CLASSIFICATION OF MONOCOTYLEDONS 237 



nales and Liliales. The Liliales are midway in the series, hav- 

 ing attained a petaloideous perianth and entoraophily, and 

 having become so diversified in structure and habit as to raise 

 a question as to their monophyletic origin. The Arnaryllida- 

 ceae introduce epigyny, and the highly specialized Iridaceae 

 complete the series. The last six families are in great need of 

 morphological investigation in the tropics where they are chiefly 

 massed. 



The genetic connection between Liliales and Farinales 

 seems clear, so that if the latter are regarded as derived from 

 the Glumales, the former must be referred to the same stock, 

 probably dissociating early from the Farinales. 



The two remaining alliances are characterized by epigyny 

 and zygomorphy, highly specialized entomophilous structures, 

 reduction and modification of stamens, and very small and un- 

 differentiated embryos. In all probability they are not genet- 

 ically related, but they resemble one another more than they do 

 the other alliances. 



IX. SCITAMIXALES.* The four families of this alliance 

 are Musaceae, Zingiberaceae, Cannaceae, and Marantaceae, to- 

 gether comprising nearly 800 species, 500 belonging to the 

 Zingiberaceae. The four families are undoubtedly genetically 

 related, although the first two are restricted to the oriental 

 tropics, and the last two to the occidental. In addition to the 

 characters mentioned above, the replacing of functional sta- 

 mens by petaloid staminodia is very characteristic, commonly 

 only one stamen being functional and even this one being peta- 

 loid. In nearly every case, also, there is a labelluni, formed 

 either by the perianth or the staminodia. The habit of the vege- 

 tative body, however, is most peculiar. The real stem is a rhi- 

 zome, but the enormous leaves, differentiated into sheath, peti- 

 ole, and pinnately veined blade, build up a false stem by means 

 of their very large and closely overlapping sheaths. 



The temptation is to derive this alliance from the Dracaena 

 region of the Liliaceae, but important anatomical features that 

 are common to all four families are opposed to this view. That 

 it i.s connected in some way with the Glumales-Farinales-Lili- 

 ales stock seems most probable ; and if so the general structures 



* SCITAMINEAE of Engler. 



