pedium longifolium (Cypripedioideae) (P1. UX, fig. 8) and 
Lecanorchis javanica (Neottioideae) (Pl. LX, fig. 4), how- 
ever, may possibly represent independent steps in the 
reduction process. Theoretically, the placentae in Phrag- 
mipedium longifolium are parietal in origin because each 
of the two placental lamellae, although situated in adja- 
cent locules, are vascularized by a common strand (note 
the connecting dotted arrow). The placentation itself is 
intermediate between the axile and parietal positions. In 
Lecanorchis javanica, we find another type of modifica- 
tion of axile placentation. The septa are broken down 
into separate placental lamellae, and the torus is com- 
pletely eliminated. The individual lamellae facing the 
adjacent ventral bundles are united by the margin, thus 
leaving a Y-shaped empty cavity in the center of the 
ovary, and the ovules are borne in two rows at the point 
of junction of the lamellae. 
These intermediary steps suggest that the reduction 
from a tricarpellate to a monocarpellate condition might 
have come about by a longitudinal division of the septa 
which eliminated the torus, followed by a gradual short- 
ening of the lamellae (Limodorum abortivum, Pl. LX, 
fig. 5), until merely traces are found along the inner wall 
(Cephalanthera alba, Pl. IX, fig. 6). 
Embryogeny. No other plant family exhibits such an 
inconsistent embryogeny as the Orchidaceae. Only the 
first and second cell generations of the zygote are con- 
sistent; the subsequent divisions apparently take place 
in a random manner. 
In the first cell generation (Pl. X, fig. 1), the zygote 
divides into a basal and terminal cell. During the second 
cell generation, the basal cell differentiates into a suspen- 
sor initial cell and middle cell, while the terminal cell 
divides by a vertical wall. From this step onward, the 
further divisions are without any definite sequence, but 
[ 66 ] 
