Par nets sia and its bearing on the Affinities of the Genus. 503 
the tanniferous cells are elongated and vermiform, and originate, according 
to Engler, by the fusion of a row of elements. They are thus scarcely 
comparable with those in Parnassia 1 and Chrysosplenium, which consist of 
individual epidermal cells, placed either singly or in groups. In these two 
genera we are not dealing with the occurrence of a definite secretory organ, 
but merely with the presence of tannin in certain epidermal cells which are 
otherwise normal. This is not a morphological character, but probably an 
expression of some peculiarity of metabolism, perhaps correlated with the 
habitat. It seems highly unlikely that such a character would be a trust- 
worthy clue to relationship. This doubt is strengthened by the fact that 
the leaf-epidermis of Chrysosplenium is markedly different in its general 
features from that of Parnassia. In the former, hairs are present, and the 
small stomata are placed in groups in connexion with subsidiary cells 
which are absent in Parnassia. 
We are thus brought to the conclusion that the case for the Saxi- 
fragaceous affinity of Parnassia gains little support from a consideration of 
the tanniferous cells. 
Having considered one or two points in which the value of the 
arguments for the relationship of Parnassia to the Saxifragaceae appears 
to have been over-estimated, we may now turn to the theory that this 
genus has an affinity with the Hypericineae. Here, on the contrary, the 
weight of favourable evidence seems to have been unduly minimized by 
many writers. For this reason it may be well to compare the essential 
organs of the flower in Parnassia and in Hypericum in some detail (cf. 
Text-figs. 3 and 4). 
In Hypericum there are either five or three carpels and in Parnassia 
generally four, though in more than one species flowers with five carpels are 
quite common. 1 2 The main difference in structure between the gynaecea of 
the two genera is that the stigmas of Parnassia are continuous with the 
placentas, whereas the styles and stigmas of Hypericum are placed between 
the placentas or septa (cf. Text-fig. 3, A and C, and Text-fig. 4, B and d). 
The dehiscence also differs, being loculicidal in Parnassia and septicidal in 
Hypericum. 
The placentation of Parnassia was described by Lindley, 3 in 1846, as 
‘ truly axile ’. Hooker and Thomson, 4 on the other hand, wrote : ‘The 
placentation in all the species is decidedly parietal, as in Droseraceae ; nor 
have I, in the earliest-examined stages, detected any evidence of this being 
a deviation from the axile type.’ It will be seen from the accompanying 
diagrams (Text-fig. 3), which represent a set of serial sections through the 
1 Thouvenin, M. : 1 . c., p. 39, PI. 10, Fig. 2. 
2 Seemann, B. : The Botany of the Voyage of H.M.S. Herala, 1852-7, p. 25, and Pace, L. 
1 . c., p. 306. 3 Lindley, J. : The Vegetable Kingdom, p. 406, London, 1846. 
4 Hooker, J. D., and Thomson, T. : 1 . c., p. 78. 
f 
