PROTEACE. 401 
All these plants have certain constant characters in common: a 
tetramerous perianth, valvate in the bud; four stamens superposed 
to the perianth-leaves; a free gynzceum with a one-celled ovary ; 
a dry fruit; an exalbuminous seed with its radicle inferior. The 
‘variable characters are: the conformation of the perianth (regular 
or irregular) ; the level of the insertion of the stamens ; the union 
or freedom of the anthers; the presence or absence of a disk, and, 
if present, the extent to which it surrounds the gyneceum; the 
form of the style, especially its stigmatiferous portion; the number 
of ovules, their direction and form (anatropous or orthotropous) ; 
the consistence of the pericarp, which may be dry or fleshy, dehis- 
cent or indehiscent. It is on these variable characters that the 
subdivisions of the order have been based. Since the days of 
R. Brown it has been divided into two grand sections in the first 
place. The fruit is indehiscent in the one (Nucamentacee); dehis- 
cent in the other (Foiiculares). But this character will sometimes 
unfortunately separate widely two genera that would be considered 
identical if the flowers alone were examined. Thus we may cite 
Andripetalum, possessing the flower of Roupala, without any ap- 
preciable difference; but as the follicle is not dehiscent in the 
former, it is quite removed from the latter in the classifications in 
vogue. Strangea, again, said to have quite the habit and inflo- 
rescence of Persoonia, has dehiscent fruits, and cannot come in the 
same tribe. /elicia, so similar in flowers and vegetative organs to 
both Roupala and Knightia, has been relegated by several authors to 
quite another series. Moreover, in collections are found numerous 
examples possessing only the flowers; there is a fair number of 
genera, more or dess contested, of which the ripe fruit is unknown, 
and whose place it is impossible to fix, if we are to make this feature 
of dehiscence or indehiscence of primary importance. Hence we 
base our divisions first on the characters of the flowers. In the 
series thus established we look for the number of seeds. This 
enables us to distinguish in Hndothriee, for instance, two secondary 
groups: Huembothriee, which has at least four seeds ; and Grevillee, 
which has at most two. Among these last the ovules may be 
descending and orthotropous, or ascending and anatropous; in this 
way we can distinguish the genera Bellendena, Roupala, Lambertia, 
&c., from Helicia and Xylomelum, which have nearly the same 
flower. We next take into account the regularity of the perianth, 
VOL. II. DD 
