THE CERATOPHYLLINAL DIVISION. 



29 



The question as to whether the placentation in Parnassia is dorsal, is one which bears with some weight on the affinities of the genus. Some time since, while examining some flowers of 

 Hypericum elodes, I was so struck with the close resemblance of the bundles of stamens to the fringed scales of Parnassia, that I felt confident that if I could meet with the flower of a Parnassia 

 with five carpels, I could then, by comparing it with H. elodes, come to a conclusion (although not perhaps to an unquestionable demonstration), as to whether the placentation was sutural or dorsal. 

 The difference between the structure of Parnassia and H. elodes (the placentation being equally parietal or very nearly so in both) is, that the latter has three carpels, instead of four or five ; that 

 the 5 fringed scales, which in Parnassia are barren, are three bundles of monadelphous stamens ; and that the five single stamens of Parnassia, which alternate with the fringed scales and 

 petals, are large pale glands, which being reduced to three are thrown out of their normal positions relatively to the petals. The cause of this difference is, that in Parnassia the flower is 

 pentamerous, the corolla, stamens, and fringed scales alternating with each other, but when the carpels are reduced to four, which is so generally the case, the stamens and fringed scales, instead 

 of becoming four each, continue to be five, by which their normal relation to the carpels is destroyed. But it is not so with H. elodes, for the five monadelphous bundles (each of which consists of 

 three stamens) become reduced to three alternating with the three carpels ; — but to obtain this two bundles are not suppressed, but unite with two others forming two double bundles, so that one 

 of the bundles (which is always opposite a petal) consists of only three stamens, while the other two consist of six, or by a deficiency of one of only five. By this junction two of the glands 

 which are between them are suppressed, and the remaining three are always opposite the three carpels* Supposing then these glands to represent the stamens of Parnassia, it would follow that, if 

 in a flower of the latter, with five carpels, the five stamens were opposite the placentas and stigmas, the placentation (the structure of the flower being the same) would be dorsal. 



Having recently examined a flower of Parnassia with five carpels, for which I am indebted to the liberality of the Rev. W. W. Newbould, I do not find from this comparison that the placen- 

 tation can be regarded as dorsal ; the carpels were rather unequal in size, but one of the placentas, which was the smallest and less likely to be displaced, was exactly opposite one of the fringed 

 scales, thus agreeing with H. elodes in being opposite the bundles of stamens; and the stamens I concluded to be alternate with the placentas, as are the glands of H. elodes. In most of the 

 instances of dorsal placentation the walls of the ovary are thickened, but in Parnassia they are very thin. P. nubicola is one of the more remarkable instances of stigmas opposite placentas, as the 

 style is slender and elongated, being nearly as long as the ovary and the stigmas are linear and spreading. 



Parnassia may be compared with Argophyllum, in which the ovary is half inferior, and the barren stamens numerous and filamentous, and it agrees with Ixerba in its loculicidal 

 dehiscence. It differs however from Brexiacece in the absence of albumen, (but the quantity in the latter is very small), and in the stamens being turned outwards, but otherwise they nearly 

 resemble them. Droseracece have been referred to as one of the nearest affinities of Parnassia, and doubtless the fimbriated stipules of P. tenella (Linn. Proc. vol. II., p. 80,) show a near 

 approach to Drosera ; and the placentas opposite the stigmas in Parnassia is a coincidence in structure with FrancoacecB, a near ally of that family. Parnassia may be further compared with 

 Hypericaeece, in the leaves of the creeping stem being opposite in pairs, which when young are connate at their bases, and in dyeing a strong clear yellow, and, if it were not for the loculicidal 

 dehiscence, it might serve to connect Brexiacece with that family. 



OCHNACEiE. 



Sauvagesia and the genera associated with it are so closely allied to Ochnacece, through Luxemburgia and allied genera, agreeing with them in their septicidal dehiscence, remarkable fimbriated 

 stipules and habit, that they may be stationed as a section of that family. They also nearly approach Hypericaeece and Droseracece, Sauvagesia (at least in one species) agreeing with Hypericum in 

 the younger stems being flattened, perhaps more so than in the two-edged species of that genus, and in the septicidal dehiscence of the fruit ; and with Drosera in its intra-axillary fimbriated stipules 

 and small embryo, which is I believe less than half the length of the albumen. In Sauvagesia the stipules extend in a reduced form into the axil, so as to join together, and Drosera has 

 this character more obviously, so that it seems not improbable that the fimbriated leaves of the latter may be accounted for by a comparison with this kind of stipules. The placentae, although 

 parietal, project considerably inwards, as much or more so than in Hypericum elodes, so that the walls of the ovary may be cut away without breaking them through. In their sensible and medical 

 properties they agree with Ochnacec®, being mucilaginous and somewhat astringent. 



* That this is the mode of formation of the flower of IT. elodes I have no doubt, because in two instances I have met with four carpels, and then there were four bundles of stamens, three of them consisting of only three stamens each, and 

 although there were only three glands, this appeared sufficiently accounted for by two of the bundles being scarcely separated, so that the space between them was insufficient to admit of the production of a gland. 



