SECRETION IN THE HYPERICACEEX. 461 
on any one of them is relieved by the absorption of the central 
cells it grows out into the space formed, becoming rounded on 
: its free surface. The growth is, however, but slight, and the 
cell by this process and by the continuance of its secretive 
activity soon loses the rest of its protoplasm, is charged with the 
secretion, and breaks up. So the progress of the absorption of 
the tissue is more rapid than the compensatory growth of the 
cells, and the cavity extends more and more towards the periphery. 
The outermost cells, however, are more vigorous than the others 
between them and the centre, and the last layer never disappears. 
Its cells grow out into the cavity of the gland as the others have 
done on the removal of the pressure; and their activity being 
greater than tbat of those, they not only grow out but divide, 
forming the papille Unger has figured. All the cells of this last 
layer seem to behave in this way. I was not able to trace 
whether the process went further than this, but it seemed from 
some sections not unlikely that the end cells of the hairs broke 
up like the original central cells. Possibly, by this means, a 
continuous process of secretion goes on. These appearances 
occur both in leaf and stem. In the former, fine branches of the 
fibrovascular bundles come into relation with the glands at some 
part or other of their surface, but do not enter them. In the 
stem the glands show no relation to the vascular-bundle system, 
being situated in the cortex. The papille do not grow into the 
glands of the ovary ; and here, too, no relationship between them 
and the vascular-bundle system obtains. 
The dark glands, the walls and contents of whose cells are of 
a violet colour, occur in both floral and foliage leaves of some 
species, while they are absent altogether from others. "They are 
spherical in form in the bracts and sepals of H. hirsutum, sphe- 
rical to ellipsoid in the petals; both spheres and ellipsoids are 
found in the leaves of H. perforatum, the former being perhaps 
the most numerous. On the petals of the same species they are 
very irregular, taking often the form of clavate patches which are 
parallel to the fibrovascular bundles. They are situated gene- 
rally nearer to the under than the upper epidermis of the leaf. 
They differ markedly in one particular from any of the glands 
hitherto described, in that they are always placed in direct rela- 
tionship with the fibrovascular system. Wherever met with in 
the green tissue, a fibrovascular bundle runs to them and gene- 
rally ends abruptly close to the glandular tissue, never, however, 
penetrating it. They thus resemble many other glands occurring 
