— 7 ~ 



the primary character based on the proportion of the adnate part of the 

 bracts to the free parts is very vague and practically useless. 



The species combined in this group by Schumann, following Baker, 

 Ridley and Gagnepain, are very heterogeneous, and some of them manifestly 

 are wrongly placed in this genus. They are: 



C.petiolata, Roxb., C. cordifolia, Wall., C, Roscoeana, Wall., C. parvifolia, 

 Wall., C. alismatifolia, Gagn., C. sparganifolia, Gagn., C. gracillima, Qagn., 

 C. Kunstleri, Baker, C. lanceolata, Ridl., C. sylvestris, Ridl. 



Having been able to examine some of them in a living state, other 

 ones by Herbarium specimens, and some ones from the excellent plates 

 and descriptions of Wallich 1 will shortly review these. 



I and 2. The first name^d species: C. petiolata, Roxb, diwd C. cordata, 

 Wall, considered as synonyms by Baker and K. Schumann, are true Curcuma- 

 species, showing all principal characteristics of the genus. 



3. C. Roscoeana, Wall, has some resemblance with Curcuma in the 

 (adnate?) pouch-forming bracts and the versatile anther with a short filament. 

 But here the resemblance ends and the differences are as follows: 



Spike: No coma, all bracts rigid, red, erect with a much recurved top 

 (free, according to Wallich, except at the broad base, adnate with the 

 edges, according to Baker). 



Petals: Dorsal lobe not cucullate. 



Staminodes: ovate, subdistant, not connate with the filament. 



Labellum: simple, not lobed, not concave, with two elevated lines in 

 the center, including a median groove. 



Anther: Terminal, articulate to the filament with a broad base; 

 thecae distant much shorter than the large connective which ends in a 

 membranaceous, ciliate crest. 



Decidedly this is not a Curcuma. Perhaps Bentham was right in reducing 

 it to Hitchenia, but the essential characters of that genus are still very 

 vague. 



4.. C. parviflora. Wall. Here the spike certainly resembles much that of 

 a true Curcuma, for there is a distinct, white coloured coma, the fiowerbracts 

 seem to be connate, forming true pouches and the flower, resembles that 

 of Curcuma by the enlargened faux and subfornicate dorsal petal. 



Still I think the most important characters of Curcuma are wanting. 

 The petals converge behind the stamen and staminodes. The latter ones are 

 free from the filament and seem to be placed in an exterior cycle. The 

 labellum is patent, recurved, not lobed not concave, without erect side parts 

 and central bar, but also without a median groove. 



Anther: terminal subarticulate and nutant with a broad base, very short 

 thecae (opening by pores?) and a very large fleshy connective prolonged 

 into a considerable crest. 



