486 DR. OTTO STAPF ON 
specialized part. This is important, because differences which 
pervade a considerable portion or the whole of the reproductive 
apparatus, or even the entire organism, determine that associ- 
ation as characters in which we perceive the phyletie unity or, 
what is the same, the congenerity of two or more species. To 
return to the seeds, described above, it is clear that they differ 
not merely dimensionally, but structurally, and in a way which 
indieates at once that the differenees must be correlated with 
further differences in the structure of the fruit generally, and 
probably also of those parts of the flower which go to form the 
fruit. The wedge shape of the seeds of Pheoneuron and Dicel- 
landra liberica is due to their being closely packed with nothing 
between them. The seeds of D. Barteri, on the other hand, 
with their rounded surface would be impossible under similar 
conditions. They mature, indeed, in the soft, though probably 
somewhat dry, pulp of a berry, the pulp not only covering them 
but, as it seems, also filling up the interstices between them. As 
is usually the case in fruits of this class, the plaeenta takes part 
in the formation of the pulp, and the epicarp remains rather thin 
and delicate. The fruit of Dicellandra Barteri is evidently 
adapted to dissemination by animals, most likely by birds. Its 
occurrence as an epiphyte on old trees (according to notes by 
Barter and W. H. Johnson) becomes thus intelligible. Phæo- 
neuron and Dicellandra liberica behave differently in this respect. 
The corky, closely packed seeds completely fill up the cells of the 
fruit. This has the shape of a berry, but the thin pericarp is 
dry and ultimately bursts as in other Melastomacee, under the 
pressure of the seeds. It dehisces along the median lines of the 
carpels from the top to the bottom, thus breaking up into five 
valves which are ultimately dropped. The seeds remain for 
some time attached to the placentas, which are transformed into 
thiek bundles of fibres and remain behind long after the last seeds 
have fallen. I would suggest that the structure of the fruits 
and seeds of Pheoneuron and Dicellandra liberica are adapted to 
dissemination by water, corky appendages to the seeds being a 
common contrivance with plants which depend on water as the 
dispersing agency, and so far all the specimens which I have 
seen have actually been collected either on the banks of streams 
or in swamps. 
The differences between Dicellandra Barteri on one side, and 
D. liberica and Pheoneuron on the other are, however, not 
