The Genus Phacelurus (Gramineae)

Abstract

Amongst the awnless Andropogoneae belonging to the subtribe Rottboelliinae Presl and enumerated by Clayton (1973), there is a small group of eight allied species which have wandered among no less than nine different genera. Two of these genera, Andropogon L. and Ischaemum L., do not belong to the Rottboelliinae (Clayton, 1972). Of those which do, Rottboellia L. is now limited to species with a fused pedicel and internode; Coelorhachis Brongn. is characterized mainly by its single racemes borne in a spathate false panicle; and Vossia Wall. & Griff. is distinguished by its caudate lower glumes. Having excluded these possibilities, there remain four available genera, Phacelurus, Pseudophacelurus, Pseudovossia and Thyrsia. The erection of these genera is not surprising for there are sufficient obvious differential characters among the eight species to invite subdivision. What is not so apparent is that these characters are only weakly correlated with one another, and that generic diagnoses lean heavily upon the selective weighting of single characters. The species are geographically scattered, so that piecemeal treatment has sufficed for regional Floras, and there has been no occasion for an overall comparison between every pair of species. When this is done, however, it becomes apparent that the customary generic treatment is unsatisfactory, being flawed by inconsistencies which have crept into the weighting of characters. In contrast to the traditional selective weighting, it is instructive to compare the species by a method which accords equal weight to the eight most useful characters (listed in the Appendix). An appropriate method employs the simple matching coefficient followed by average link clustering. It is hardly worth using a computer, for a small matrix can be quickly worked on a desk calculator. For those unfamiliar with the method, a suitable algorithm is explained in detail by Sokal & Sneath (1963). The resulting dendrogram is shown in Fig. i. It comes very close to an intuitive assessment of relationships, which are summarized in the dichotomous key given below. The dendrogram suggests a division into four entities along the 6o per cent phenon line. These are at variance with any of the earlier generic treatments, but indicate a more natural division into little groups of very closely related species. The question of their rank hinges upon whether homogeneity is thought to be the primary generic criterion, in the sense that character variation within each genus should be demonstrably less than that between adjacent genera; or whether recognition should be given to an overall similarity of spikelet structure, viewed within the context of neighbouring parts of the classificatory system. In my view the former course leads to the creation of taxa which might be more appropriately termed superspecies. I have not therefore attempted yet another generic