Seed in the Scitamineae . 
35 
around the endosperm ; while in Strelitzia the extreme 
reduction of this tissue to a useless remnant is realized. 
The members of the Scitamineae present a progressive 
series of stages in the development of the endosperm. In 
the Musaceae this tissue is abundant and starch-bearing, 
though its outer cells may form an aleurone-layer ( Strelitzia ). 
In the Zingiberaceae the endosperm, though several cells 
in thickness in the lower part of the embryo-sac, contains 
only aleurone. In the Cannaceae this tissue is reduced 
to a single aleurone-layer lining the cavity ; and in the 
Marantaceae it is probably not represented in the ripe seed. 
The embryo-sac and embryo are practically straight in 
all of the Scitamineae, except most Marantaceae. In every 
species of the group examined, the entire egg-cell contributes 
directly to the formation of the embryo, without any develop- 
ment of a suspensor. 
Coming now to questions of affinity, we find in the 
characteristic formation of the micropylar collar, in the per- 
sistence of the micropylar epidermis over the apex of the 
embryo-sac, in the very common development of a micro- 
pylar aril, and in the direct development of the embryo, the 
features of the seed which are to be regarded as most charac- 
teristic of the Scitamineae in general. 
The uniformity of their ovules, the campylotropic develop- 
ment of their seeds, and the differentiation of their perisperm- 
canals, support the indications of their floral structure that the 
Marantaceae form a natural and closely related group of 
genera. 
The Cannaceae show no near affinities in their seed- 
development with any other group of Scitamineae. In the 
development of the bulk of the seed from the chalazal 
portion of the ovule, in the coalescence of the micropyle and 
hilus to form the germinal slit, and in the presence of stomata 
in the testa, Canna stands quite alone. Therefore, while we 
fail to obtain light on the origin of these peculiarities or the 
phylogeny of the genus, we are quite justified in separating 
it as the type of a distinct family. 
D 2 
