ON THE PHYLOGENY OF ORCHIDS. 479 
most Conantherea, are very distinct in the Apostasiee, but the ovary is 
still trilocular and the style slender, with three fertile stigmatic lobes. 
We have only to cancel the three staminodes of Tecophilea in order to 
obtain the diagram of Newwiedia ; if we reduce the anterior stamen to a 
staminode, we have the diagram of Apostasia; and if we suppress it 
entirely, that of Adactylus. 
From a diagram like that of Tecophilea we can also construct that of 
Cypripedie@, by strengthening the symmetrical structure of the perianth, 
and reducing the anterior stamen to a staminode, while the three 
staminodes of Tecophilea are entirely wanting. The American genus 
Phragmipedium still retains the trilocular ovary, while in the geronto- 
geous Paphiopediwm the ovary is trilocular only at the apex, and in 
the northern circumpolar genus Cypripedium the septa have entirely 
disappeared. 
Of the monandrous orchids the Neottie@ come nearest to Conantheree 
and Philydracee. In Thelymitree the perianth is nearly radiate; in 
addition to the fertile stamen there are two very large staminodes, which 
are also present in Diuridez, but in combination with a distinctly 
symmetric perianth. The stamen is quite normal in these plants, the 
anther upright, on a short filament, and only the transformation of the 
“unpair’’ lobe of the stigma into a rostellum gives quite a new character, 
developed as an adaptation to the fertilisation of the flowers by insects. 
When a column is produced, as in Cephalanthera, it has been observed 
that it arises quite late; ina bud of 4 mm. length of C. rubra there is 
no trace of a column to be found, from which it may be concluded that 
the length of the column is not an old or important character. 
I think the contact of the apex of the anther with the top of the 
rostellum, as it occurs in Neottiee and in Cyanella, is an older type than 
the complicated manner in which (in Ophrydee) the pollinia develop 
their caudicles towards the base of the anther. On the other hand, the 
long stigmatic processes of Habenaria and allied genera come nearer to 
the structure of Amaryllidacee than the sessile shining stigmatic surfaces 
of most Neottiee. Perhaps we may infer that the different development 
of the rostellum was the cause why the pollinia applied themselves either 
to its apex or to the base of the anther. 
Besides, in Apostasiee, Cypripediee, Neottiee, and Ophrydee we do 
not find articulated leaves ; yet insome Liparidea, as Malaxis, Microstylis, 
of which the former genus with its upright anther also appears to be an 
old type, the pollinia are waxy without any appendix, while they are 
granular in Apostasiee, Cypripediee, and most Neottiee. The sectile 
pollen of some genera of the latter is very similar to the parcel-divided 
pollinia of Ophrydeew, because we find many pollen masses only in 
plants highly adapted to fertilisation by insects. The granular or sectile 
pollinia would also be a reason for considering the above-named groups 
as less changed by adaptation than the rest of the orchids; we may infer 
that the next step was the formation of caudicles; and the last, the 
existence of a distinct stipes or gland which separates from the rostellum. 
If we compare the morphological structure of the whole plant, we 
find that all orchids with granular or sectile pollen have terminal in- 
florescences, while in the Hpidendree and Vandee of. Bentham some 
