DICELLANDRA AND PHHONEURON. 485 
testa is thin and brittle, with the outer walls of the epidermal 
cells much thickened, and often produced into short papille 
(Pl. 19. figs. 8, 10, and fig. 1 in the text), thereby causing the 
granular appearanee mentioned above. The embryo is more or 
less ovoid, and occupies the part of the seed underlying the granu- 
lar portion of the testa (Pl. 19. figs. 7, 8, and fig. 1 in the text). 
In Pheoneuron, on the other hand, the seeds are approximately 
of the shape of a short, thick wedge ; the raphe is much larger and 
consists of a uniform spongy or corky tissue, whilst the surface 
of the seed is dull brown and smooth, and the embryo is more 
or less cylindrie and oceupies the anterior side of the seed down 
to or beyond the base of the raphe (Pl. 19. figs. 14-16, and fig. 2 
in the text). 
These differences in the structure of the seeds are, in my 
opinion, much more fundamental than those adduced from the 
andreecium, on which Gilg based the distinction of Dicellandra 
and Pheoneuron. It is true, our present system of Melastomaces 
rests to a very great extent on the structure of the androecium— 
more so than is the case in any other order of Phanerogamx,— 
and particular stress has been laid on the homoeandry or heterandry 
of the flowers. But, how do the stamens differ of the outer and 
inner whorl in Dicellandra, as defined by Gilg? They differ in 
the size of the anthers and of the small appendages at the base 
of the connective, and, what is considered more important, in the 
presence of a filiform prolongation of the connective between 
the anther and the basal appendages in the outer whorl of 
stamens. The difference, I may add, in the size of the anthers 
is considerable in Dicellandra Barteri, but slight in D. liberica 
(Pl. 19. figs. 2, 3, and 12, 13). On the other band, Pheoneuron 
is supposed to have all the stamens alike; but a closer exami- 
nation reveals even here, at least in some of the species which I 
describe in the second part of the paper, a tendency towards 
heterandry, which manifests itself in a very slight prolongation 
of the base of the connective, so that the anticous appendages 
are somewhat removed from the anther in the outer whorl of 
stamens (Pl. 19. figs. 17, 18). It is clear that the differences 
used so far to distinguish Dicellandra and Pheoneuron are 
purely dimensional: they show themselves at a relatively very 
advanced stage, and they do not point to greater differences in 
the general structure of the flower, such as would manifest them- 
selves at an earlier period, and in more than one particularly 
