382 DEVELOPMENT OF ELASMOBRANCH FISHES. 



of large elongated cells with oval nuclei. The remainder of the 

 rudiment forms the commencement of the true nerve. 



The anterior root, which, at the close of stage I, formed a 

 small and inconspicuous prominence from the spinal cord, 

 grows rapidly during the succeeding stages, and soon forms an 

 elongated cellular structure with a wide attachment to the spinal 

 cord (PI. 14, fig. 5). At first it passes obliquely and nearly 

 horizontally outwards, but, before reaching the muscle-plate of 

 its side, takes a bend downwards (PL 14, fig. 7). 



I have not definitely made out when the anterior and pos- 

 terior roots unite, but this may easily be seen to take place 

 before the close of stage K (PI. 12, fig. 18). 



One feature of some interest with reference to the anterior 

 roots, is the fact that they arise not vertically below, but alter- 

 nately with the dorsal roots, a condition which persists in the 

 adult. 



Although I have made some efforts to determine the even- 

 tual fate of the commissure uniting the dorsal roots, these have 

 not hitherto been crowned with success. It grows thinner and 

 thinner, becoming at the same time composed of fibrous pro- 

 toplasm with imbedded nuclei (PI. 14, figs. 8 and 9). By stage 

 M it is so small as to be quite indistinguishable in transverse 

 sections ; and I have failed in stage P to recognize it at all. I 

 can only conclude that it gradually atrophies, and finally 

 vanishes without leaving a trace. Both its appearance and 

 history are very remarkable, and deserve the careful attention 

 of future investigators. 



There can be little doubt that it is some sort of remnant of 

 an ancestral structure in the nervous system ; and it would 

 appear to indicate that the central nervous system must origi- 

 nally have been formed of a median and two lateral strands. 

 At the same time I very much doubt whether it can be brought 

 into relation with the three rows of ganglion-cells (a median and 

 two lateral) which are so frequently present on the ventral side 

 of annelidan nerve-cords. 



My results may be summarised as follozvs: Along the ex- 

 treme dorsal summit of the spinal cord there arises on each 

 side a continuous outgrowth. From each outgrowth processes 

 corresponding in number to the muscle-plates grow downwards. 



