TAXONOMY OR CLASSIFICATION 157 



economic value of included species. Keys are given to all 

 genera native or commonly cultivated in the United 

 States. 



198. Characters used in classification. — The con- 

 sensus of botanical opinion is that genetic relationships 

 among phanerogams are best shown by the structure of 

 the flowers. Grasses are no exception to this rule and 

 hence the classification is based upon the structure of the 

 spikelets. The preceding key indicates the characters 

 used in classifying the tribes. The classification appears 

 to be somewhat artificial, but nevertheless it brings to- 

 gether in the same tribe genera that are evidently related. 

 But it also in some cases, separates into different tribes 

 genera that are closely related. Reference will be made 

 to some of these cases again in the appropriate place. 



199. Phylogeny. — As to the phylogeny of the grasses, 

 it is probable that the most primitive existing forms are 

 those in which the spikelet consists of a series of flowers 

 in the axils of herbaceous bracts. The simpler genera of 

 Bambusese, such as Arundinaria, probably represent the 

 lower or more primitive forms. It must not be understood 

 that this tribe is, as a whole, less developed than the other 

 tribes. Some genera are highly developed. There is good 

 ground for believing, however, that the Bambusese arose 

 from forms more primitive than those that gave rise to 

 the other tribes. The Festucese and Hordese probably come 

 next in phylogenetic development, while the Andro- 

 pogoneae and Panicese are highly developed. The exact 

 relationship of the various tribes and the smaller groups 

 is, of course, for the present largely a matter of conjecture 

 and individual opinion. Phylogenetic ideas are expressed 

 by the grouping of forms rather than by attempting to 

 trace lines of descent. We may group allied species into 



