Tort] SHARP—PHYSOSTEGIA 227 
walls (fig. 12), and further similar divisions give rise to a large- 
celled, thin-walled tissue which fills the endosperm lobe (fig. 13). 
This endosperm formation may cease abruptly at the narrow portion 
of the sac (fig. 14), but usually extends for a little distance into 
the micropylar lobe (fig. 16). The two-ranked arrangement so 
conspicuous in the endosperm lobe in fig. 13 and in the micropylar 
lobe in fig. 16 is doubtless due to the longitudinal separation of the 
embryo sac into two parts as described above. 
The cessation of endosperm formation at an iideinite point 
results in nuclei being left free in the cytoplasm of the micropylar 
portion of the sac (fig. 13). These nuclei, usually two in number, 
enlarge (fig. 14) and may occasionally divide, the walls which 
appear on the spindle fibers being evanescent. Often the nuclei 
were observed fusing. Consequently, from one to at least four 
may be present in stages somewhat later, but they play no further 
active part, and disorganize with the other contents of the micro- 
pylar lobe (figs. 18 and 19). 
In embryo sacs which show a wall at the first division of the 
endosperm nucleus it is usual for the sac to be thereby separated 
transversely into two chambers, and for endosperm to be formed 
only in the micropylar one. Among such cases the endosperm 
may pass through a free nuclear stage, as in Sagittaria (SCHAFFNER 
12), Limnocharis (HALL 5), and Ruppia (MuRBECK 11); or walls 
may be formed at all of the divisions, as in Ceratophyllum (StrRas- 
BURGER 13) and the Nymphaeaceae (Cook 1 and 2). Less fre- 
quently both daughter nuclei resulting from the division of the 
endosperm nucleus take equal parts in the direct formation of 
cellular endosperm, as reported for Peperomia pellucida (JoHN- 
SON 8), Heckeria (JOHNSON g), and Datura laevis (GUIGNARD 4). 
From the above account it is seen that essentially this is the mode 
of endosperm formation in @/ysostegia, and in this sac the main 
point of interest lies in the fact that the first wall is longitudinal 
rather than transverse. The factors governing the orientation 
of the spindle and the consequent position of the wall are not at 
all clear, and the feature is probably best regarded as a minor 
peculiarity rather than a character of much significance. 
The restriction of endosperm to the antipodal portion of the 
