28 ME. G. BENTHAM ON GBAMINEjE. 



The consequence is that there are a considerable number of 

 sjiecies in which the grain has been described by some as ad- 

 herent and by others as free, and which have consequently been 

 transferred from one geniis to another. Yet, if not taken too 

 absolutely, the character is sometimes a useful one, assisting, for 

 instance, in the arrangement of the genera of some of the sub- 

 tribes of the difficult tribe Pestucea?. 



Considerable importance was attached by the earlier agrosto- 

 logists to the presence or absence of the awn on the back or 

 apex of the flowering glume ; but this has subsequently been 

 found to be subject to great variations. The spiral twist, how- 

 ever, in the lower part of the awn in some genera is more con- 

 stant, and in the ' Flora Australiensis ' I had taken it as an essential 

 character of some tribes or subtribes ; but there are more ex- 

 ceptions than I was then aware of. The awn, when present, 

 is generally twisted in Audropogoneee, Tristegiueae, Agrostideae, 

 and Aveuacese, and not in Pauiceae, Chloridea;, Festuceffi, or 

 HordesB ; but it is sometimes very slightly so in a few species of 

 the latter group, and in the former tribes, where the awn is much 

 reduced, if there be any twist it is scarcely perceptible. In all 

 the tribes, also, the awn is occasionally, and in the straight- 

 awned ones frequently, altogether deficient ; and in some genera, 

 as in Sfipa for instance, where it is usually twisted, there are 

 exceptional species in which it is straight or curved only. The 

 character must therefore generally be used with more or less 

 of reservation. 



The partial or absolute separation of the sexes or the increase 

 in the number of stamens observed in a few genera have been 

 occasionally introduced amongst tribual characters ; but further 

 observation has shown that they occur amongst Q-ramine£e of 

 very different affinities, and have thus proved to be often of no 

 more than generic value, although in one tribe, the Maydese, the 

 absolute unisexuality of the spikelets may be constant, 



Differences in the size of the embryo, in the form of the so- 

 called scutellum on the caryopsis (indicative, apparently, of the 

 hilum of the seed), or in the longitudinal groove or cavity fre- 

 quently observable on the caryopsis, have been sometimes brought 

 forward as absolute generic, if not tribual, characters, and they 

 may often be really important; but we know, as yet, too little 

 about them to test their value fairly. Herbarium specimens 

 rarely supply ripe fruits, and they have been carefully observed 



