers is constant in number, form, and orientation. There 
are always eight epipetalous staminodes alternating with and 
opposite to the corolla segments. In the male flowers a 
variable number of stamens occur, but coalescence of the 
filament occurs so frequently that the resultant number of 
groups is relatively small. In one case there were sixteen 
stamens, either single or in groups of two and three, while in 
another flower thirty-two stamens occurred showing various 
degrees of coalescence, but in each of these the resultant 
groups were eight in number and arranged so as to alternate 
with the corolla segments. Hiern* mentions the occurrence 
of as many as thirty-two stamens on eight filaments. 
This being the case, we have another example in which 
the origin of unisexuality can be easily understood, since the 
staminal whorl always forms a series of eight. The monoe¬ 
cious state would represent the intermediate between a 
* Hiern, l.c., p. 208. 
