HIGHLY SPECIALIZED UNISEXUAL SPIKELETS 93 



a very short slender axis, in the axils of broadened 

 sheaths at the summit of short culms and very much 

 overtopped by the blades. If we cut off the base of 

 one of these little heads we have three to five such 

 spikelets, as shown in Fig. 82, B, which had been 

 held rigidly together by the hard thickened base, 

 the overlapping backs of the second glumes forming a 

 thick white wall surmounted by their green-toothed 

 summits. The base can be nothing but a rachis, 

 shortened, broadened, and thickened, though all 

 trace is lost of the junction of the spikelets and the 

 rachis. Fig. 82, D, is a diagram of half a head, show- 

 ing one of the two rows of spikelets. Compare this 

 with Fig. 51, E, showing a great number of spikelets 

 on an elongate rachis. As in Bouteloua gracilis the 

 spikelets stand nearly at right angles to the rachis 

 and the first glume is inward, that is, it would be 

 against the rachis, as in Bennuda-grass, if the spike- 

 lets were appressed. The difficulty the student en- 

 counters in comprehending the pistillate iYiflores- 

 cence of buffalo-grass is not so much due to its 

 complexity as to the difficulty of dissecting the 

 rigid little structure, and also to the suppression 

 or deformity of some of the organs. When spike- 

 lets are closely crowded, as in the Cenchrus bur, 

 in some species of millet, and in buffalo-grass, some 

 of them are nearly always deformed from pressure. 

 In buffalo-grass one or two of the spikelets in a 

 head are not fully developed. The first glume is 

 commonly reduced to a minute scale in two or 



