114 



NATUHAL TTTSTOIiY OF PLANTS. 



Pitlospon'cp. Finally, I^entham & Hooker' have recently divided 

 the order into three tribes, DcUmecv, Dilleniea, and Hihhertiece^' 

 based on the form of the anthers. We have elsewhere attempted 

 to show' how this classification, often serviceable in practice, is 

 yet by no means exact, and how the same form of stamen may be 

 observed in genera of any of the three tribes indifferently. Hence 

 we have tried to establish a certain number of series, of which the 

 respective genera have been described above, and which are founded 

 first on the general structure of the gynseceum, and then on that 

 of the androceum. i)///£?;zm is in our eyes the prime centre around 

 which are grouped the genera in which the carpels are more or less 

 united into a plurilocular ovary, while at the same time the stamens 

 are indefinite. In all the other BUleniacece the carpels are inde- 

 pendent of one another, and the unilocular ovaries have a parietal 

 placenta in the inner angle. But among these the stamens 

 may be indefinite as in Hihhertia, or twice as numerous as 

 the petals, or grouped in exactly as many bundles as there are 

 pieces in the perianth, as occurs in Candollea. Thus Hihhertia 

 and Candollea become two other centres or heads of series, usually 

 easy to separate in practice, but between which we should be the first 

 to recognise that there are inevitable points of contact, such as are 

 always found in orders like the one under consideration.^ 



' Genera, 10, 11. 



2 Planchon has reproduced (see de Linden, 

 .3, 4) the opinions of the English authors, and 

 admitted tlieir three principal groups; but he 

 makes a fourth for the genera Wormia, Acro- 

 irema,-A.\\i\. Sc]inmachena,\i\\\c\\, he says, "are 

 more or less abnormal, and do not lit well in any 

 of the divisions." We have shown (Adansonia, vi. 

 27(>) in what this assertion is too absolute, and 

 how closely analogous are Wormia and Dillenia, 

 Schunuwheria and llemisiemma, Acrotrema and 

 Tetracera, at least in flower and fruit. 



3 Adansonia, vi. 269, 278. In several Tetra- 

 ceras and Darillas the same flower contains one- 

 and two-celled extrorse and in trorse anthers. We 

 have liocn shown several flowers of T. sene- 

 ffalenxix, w hicli had introrse anthers to the inferior 

 stamens, while all the superior or innermost 

 stamens had extrorse anthers. In T. ohovafa, 

 the summit of the filament swells into a connec- 

 tive of variable form, sometimes entire, sometimes 

 bifid to a vaiiabloextent ; the cells are then borne 

 on distinct branches (fig. 143). In T. voluhilis, 

 the stamens are all unlike. The connective swells 

 gradually, or suddenly, into an obpyramidal, or 



club-shaped head (fig. 144); the outermost 

 stamens, very short, may be quite sterile. In T. 

 sarmentosa, the filaments are free, or slightly 

 coherent at the base. In Davilla rugosa we 

 have seen introrse and extrorse anthers in the 

 bud. Acrotrema, which is said to have " Sta- 

 mimim Jilamenta hand d 'data," may have parallel 

 marginal anther cells, or a connective swollen 

 into a head like Tetracera (fig. 151), or the 

 anther cells may be porricidal and close together 

 for their whole length (fig. 152). The Jlibhertias, 

 in which the cells are long, narrow, parallel, and 

 close together (fig. 130), may have anthers with 

 short, dilated connectives, and short cells, like 

 those of Tetracera (see fig. 131). Such occurrences 

 have been pointed out by F. Mxjellee {Fragm., 

 ii. 2), especially in H. stellaris, of which the 

 anthers are broader than they are long. 



' Thus we have shown how the HihhertiecB 

 and the Belimece, come through Trisema and 

 Delima respectively to present the same perianth, 

 the same androceum with indefinite elements, 

 and the same gyna;ceum. We have also re- 

 cognised the common links between Acrotrema 

 and Scliumacheria, and the Dilleniem, TetracerecB, 



