Jlermbstcedtia.] cvi. amarantace.e (baker and clarke). 25 



Species 15, iu Tropical and South Africa. 



*Eu-HERMBST.EDTrA. — Sterile stamens alternate with and distinct from tlj6 

 fertile filaments, oblong, divided at the top into two lanceolate lobes. 

 Leaves Ih in. long, linear. 



Sterile stamens about as long as the fertile . .1. H. recurva. 



Sterile stamens much longer than the fertile . . 2. H. elegans. 

 Leaves A-1 iu. long. 



Leaves narrowly oblong, parallel-sided . . . Z. H. dammarensis. 



Leaves widely obovate . . . . . . 4. H. Gregoryl. 



**GoMPnRO-HERMBST.EDTiA. — Filameuts 5, oblong, slightly widened upwards 

 with two lobes at the top, oblong, divergent or unequal, or very short and almost 

 obsolete ; between these lobes is the sessile anther. 



This subgenus has been put back into Celosia by Schinz and Lopriore ; a view 

 which involves the assumption tliat the 5 stamens which are fertile in Ea-Hermh- 

 sf(Bdtiaa,re sterile in Gomphro-Hermbsta:dtia. It is preferable, with J. G. Baker, 

 to suppose the bifid processes of the former to be yet more deeply bifid in the latter 

 and fused in separate halves with the fertile filaments. What Lopriore regards as 

 a very broad filament is then a linear filament fused on either side with half a 

 sterile filament. 



Lateral lobes at the top of the filament manifest 

 Style 3-fid. 



Perianth J-j^in. long . . . . . o. H. argenteiformi*. 



Perianth ^ in, long or less. 



Heads short ovoid . . , . . . 6. JfiT. Schinzii. 



Heads cylindric . . . . . .7. H.falcata. 



Style 2-fid . . . . . . . 8. J£. linearis. 



Lateral lobes at the shoulder of the filament obscure or 

 obsolete. 

 Plant nearly glabrous . . . . . . 9. JI. longistgla. 



Stem and leaves jiubescent, at least when young. 



Stem liairy even in age ..... 10. H. xcahra. 



Stem in age nearly glabrous .... 11. M. Fleckii. 



***PsErDO-HERiiBST.i:DTiA — Filaments 5, depressed-ovate; on the summit of 

 each is a short or minute fertile filament. This subgenus has also been put back 

 into Celosia by Schinz and Lopriore. It is sn])posed here that the broad ovate part 

 of the filaments is formed of the fertile filaments fused in some way with the sterile. 

 At all events, the line between H. scahra (with the lateral processes at the top of the 

 filament sometimes nearly obsolete) and H. angolensis (with no trace of such pro- 

 cesses) is so narrow that the two species must be placed in the same genus. 



Leaves 1^ by ^ in., o])long 12. if. angolensis. 



Leaves A by \ in., obovate ..... 13. H. sjiathulcBfolia. 



1. H. recurva, C. B. Clarke. Rhizome thickened, very long, per- 

 pendicular. Stem 1 ft. long, divided repeatedly from the base. Leaves 

 l-\\ in. long, linear, the lovvrer linear-oblong, nearly glabrous. Spikes 

 ^ in. broad, continuous, simple, J-IJ in. long. Perianth \ in. long; 

 segments oblong, triangular at the tip, striate on the back, more or less 

 pink on the back when young, fuscous in fruit. Filaments linear, 

 shorter than the staminal tube ; anthers small, round ; pollen globose, 

 1^0 11 in diam., tubercled ; interjected processes (infertile stamens ?) 

 scarcely longer than the stamens, quadrate-oblong, divided at the top 

 into two lanceolate spreading segments. Style very short, 3-tid. Cap- 

 sule ovoid, with few seeds. — II. Caffra^ Moquin in DC. Prodr. xiii. ii. 24G, 



FORES 



\ i. .L.5 



