VIVIPAROUS FISHES OF THE PACIFIC COAST. 
463 
migrated in through the gill-slit. In the ciliated gullet the cells are 22/u high. On 
the roof this tract extends farther caudad than on the floor, where it does not extend 
beyond the gill thickenings. The walls of the mid gut are again of a low epithelial 
nature, while the cells lining the hind gut are of quite another nature. They are high 
and the small nucleus is situated near the base of the cells, where the contents are 
slightly granular. The center of the cells is a large unstainable space very variable 
in size. The free ends of the cells are again granular. Some of these cells contain 
deeply stained bodies similar to those found in the lumen, and it seems very probable 
that these bodies have been swallowed by the cell and are in process of reduction by 
intracellular digestion. At the anus the epithelium of the roof of the intestine is 
continuous with that of the floor of the combined segmental ducts. The cells of the 
floor of the intestine are continuous with those of the ventral surface. The segmental 
duct empties just behind the anus and not into a cloaca or iuto the intestine. 
In an earlier stage it was noticed that the mouth was first indicated just behind 
the eye. The conditions obtaining in larvae of this stage are shown in figs. 167 to 171 
(3-2 mm.); the mouth is essentially like the other gills, especially the spiracular, 
and agrees in all major points with the condition described by Dohrn. That is, the 
mouth is further developed laterally than medially and some distance behind the 
point where it attains its full development. In the larvae 1mm. long I was unable to 
trace the hypoblast cells much beyond the notochord. In other words the alimentary 
tract begins in the hyobranchial region in those larvae. It soon extends forward in 
the median line and, as far as I could determine, the outgrowth of hypoblast to form 
the hyomandibular slit takes place later than that to form the hyobranchial. In larvae 
1-8 mm. long the hypoblast extends outward to the ectoderm just behind the eye. 
This I have identified as the first indication of the future mouth (figs. 154, 181). In 
these larvae the hypoblast does not yet extend to the anterior end, and the mouth is a 
strictly bilateral structure. The hyomandibular evagination is separated from the 
mouth evagination by a more restricted region of hypoblast (fig. 153). This evagina- 
tion does not differ materially from the mouth evagination. In each case the ectoderm 
is two layers thick where the hypoblast touches it. But one of the gill evaginations 
has been completed and the second is in process of formation (figs. 150 and 151). The 
details of the completion of the mouth have not been traced. It is not functional 
when the larvte are 4 mm. long and in fact the lumen does not extend forward any 
farther than in the 3 mm. larvae. 
The mouth is completed shortly after the larvae have reached 4 mm. This late- 
ness of the appearance of the mouth seems to me to be one of the most remarkable 
circumstances connected with the development of the alimentary tract. Thus 
while the intestine becomes functional when the larvae have reached a length of 1 
mm. the mouth is not formed till they are over 4 mm. long — not in fact till the liver 
has long been functional, the air-bladder well developed — not till all the glands derived 
from the hypoblast are well developed. Not only is the mouth late in appearing but 
the whole canal from the gill-cavity forward is also late in forming. While this may 
be due to retardation, since ingress to the canal is had through the first gill-cleft, the 
conditions impress one with the suggestion of Dohrn that the present mouth of verte- 
brates is not the original mouth, but is of comparatively late origin. How one struc- 
ture may replace another as a mouth is well illustrated by Cymatogaster , where a new 
structure, the hyobranchial gill-cleft, functions as a mouth for a long time. It would 
