19 



group, but esiiecially from their fauna. Their structural peculiarities and the 

 large number of sijecies \\i>uld certainly indicate a very ancient occupation of 

 this group of islands by their imniigraut ancestors. These four genera are in 

 reality so closely related to each other that botanists have been imable to set 

 exact specific limits, and it is true that some species, as for example Clermoniia 

 ^Yaimeac, puzzle the botanist as regards their generic value. This was however 

 brought about by the basing of these genera, by Ch. Gandichaud on irrevelaut 

 characters, which, as new species were discovered, became untenable. It was 

 due to the following of Gaudichaud's classification by later botanists that heter- 

 ogenous sj)ecies were brought together and closely related ones were separated. 

 Ilillebraud, who has been by far the best connoisseur of our tlora has arranged 

 them in what must be considered the best possible classification. 



THE BACCATP] GENERA OF THE HAWAIIAN LOBELIOIDEAE. 



To this group belong the already mentioned genera VJcnnontia, Cyaii<a, 

 Delissea, and Rollandia. Of these, the genus ('yaiua possesses by far the most 

 species, and is with Clcnnoiiiia, still in the process of evolution, while the genus 

 Delissea is to be considered in a decadent stage. The characters of these four 

 genera are not very well ilefined and as llillebrand pointed out were not quite 

 satisfactoril\- arranged. Gaudichaud estalilished these genera mainly on the 

 length of the ealyeine lobes, a character not at all relialile as is well illustrated 

 by certain species of CUrinnntin still in thi' process of evolulinn. Ilillebi-and 's 

 classification was based on more extensive material, and though a great many new 

 species have been discovered since Ilillebraud, his classification need not be al- 

 tered; on the contrary the newly discovered species confirm his classification. 

 The genus CIcn)W)itia is at once recognized by its cymose inflorescence, and also 

 by its peculiar branching habit, which to some extent occurs in a few species 

 of Cyanea. The three remaining genera have all a racemose inflorescence. Cler- 

 iiiontiae are usually trees or shrubs of which the latter type is quite commonly 

 epiphytic. The type of branching is candelabra-like as can be seen by the accom- 

 panying illustrations. The crowns are either dense and I'ouuded, or open and 

 loosely branched, in one instance (Ch rmontia Halcakalciisis) the branches are 

 stout and ascending. (See Plate VIII.) Their corollas are rather large and fleshy 

 (CI. arborescens CI. drepanomorplia, etc.) or thin and slender as well as small, 

 to this category belong Clcrmontia parriflora, CI. micrantha and CI. midfi flora 

 the latter in some respects forming with CI. pcrsicifolia the intei'mediate of the 

 fleshy and large flowering species. Their foliage is usually thick leathery and 

 shiny above, with two exceptions (CIcnnontia parri/hira and Clermontia Ka- 

 I'cana). All species have a tendency to become, or indeed are, more or less 

 epiphytic. The large orange-colored berries are a source of food for the birds 

 which deposit the seeds with their droppings in the forks of trees, or on thick 

 mossy branches, or in the fibrous masses of the stout trunks of the tree ferns 

 Cibotiiim Mcmiesii and (.'. Cltamissoi. Clermontia Peleana has been observed l)y 

 the writer in the uppermost branches of Ohia lehua trees (Metrosideros coUiiia 

 polymorpha) on Hawaii, some sixty to eighty feet up in the tree. (PL IX.) 



Clermontia Ilalcalialcnsisi which is the largest of all species of Clermontia 

 is t(>rrestrial (inl.w a1 least so far as lias been observed. Thcv are never single 



