1 62 Sprague . — On the Heteranthus Section 
tive characters, and the very numerous differences in floral 
structure displayed by the different species. The first is 
obviously correlated with the similarity of habitat which 
obtains throughout the group, all the members of which grow 
in gravelly soil among rocks on the upper courses of rivers. 
As regards the internal classification of the section, the 
relationships between the several species are too intricate 
to admit of any satisfactory natural grouping. C. rigidula 
is separated sharply from all the rest by the dichasial branch- 
ing of its inflorescence, and has perhaps its nearest ally in 
C. setosa. 
Geography. 
The following table shows at once that the section is 
characteristic of the Andes (including in this term the Coast 
Andes of Venezuela), and in fact inhabits the upper parts of 
Engler’s subandine (4) region. The only species occurring 
outside this limit are C. rigidula and C. setosa. We have 
unfortunately no locality for C. rigidula more precise than 
Guiana, but it may fairly be assumed from what we know 
of the habitats of the other species that C. rigidula comes 
from the mountains of the interior of Guiana, and possibly 
from the Roraima region. The early isolation of the Guiana 
mountains would explain the separateness of C. rigidula. 
Taking C. setosa (distribn. Andes and Tobago) next into 
consideration, we find that it is really only an apparent 
exception to the andine distribution of the section, for it is 
a well-known fact that both Trinidad and Tobago are 
geologically related rather to the South American mainland 
than to the other West Indian islands, and form the con- 
tinuation eastwards of the Coast Andes of Venezuela. The 
occurrence of C. setosa in Tobago is an excellent illustration 
of the South American affinity of the flora of that island first 
remarked by Eggers (6). 
The distribution of C. epilobiifolia (Andes proper and 
Venezuelan Coast range) illustrates the truly andine character 
of the Venezuelan Coast range recently pointed out by 
