32 MISSOURI BOTANICAL GARDEN. 
in a more or less pronounced depression forming outwards 
a gibbosity of the tube. Besides this character there is a 
difference in the vernation of the foliage, the leaves being 
flat or carinate, while in all other species I have found them 
convolute in bud or occasionally involute, and afurther differ- 
ence is the absence of the superposed accessory buds so com- 
mon in most other species. The remaining species of the 
subgenus Chamaecerasus can be easily divided into further 
sections by the pith of the branches, which is either well 
developed or evanescent: in the first case the branches are 
solid, in the second they are hollow. This behavior of the 
pith is a concomitant of a difference in the union of the 
bractlets. In the solid-branched species there is a tendency 
of the bractlets of one flower to unite more or less with the 
bractlets of the second flower of the same pair or to form 
altogether a complete cupula, and at the same time there is 
a tendency of the ovaries to unite, while in the hollow- 
branched species only the bractlets of the same flower unite, 
or all are free, as are the ovaries. The hollow-branched 
species form aremarkably homogenous section, for which I 
propose the name Coeloxylosteum, in reference to the hollow 
branches. The other section, however, for which the name 
Isika of Adanson in a somewhat enlarged sense may be 
used, presents a very great yariety of different characters, 
especially as regards the bractlets: sometimes these are en- 
tirely wanting, sometimes perfectly distinct and sometimes 
wholly connate, even so far as to form with the inclosed 
ovaries a kind of pseudocarp. The limb of the corolla 
varies from almost regular to distinctly two-lipped; the 
ovary varies from two- to three-celled and the bud scales 
and other vegetative characters also show great differ- 
ences. Though these differences afford good characters 
to subdivide the species of this section into a number of 
well-defined subsections, it is rather difficult to group satis- 
factorily the subsections according to their affinities, since 
their relations cannot be expressed by linear sequence. 
