384 BOTANICAL GAZETTE [NOVEMBER 
time it somewhat resembles the embryo of Sparganium simplex as 
described by CAMPBELL (3); but without the younger stages of S. 
simplex and more of the young stages of NV. advena (?) it is impos- 
sible to say whether this resemblance is more than superficial. From 
the large end of this conical embryo a cotyledonary ridge is now pro- 
duced which extends almost entirely around this end and almost com- 
pletely encloses the plumule (fig. 17), while in N. advena of the north 
it extends only a little more than half the distance around the embryo. 
This monocotyledonous character is very evident in embryos dis- 
sected out of the sacs. Two cotyledonary lobes are next developed 
from this cotyledonary ridge, thus giving it the dicotyledonous char- 
acter (jig. 18). The development of this dicotyledonous character 
is much earlier and much more pronounced than in N. advena of 
the north. A number of embryos were cross sectioned and examined 
very carefully and considerable variation was found in the prominence 
of this character. In one case it was so great as to give the appear- 
ance of two equal cotyledons (fig. 19). The development is strik- 
ingly similar to that of Nelumbo as described by Lyon (18) and 
York (28), 
SCHAFFNER (26) dissected the advanced embryos of VV __ advena 
out of the sacs and clearly demonstrated the formation of the two 
cotyledonary lobes. He did not contradict my conclusions, as stated 
by Morrrer (20), but made his studies from older embryos than 
I was able to secure at the time my studies were made. 
The embryos of the two species of Castalia show some differences 
and also differ from Castalia odorata, In Castalia ampla a Pro 
embryo is formed which may consist of as many as six cells in linear 
arrangement (figs. 20, 21). The terminal cell then divides by.* 
longitudinal wall, which is followed by a similar division in the next 
cell (fig. 21). The four cells thus formed then divide by a second 
longitudinal wall at right angles to the first. By repeated division, 
this mass of cells now forms a spherical embryo supported by 2 sue 
pensor of four or five cells in linear arrangement (fig. 22). One we 
more of these suspensor cells, usually the basal, may divide longi 
tudinally. On account of the seed pods sinking soon after fertiliza- 
tion it was impossible to follow the development in this species further. 
In Castalia pubescens the embryo develops in the same manner, 
