had they been developed. Not only do they stand in this position, but the column in some cases, as in Cephalanthera, has on each side a prominent ridge, running from them to the bases or mid-ribs of the two upper petals; that is, in the proper position of the filaments of these two stamens. It is, again, impossible to doubt that the two membranes of the clinandrum in Malaxis are formed by these two anthers in a rudimentary and modified condition. Now, from the perfect clinandrum of Malaxis, through that of Spiranthes, Goodyera, Epipactis latifolia, and E. palustris (see fig. 16, p. 101, and fig. 15, p. 94), to the minute and slightly flattened auricles in the genus Orchis, a perfect gradation can be traced. Hence I conclude that these auricles are doubly rudimentary; that is they are rudiments of the membranous sides of the clinandrum, these membranes themselves being rudiments of the two anthers so often referred to. The absence of spiral vessels running to the auricles is by no means sufficient to overthrow the views here advocated as to the much disputed nature of these structures; that such vessels may quite disappear, we have proof in Cephalanthera grandiflora, in which the rostellum and its vessels are completely aborted.
Finally, then, with respect to the six stamens which ought to be represented in every Orchid: the three belonging to the outer whorl are always present, the upper one being fertile (except in Cypripedium), and the two lower ones invariably petaloid and forming part of the labellum. The three stamens of the inner whorl are less plainly developed, especially the lower one, a3, which, when it can be detected, serves only to strengthen the column, and, in some rare cases, according to Brown, forms a separate projection or filament; the two upper anthers of this inner whorl are fertile
