front of each pollinium, where it comes into contact with the back of the uppermost part of the rostellum. From the number of these threads this middle line looks brown, and each pollen-mass here shows a tendency to divide longitudinally into halves. In all these respects there is a close general resemblance to the pollinia of the Ophreæ.
The line where the parallel threads are the most numerous is the line of greatest strength; elsewhere the pollen-masses are extremely friable, so that large portions can easily be broken off. In the bud-state the rostellum is curved a little backwards, and is pressed against the recently-opened anther; and the above-mentioned slightly projecting bundles of threads become firmly attached to the posterior flap of the membranous cap of the rostellum. The point of attachment lies a little beneath the summit of the pollen-masses; but the exact point is somewhat variable, for I have met with specimens in which the attachment was one-fifth of the length of the pollen-masses from their summits. This variability is so far interesting, as it is a step leading to the structure of the Ophreæ, in which the confluent threads, or caudicles, always spring from the lower ends of the pollen-masses. After the pollinia are firmly attached by their threads to the back of the rostellum, the rostellum bends a little forwards, and this partly draws the pollinia out of the anther-cells. The upper end of the anther consists of a blunt, solid point, not including pollen; this blunt point projects slightly beyond the face of the rostellum, which circumstance, as we shall see, is important.
The flowers stand out (fig. A) almost horiozontally from the stem. The labellum is curiously shaped, as may be seen in the drawings: the distal half, which projects beyond the other petals and forms an excellent
