We thus see that, comparing the flower of Graminece with the normal liliaceous plan (fig. 5), it differs in the complete suppression of the outer row and the posterior member of the inner row of the perianth-leaves, of the whole inner row of stamens, and of the anterior carpel, whilst the remain-
Fig. 5—Diagrams of the ordinary grass-flower. 1, actual condition; 2, theoretical, with the suppressed organs supplied. a, axis; b, flowering glume; c, palea; d, outer row of perianth leaves; e, inner row; f, outer row of stamens; g, inner row; h, pistil.
ing members of the perianth are in a rudimentary condition. But each or any of the usually missing organs are to be found normally in different genera, or as occasional de velopments.
Fruit.—The ovary ripens into a usually small ovoid or rounded fruit, which is entirely occupied by the single large seed, from which it is not to be distinguished, tin thin pericarp being completely united to its surface. To this peculiar fruit the term caryopsis has been applied (more familiarly "grain"); it is commonly furrowed longitudinally down one side (usually the inner, but in CW-rand its allies, the outer), and an additional covering is not unfrequently provided by the adherence of the persistent palea, or even also of the flowering glume (" chaff " of cereals). From this type are a few deviations ; thus in Sporobolus, &c. (fig.
Fig. 6—Fruit of Sporobolus, showing the dehiscent pericarp and seed.
6), tli3 pericarp is not unite 1 with the seed but is quite distinct, dehisces, and allows the loose seed to escape. Sometimes the pericarp is mem branous, sometimes hard and brittle, whilst, on the other hand, in some genera of JDambusece, it becomes thick and fleshy, forming a " berry," or rather a drupe. In Meloccuma baccifera this of forms a fruit 3 or 4 inches long, with a pointed beak of 2 inches more ; it is indehiscent, and the small seed germinates whilst the fruit is still attached to the tree, putting out a tuft of roots and a shoot, and not falling till the latter is 6 inches long. A similar germina tion also occurs in Pseudostachyum compactum, which has the largest fruit of the order.
Seed.—The testa is thin and membranous but occasion ally coloured, and the embryo small, the great bulk of the seed being occupied by the hard farinaceous endosperm (albumen) on which the nutritive value of the grain depends.
Fin. 7. A grain of wheat. l,back, and 2, front view; 3, vertical section, showing ( /) the endosperm, and () embryo; 4, commencement of germination, showing (6) the pileola and (c) the secondary rootlets surrounded by their coleorrhizas. The embryo presents many points of interest. Its position is remarkable, closely applied to the surface of the endo sperm at the base of its outer side. This character is abso lute for the whole order, and effectually separates Graminece from Cyperacece. The part in contact with the endosperm is flat and plate-like, and is known as the sciitellum. Some difference of opinion is held on the nature of this ; but it is probably correctly regarded (as by Van Tieghem) as the main portion of the cotyledon, the white epigoeal sheath (pileola) protecting the plumule which is often described as the cotyledon being the ligule only of that organ greatly developed. The radicle is inferior, broad, and blunt ; the primary root is very slightly developed in germination, but several secondary lateral ones burst through above its point, and thus become surrounded at their base with little sheaths (coleorrhizae).
II. Classification.—Gramineæ are thus sharply defined from all other plants, and there are no genera as to which it is possible to feel a doubt whether they should be referred to it or not. The only order closely allied is Cyperacece^ and the points of difference between the two have been alluded to above, but may be here brought together. The best distinctions are found in the position of the embryo in relation to the endo spermlateral in grasses, basal in Cyperacece and in the possession by Graminece of the 2-nerved palea below each flower. Less absolute characters, but generally trust worthy and more easily observed, are the feathery stigmas, the always distichous arrangement of the glumes, the usual absence of more general bracts in the inflorescence, the split leaf-sheaths, and the hollow, cylindrical, jointed culms, some or all of which are wanting in all Cyperacece. The same characters will distinguish grasses from the other glumiferous orders, Restiacece, Enocaulonece, and Desvaiixi- (icece, which are besides further removed by their capsular fruit and pendulous ovules. To other monocotyledonous families the resemblances are merely of adaptive or vege tative characters. Some Commelyuacece and Marantaccce approach grasses in foliage; the leaves of Allium, &c., possess a ligule; the habit of some palms reminds one of the bamboos ; and Juncacece and a few Liliacece possess an inconspicuous scarious perianth. The great uniformity among the very numerous species of this vast family renders its classification very difficult. The difficulty has been increased by the confusion resulting from the multiplication of genera founded on slight characters, and from the description of identical plants under several different genera in consequence of their wide distribution. No characters for main divisions can be obtained from the flower proper or fruit ; though both Kunth and Eeich- enbach have used them especially the form of the styles and stigmas, thelodicules and the caryopsis they have not been generally employed by botanists, who have found it necessary to trust to characters derived from the usually less important inflorescence and bracts. The earlier authors made the general arrangement of the spikelets (spicate, paniculate, &c.) the basis of their classi fications. Palisot de Beauvois s main divisions are founded on the existence in the same inflorescence of similar or diverse spikelets. Fries has proposed a division into Eury- nnthece and Clisantkece, according to the condition of the flowering glume and pale, separated or close during inflor escence. Dumortier gives a classification into Rachidece, Rachilliftorce, and Calliflorce ; the first with the spikelets inserted into excavations of the rachis, the second with the flowers of the free spikelet inserted on its axis (rachillus), and the last with the flower or flowers borne on the short callus of the glumes. Both these systems, and others which have been suggested, possess merits of their own, but they have not as yet been found capable of application to the order as a whole, having been framed for the accommodation mainly of European genera. For such general treatment no better primary divisions have been found than those proposed by Robert Brown
in 1810 and further elaborated in 1814, which have
