Trinius, in his elaborate monograph of the tribe, divided it into three primary groups or subtribes-Vilfea with the callus scarcely prominent or quite obsolete, Agrostem with the callus globular, and Stipeæ with the callus obconical. In this I feel unable to follow him. In the first place he does not appear to have considered what the so-called callus really is. It is not, as the name would suggest, an appendage to the base of the flowering glume, or, as he would have termed it, to the flower, but only the upper or principal part of the rhachilla or axis of the spikelet, to which the glume and its enclosed flower are attached, and which breaks off immediately above the persistent empty glumes. Its shape depends on the distance at which the flowering glume is attached above the empty ones, a distance very variable throughout the Order. And although the long or the short interval may be more prevalent or even constant in some genera, yet I have never found the variations so precise as to be defined by actual measurement, and the species are numerous, even in Stipa itself, where it is doubtful whether we should call it long or short. It is sometimes a useful accessory character, but, I believe, never positive enough to be regarded as subtribual. It is true that no other simple absolute character has yet been proposed for the subdivision of the tribe; but we are obliged, here as elsewhere, to take a combination of characters, to each of which an occasional exception must be allowed. Acting on this principle, we might, whilst following in many respects the arrangements of Kunth and others, admit thirty-seven genera of Agrosteæ, distributed in four fairly natural subtribes, all four of wide geographical range, but chiefly in temperate regions, the tropical species mostly confined to mountain districts, and no genus, except a few monotypic ones, exclusively tropical.

Our first subtribe, STIPEE, is the long-established one of that name, slightly extended so as to include Oryzopsis, Muehlenbergia, and their immediate allies, the close connexion of which with Stipa has been frequently suggested. The subtribe thus formed would be characterized by the paniculate inflorescence not condensed into the cylindrical spike of Phleoideæ, by the rhachilla of the spikelet not produced beyond the flower except in the single species of Brachyelytrum, by the awn of the flowering glume terminal, not dorsal as in Euagrosteæ, and especially by the grain being very closely enveloped in the fruiting glume. In the majority of species these characters are well marked; but

in the larger genera there occur occasional exceptions more or less decided, which prevent our taking any single one of them as an absolute test. The subtribe would include the following eight genera, in the first five of which the fruiting glume is more or less hardened or rigid as in Panice; in the succeeding three it is thinner, though still closely pressed on the grain.

1. ARISTIDA, Linn., is now a genus of at least a hundred species, abundant in all the warmer regions of the globe, but also represented by a few species in Europe and temperate Asia, and by several in North America. With few exceptions it is most readily recognized by the long, fine, three-branched awns, the lateral branches opposite and spreading. Doell adds to the generic character three lodicules as in Stipa; and Nees describes three lodicules in some South-African species; but all other Agrostologists describe two only, and I have never found more than that number. It is probable that both Nees and Doell mistook for the third lodicule the palea, which in many species is very thin and scarcely, if at all, larger than the lodicules. The genus is divided into three fairly marked sections, which Beauvois, Nees, and some others have raised to the rank of genera. In (1) Chataria, Beauv., the flowering glume is continuous with the awn without any articulation, and though much longer than the empty glumes, and often much attenuated at the end, is neither quite awn-like nor decidedly twisted below the branches. Amongst its species, Curtopogon was proposed as a genus by Beauvois for the North-American A. dichotoma, Mich., in which the lateral branches of the awn, instead of diverging from the central one, are short and erect at its sides, showing more or less distinctly that they are continuations of the lateral nerves of the glume. It is probable that this is the case throughout the genus, only that the lateral nerves before they diverge are so closely consolidated with the central one as to be undistinguishable from it. The genus Ortachne was proposed by Nees for two or three Mexican or Columbian plants, originally published by Kunth as species of Streptachne, Br., and afterwards transferred by him to Aristida, in which the lateral branches of the awn are very short, sometimes minute or even quite obsolete, thus nearly connecting the section Chataria of Aristida with the section Aristella of Stipa, but in the narrow base of the rhachilla, and some other minor points, nearer to the former than to the latter. Ortachne retorta, Nees (in Steud. Gram.), is probably a true Stipa. In

(2) Arthratherum, Nees, the awn is decidedly articulate on the glume and much twisted above the articulation below the branches, the flowering glume itself much shorter than the lower empty glumes, instead of exceeding them as in Chataria. In (3) Stipagrostis, Nees, the awn is articulate on the glume, as in Arthratherum, but scarcely twisted, and above the branches elegantly plumose, the branches also being plumose in some species; whilst in others, forming Figari and De Notaris's proposed genus Schistachne, the central awn alone is plumose, the lateral branches short and glabrous. All, however, are most conveniently included in the great genus Aristida.

2. STIPA, Linn., is almost as numerous and as widely spread as Aristida. It is also strongly characterized, as to the great majority of species, by the narrow, rather hard fruiting glume, carrying off a rather long or obconical internode of the rhachilla (or so-called callus), by the long undivided awn more or less articulate on the glume and usually twisted at the base, and by the presence of three lodicules; but the exceptions to one or more of these characters are more numerous than in Aristida; the internode of the rhachilla varies much in length and in shape, the articulation and twist of the awn gradually disappear in some species, and the third lodicule, though often as large as the others, is sometimes much smaller or even quite obsolete. The genus is also not so clearly divisible into sections as Aristida, although several genera have been proposed for more or less aberrant species. Macrochloa, Kunth, includes S. tenacissima, Linu., and S. arenaria, Brot., both from the Mediterranean region, remarkable for their large membranous glumes, the flowering one shortly bifid at the apex. In Aristella, Bertol., founded on S. aristella, Linn., a European and Mediterranean species, in Streptachne, Br., a single Australian species, and in Orthoraphium, Nees, two or perhaps three East-Indian species, the flowering glume is 2-toothed or shortly bifid at the apex, the awn scarcely or not at all articulate, and the internode of the rhachilla very short, though still perhaps slightly thickened under the flowering glume. The S. aristella, however, is very closely connected with typical Stipe through S. sibirica, Lam., S. Redowskii, Trin., and S. altaica, Ledeb. Jarava, Ruiz and Pav., was founded on S. jarava, Kunth (S. eriostachys, Cav., S. papposa, Nees), a widely-spread WestAmerican species, to which the small spikelets in a long narrow dense panicle, with the flowering glumes crowned under the awn

by a pappus-like ring of long hairs, give a very peculiar aspect; but precisely similar flowering glumes are observable in several South-American species with very various habits. In the European S. pennata, Linn., and a few other American as well as Old-World species, the awn itself is (almost entirely, or for a short distance above the base) plumose with long spreading hairs. Lasiagrostis, Link (Achnatherum, Beauv.), was proposed as a genus for the European S. Calamagrostis, Wahlenb., and extended by Nees and Trinius to several African and Asiatic species, only differing from other small-flowered Stipe in the flowering glume itself being plumose with spreading hairs, either below the middle. or in its whole length; and in S. mongholica, Trin., forming the genus Ptilagrostis of Grisebach, these hairs extend to halfway up the awn. S. verticillata, Nees, from Australia, and Apera arundinacea, Hook. f., from New Zealand, two plants closely resembling each other, though specifically distinct, connect Stipa with Muchlenbergia. They have the inflorescence and small spikelets of the latter genus; and in S. verticillata the awn is generally persistent, though the articulation is distinctly traceable on the flowering glume; in S. arundinacea the awn is very deciduous; in this species there is usually but one stamen, whilst in S. verticillata there are the normal three. S. rariflora (Muehlenbergia rariflora, Hook. f.), from Antarctic America, is another species closely allied to the above two; and all three appear to be better placed under Stipa than under Muehlenbergia.

3. ORYZOPSIS, Mich. (Urachne, Trin.), is a genus of about fourand-twenty species, from the temperate and subtropical regions of the northern hemisphere or from extratropical South America, very rare within the tropics, most of them often regarded as awned species of Milium, but really more nearly connected with Stipa, from which they chiefly differ in the broader fruiting glume, often oblique at the top, the awn usually short, slender, and twisted, and very deciduous. The genus divides readily into three sections, regarded by some as distinct genera, but all united into one by Trinius and others. 1. Piptatherum, Beauv., comprises the Old-World species, often included in Milium as a section, with awned glumes, and really connecting in some measure the two genera. The obliquity of the fruiting glume is much less marked than in the typical species of Oryzopsis; and the rhachilla of the spikelet is glabrous. 2. Euoryzopsis or Oryzopsis proper, including the proposed genera Caryochloa,

Spreng., Piptochatium, Presl, and Nassella, E. Desv., is entirely American, with the typical character of the genus, and the rhachilla bearing a ring of hairs under the flowering glume. 3. Eriocoma, Nutt. (Fendleria, Steud.), differs from Euoryzopsis only in the long silky hairs clothing the fruiting glume.

4. MILIUM, Linn., was formerly extended to several unawned Panicea with only two empty glumes, but is now reduced to five or six European or temperate Asiatic species, one of which is also spread over North America, all removed from Panicaceæ as having the empty glumes persistent below the articulation. They differ from Oryzopsis chiefly in their obtuse absolutely unawned flowering glume. 5. ACIACHNE, Benth., is a single dwarf tufted dicecious grass from the higher mountains of Peru and Colombia. The female individual, with only one spikelet terminal on the peduncle, is fully described and figured in the last part of Hooker's 'Icones.' The male plant, if correctly matched, of which I am by no means certain, has a loose almost simple panicle with precisely the glumes of the female, but enclosing stamens only. In the few specimens seen the leaves are much longer than in the numerous females from various localities, which makes me rather doubt the specific identity of the two.

6. MUEHLENBERGIA, Schreb., has nearly sixty known species, chiefly American, extending from the Andes of South America over the northern continent generally, with a very few from central or eastern Asia. They connect, in many respects, Stipa with Agrostis. In general they come very near in technical character to the smaller-flowered Stipa, differing in the still smaller spikelets with thinner though still closely appressed and narrow fruiting glumes, and usually with a more or less hairy rhachilla. From Agrostis and its immediate allies they may be readily distinguished by this narrow appressed fruiting glume with a terminal never dorsal awn; a very few unawned species are scarcely separable from Epicampes, except by the shape of the glume. There is a considerable variety in the inflorescence and in the proportions of the glumes, but nothing definite enough to establish good sections, although several separate genera have been proposed. In the original M. diffusa, Schreb., and its immediate allies, the panicle is usually long, narrow, and dense, and the lower empty glumes are very minute; whilst in Trinius's proposed section Acroxis both the lower glumes or one only of them are nearly as large as the flowering one; but throughout the