20 ME. C. B. CXABKE ON INDIAN SPECIES OE CTPEEUS. 



nearly all the flowers of a spikelet or of a plant, so that errors 

 may easily arise here. 



The tissue of the filaments is of lax oblong cells, and in many 

 species (as G. globosus) this is so lax as to give a scabrous-papil- 

 lose character. The filaments look here rough to the naked eye, 

 but there is hardly a species in which the microscope will not 

 reveal a structure essentially the same. 



The persistency of the filaments is very characteristic of some 

 species (as of G. globosus), where they remain attached, without 

 anthers, after the glumes have fallen. 



The anthers themselves are always normal, oblong or linear- 

 oblong, yellow, red, or white, with purple spots. 



The anthers are in the majority of species simple, muticous, or 

 with a minute apiculus at the apex ; but in many species the 

 anther has a lanceolate scarlet crest, very constant in size for 

 each species ; in some minute and not always to be made out, in 

 others prominent, in G. platyphyllus (fig. 32) two thirds the length 

 of the anther. In G. leevigatus (fig. 33) the crest is much smaller ; 

 in G. platystylis (fig. 7) and several others it is small, truncate ; 

 in G. Haspan (fig. 34) the crest is obsolete, only a few of its linear 

 acute papillse sometimes developing. 



The anther-crest is wanting in a majority of species ; but where 

 present it is, from its constancy, a very valuable character in the 

 separation of species. In G. inundatus, Koxb., there is a depressed 

 rudimentary crest exactly as in the closely-allied C. Monti. In 

 G. procerus, Rottb., on the contrary, the anther is absolutely 

 ecristate and similar to that of the allied C. pilosus. 



The crest of the anther also has a subsectional, but no sec- 

 tional or subgeneric value. A small hispid crest is charac- 

 teristic of the whole group Elegantes. C. elatus, Linn., is by its 

 anther-crest closely joined to C. platyphyllus, and G. pannonicus 

 to C. IcBvigatus. There is no clearly cristate anther in any species 

 of the subgenera Pycreus, Mariscus cum Leptostachyo, Diclidium. 

 Nevertheless the species with a highly -developed crest are scat- 

 tered in Juncellus and in remote sections of Eucyperus, so that 

 the character cannot be used as a primary one in the delimita- 

 tion of the larger divisions. 



(9) The Style. 



The style i3 trifid or bifid, or in one species, G. cephalotes, sub- 

 entire. The style is in each bpecies, section, and subgenus of 



