41 



TraciiiLj- the sections of C. levinseni ventriilh' it will be seen that the collar-cavities become 

 separated from one another b\- the diverticulum of the phannx (PI. X, tig. ii8) and begin 

 to pass round the ah'inentar\- canal to the ventral side. The dorsal projection of the metasome 

 [met.) containing the oviducts appears behind the colhir-region, and the two collar-cavities become 

 more and more separated from one another dorsally by the third body-cavities (figs. i 1 8 — - 

 124, ó.i\''). The operculum originates ventro-laterally from the collar (figs. 118 — 122), as has 

 already been described; and the direct continuations of the dorsal parts of the collar-cavities, 

 specially connected, as has been shewn above, with the arm-bases, terminate ventrally in the 

 regions containing the collar-canals (fig. 122, cc). The more anterior portions of the cavities 

 (cf. PI. XII, fig. 151) are, however, continued round the ventral side of the alimentary canal, 

 where they become completely confluent (fig. 122). As soon as the base of the operculum has 

 been passed, a ventral collar-mesentery makes its appearance (fig. 123). The collar-cavities may 

 be seen, in several sections further, on the ventral side of the third body-cavities (fig. 124), 

 the reason for which will at once be apparent from an inspection of the sagittal section (PI. IV, 

 fig. 34). The ventral mesentery persists to the posterior end of the collar-region (PI. IX, fig. 107). 



The relations of the collar-cavities in C. dodecalophtts differ in no important respects 

 from those above described. This will be clear from an inspection of figs. 151 — 146 (see p. 34); 

 but attention may specially be called to fig. 151, which shews that the collar-canal occupies 

 the ventral end of the part of the cavity which encircles the alimentary canal, and is a direct 

 continuation of the cavity of the arm-base. 



The same arrangement is shewn, for C. gracilis, in the series of obliquely sagittal sections, 

 PI. V, figs. 43 — 51, and in the frontal sections, PI. \T, figs. 63 — 68; and, for neuter individuals 

 of C. sibogae, in PI. XIII, figs. 182—185 Oöu"g). ?!• VIII, fig. 93 and PI. VII, figs. -]■], 78. 



Collar- pores or Collar-canals. 



There are no organs in Cephalodiscus which have given me more trouble than the 

 collar-canals, and this in consequence of the difficulty of forming a reasonable theory as to their 

 mode of action. The general facts are perfectly clear. Each collar-pore or collar-canal is an 

 ovoid body which lies with its long axis more or less transverse to that of the body (PI. XI, 

 fig. 131, £•. r.). On the median side, the canal opens by a large aperture into the collar-cavity. 

 On the outer side it opens by an even larger aperture [c. c e') to the exterior. It follows from 

 this disposition of the organ that both apertures may be cut in a single section which jjasses 

 either in a transverse plane (fig. 131) or in a frontal plane (PI. X, fig. 121); and that a section 

 which passes transversely to the long axis of the organ must have a more or less sagittal 

 direction (PI. XI, fig. 140J. But with varying states of contraction of the collar, these relations 

 ma)- be somewhat distorted. 



The collar-canal lies at the ventral end of that section of the collar-cavity which is 

 directly continuous wMth the arm-base (PI. XII, fig. 151 ; PI. V, figs. 46 — 49; PI. XI, fig. 140). 

 Its ventral wall is in immediate contact with the dorsal wall of the gill-slit (figs. 46, 131, 140), 

 but it rests partly on the septum which divides the collar-cavity from the trunk-cavity (figs. 49, 

 140, 155J. This wall is composed ot a high epithelium with numerous nuclei, im[)lying narrow, 



smoGA-EXPEDITIE XXVI bis. 6 



