OP CEYLON AND INDIA. 
395 
fig. 8, a few examples are shown, selected almost at random 
from the few flowers at my disposal, and showing all forms 
from simple subulate to the obcuneate and deltoid toothed 
forms like those of the American Lophogyne. 
The fruit is almost, but not quite, isolobous, and has 
usually twelve, but often more, ribs. The extra four ribs, as 
compared Avith the fruits of the other species, are intercalated, 
and often do not run up to the top of the fruit, much as is 
the case in the extra ribs often found in the Burmese form 
of Dicræa Wallichii. This difference alone is certainly not 
sufficient toalloAv of this species being generically separated 
from the other Hydrobryums, with Avhich it agrees so closely 
in thallus morphology, spathe, and other floral characters. 
The genus Hydrobryum thus contains some of the most 
peculiar of the many peculiar plants that we have described. 
If only the species last described, H. olivaceum and H. Griffi- 
thii, were knoAvn, we might be somewhat puzzled to know^ 
what to make of them, but the intermediate steps afforded 
by H. sessile and H. lichenoides lead back Avithout any serious 
discontinuity to thethalli in the forms already dealt wdth. 
The dwarfing of the secondary shoots, and the enlargement 
of the thallus to do the work of assimilation or to produce 
more secondary shoots for that purpose, which we have 
already seen in Dicræa, &c., is carried to an extreme in this 
genus, and Ave get here forms as dAvarf in every respect as 
Lawia itself. Probably directly correlated Avith this is the 
fact that the tAvo genera have very similar habitats, and that 
the species are very often mixed with one another. Both 
inhabit water that is liable to become shallow very quickly, 
and consequently perhaps are the genera which are found 
at the greatest elevations, i.e.^ on the whole in the smallest 
streams, and in the most northern localities, i.e., in the 
rivers which most rapidly run dry. During the early part 
of the vegetative season the Avater is usually deep enough. 
