[6 7 J THE EVOLUTION OF THE FINS OF FISHES. 1047 



partially homologous with the interspinous bones of norraal forms, for 

 at the proximal cods of the basiradial cartilages there is a constriction 

 at the point where they are continuous with the neural spines proper, 

 which become suddenly larger proximad of the constriction. Moreover, 

 these bars bifurcate at the i)roximal ends, the bifurcation passing over 

 either side of the spinal cord, and the ends finally rest on the upper 

 aspect of the chorda, thus constituting the true neural arches, of which 

 there are evidently several to a single centrum. We therefore have 

 in the young sea-horse neural arches, interspinous and basilar inter- 

 neural elements represented in cartilage as the supports of the dorsal, 

 which are in no sense to be regarded as " rays," but as epaxial vertebral 

 arches and their distal appendages. The intersi)inous portions of the 

 extremities of the successive arches are fused together, as observed by 

 McMurrich in Siphostoma of a similar age, but just proximal to the basi- 

 lar interneural cartilaginous nodules or actinophores there is a general 

 flattening of the cartilage cells and a tendency to indicate that an 

 articulation is in process of formation between the interneural and 

 median actinophoral elements. 



The structure of the distal part of the anal of the young of Hippo- 

 campus is exactly simihir to that of the dorsal. 



It is interesting to note, also, that the neural arches in both Hip- 

 pocampus and Siphostoma correspond in number with the rays, which 

 in other forms are often in excess, together with their homonymous 

 interneural ])ieces, of the number of homonomous neural spines. Here 

 the neural spines are in excess of the number of vertebral centra by 

 about five to each centrum, only about three centra underlying the 

 dorsal of Hippocampus. In Siphost07na, on the other hand, nine verte- 

 bral centra afltbrd arches to support the rays of the dorsal, which num- 

 ber about thirty-five, or an average of about four rays to each centrum. 

 Each vertebral centrum of these Lophobranchiate genera therefore de- 

 velo])s a number of distinct neural arches. Their number seems to be 

 about five to six to each centrum. The vertebras of these types may 

 therefore be said to hepenta- and hexacantlious dorsally. 



Proximally these arches, in Hippocampus and Siphostoma, tend tocon- 

 cresce or be drawn closer together iu an anteroposterior direction. In 

 the first genus all of the neural arches are thus drawn together proxi- 

 mally, the effect of this concrescence being most visible in the oppositely 

 inclined position of the anterior and posterior neural elements, respect- 

 ively. It follows from this that we may possibly be justified in assuming 

 that the dorsal at one time, in some ancestral form of Hippocampus, was 

 supported by more centra than at present. In Siphostouia this tendency 

 toward proximal concrescence is manifested by the group of neural 

 arches belonging to single centra of the series w'hich afford the dorsal 

 flu its support. 



Th i fin rays of both these forms probablj^ originate in mesoblast, as 

 I have shown to be the case in Salmo. What lends additional probii- 



