formed utriculi. These are believed to be connected with the penetration of the pollen-tubes; and their absence in the rostellum probably accounts for its not being penetrated. If the structure of the rostellum and of the stigma is as here described, their only difference consists in the layer of cells which secrete the viscid matter being thicker in the rostellum than in the stigma, and in the utriculi having disappeared from the former. There is therefore no great difficulty in believing that the upper stigma, whilst still in some degree fertile or capable of penetration by the pollen-tubes, might have gradually acquired the power of secreting a larger amount of viscid matter, losing at the same time its capacity for fertilisation; and that insects smeared with this viscid matter removed and transported the pollen-masses in a more and more effective manner to the stigmas of other flowers. In this case an incipient rostellum would have been formed.
In the several tribes, the rostellum presents a marvellous amount of diversity of structure; but most of the differences can be connected without very wide breaks. One of the most striking differences is, that either the whole anterior surface to some depth, or only the internal parts become viscid; and in this latter case the surface retains, as in Orchis, a membranous condition. But these two states graduate into each other so closely, that it is scarcely possible to draw any line of separation between them: thus, in Epipactis, the exterior surface undergoes a vast change from its early cellular condition, for it becomes converted into a highly elastic and tender membrane, which is in itself slightly viscid, and allows the underlying viscid matter readily to exude; yet it acts as a membrane, and its under surface is lined with much
