250 MR. A. W. HILL—REVISION OF 
Herbaria (v. Bull. Soc. Bot. France, xii. 1865, p. 82); it is also the plant, collected by 
Mandon under the number 796, which is labelled “ M. anthemidifolium, A. Gray, var.,” 
in the Herbaria of Paris and Kew, though at Paris it bears in addition a note by 
Weddell, * Malvastrum verisimiliter, spec. nov.” Besides these two numbers, Pearce's 
specimen, placed by Baker fil. under M. Pearcei, also belongs to this species. 
The anthers are few and dark purple, and are almost sessile on the staminal column, 
the free portions of the filaments being very short. 
Some difficulty has been experienced in attempting to assign the two preceding species, 
N. Matthewsii and .N. purpurascens, to their proper position in the genus. 
In both species the lamina is deeply cut into five segments, and the segments have 
but few lobes; the lobes are long, linear-oblong in shape, with the margins slightly 
inrolled; the upper surface of the whole lamina is covered with a dense tomentum, 
while the lower is glabrous (Pl. 29. fig. 11). Though these species appear to belong to 
the palmatifid subsection, their nearest allies are to be found probably amongst the 
bipinnatifid forms, and it also seems certain that these two species are not really allied 
to one another. NV. Matthewsii, with its almost polypetalous corolla, seems to be allied 
more closely to JV. stenopetala than to any other species; in its leaf, too, it resembles 
that species in certain respects (cf. Pl. 29. fig. 17). N. purpurascens is also a very 
distinct and well-defined species, although it has been previously confused with 
N. Mandoniana, and has been distributed to Kew and other herbaria with this species 
under the number Mandon, 797. 
The following group of species with bipinnatifid leaves appears to be in reality an 
offshoot from the forms having the palmatifid type of leaf, similar to the pedicularifolia 
section of the genus already described. The central lobe of the palmatifid leaf has 
in this case developed further and has produced lateral lobes, which have superseded 
in importance the original lateral lobes of the 5-lobed leaf, and the bipinnatifid character 
is thus given to the whole leaf. The various species ean be easily separated into groups 
—as can be seen from the key—on account of the distribution and character of the 
tomentum. As a rule the species are hairy on the upper surface of the lamina only and 
the hairs form a more or less dense felt of definite stellate hairs. In some species, 
however, viz. N. acuminata, N. sericea, and N. argentea, both surfaces of the leaf have 
a tomentum of long hairs and a few species (e. g., N. Castelneana, N. cinerea, and 
N. glauca) have a compact densely-stellate felt all over the lamina. Another point of 
considerable importance in this group is the character of the hairiness of the vagina. 
In some species, of which JV. anthemidifolia may be taken as a type, the upper surface 
of the vagina is covered by a dense mass of long silky hairs (Pl. 30. fig. 1), while in 
other species, such as N. aristata, it is quite glabrous. Of the species included in the 
key under this subsection, N. stenopetala and N. epileuca do not appear to be very 
closely allied to the rest of the species. They differ particularly in their crimsoU 
almost polypetalous corollas. N. purpurascens and N. Matthewsii also appear to 
occupy somewhat isolated positions, probably due to our imperfect knowledge of all 
the representatives of this genus. 
