on the outer or lower side ; a fact, however, wliich can only 
be ascertained before the opening of the corolla, the flowers, 
in the greater number of species, ])cc()niinij rcsupinate in 
the expanded state, — a circumstance which does not appear 
to have been before remarked. The relation, thererore, 
not only of the corolla, but of the calyx and stamina to the 
axis of inflorescence, is different in these two tribes. 
2dly. In Goodenoviae, the greater part of the tube of 
the corolla is formed by the cohesion of 5 laciniae, the 
distinct inflected margins of which are, in most cases, 
visible nearly to its base. These laciniae are> in some 
cases, unconnected, as in Diaspasis, and more remarkably 
still, in Ci/phia, which is actually pentapetalous. I have 
observed no such structure in Lobelia. 
3dly. At the period of burstinc: of the aiitlicrjp, the 
stigma in Lobelia is almost completely evolved, and 
capable of receiving impregnation from the pollen of the 
sam^ flower : the function, therefore, of its surrounding 
pencil is similar to that of the hairs, which* are almost 
equally obvious in many compositae, especially Cinaro- 
cephalae. On the contrary, in Goodenoviae, the stigma at 
the same period is hardly visible, and is certainly not 
then capable of receiving impregnation from the pollen 
of its proper flower : it is, therefore, either impregnated by 
the antherae of different flowers, or, in some cases, at 
a more advanced stage, by the pollen of its own an- 
therae, which is received and retained in the Indusium. 
To these arguments for the exclusion of Lobelia I may 
add, that, in the greater part of Goodenoviae, with dehiscent 
fruit, the dissepiment is parallel to the valves of the cap- 
sule, in which respect they differ equally from Lobelia 
and the valvular-fruited Campannlaceae ; and lastly, that 
many species of Lobelia, as well as Campannlaceae, con- 
tain a milky juice, of which there is no instance in Goode- 
noviae. If, therefore, in Lobelia the pencil surrounding 
the stigma, and the irregularity of the Corolla, which, how- 
ever, in some species is hardly perceptible, be considered 
as characters sufficient to separate this extensive genus 
from Campannlaceae, it may form a separate order, admit- 
ting^ perhap$> <^9ui)jsUyisicm several distinct geneik.** 
