28 ULLETIN : MUSEUM OF COMPARATIVE ZOOLOGY. 



two 1 And if two, what is theit mode of uniou 1 "We have no hesita- 

 tion iu answering that two median plates are present, and that the an- 

 terior overlaps the posterior, as in Coccosteus. The evidence appears 

 perfectly decisive, and is of twofold nature ; it depends upon a prom- 

 inent surface elevation over the very region where we should expect the 

 boundary between two median ventrals to be, and upon the fact that 

 two centres of ossification are discernible. 



The surface elevation referred to is palpably of the same nature as 

 those prominences which are formed by the hinder extremities of the 

 antero-ventro-laterals where they are superimposed upon the posterior 

 pair of ventro-laterals. All of these elevations are more or less eroded 

 iu the specimen, but the one under consideration is scarcely more so 

 than the others. If it were a purely fortuitous bulge of the surface, we 

 should expect similar ones to occur elsewhere, whereas the prevailing 

 aspect of the plates is flat and smooth. But inasmuch as the only re- 

 maining elevations are found at those places where we know for certain 

 that boundaries occur, and as this occurs at the only place in the median 

 line where we should expect to find a boundary, we are compelled to 

 look upon this as a significant, not an accidental feature. Moreover, 

 the shape of the elevation corresponds with the tapering extremity of 

 the antero-ventro-mediau, superimposed upon the posterior element ; 

 and the outline of the latter is seen to be perfectly normal as compared 

 with homologous plates, when we cut it oflF at this point. In fact, it is 

 noteworthy that the shape of the postero-ventro-median bears a marked 

 similarity to the bone last described (supra, p. 26), and shown in 

 Plate 5, Fig. 1. 



But still more pertinent evidence as to the existence of tw^o median 

 ventrals is furnished by the structure of the plates themselves. It is 

 apparent at a glance that in the postero-ventro-median ossification pro- 

 ceeded from a single centre, which was nearly coincident with the centre 

 of the plate itself. On holding the slab so as to reflect light at a proper 

 auffle, the course of vascular (Haversian) canals can be seen very dis- 

 tinctly, especialh^ at the right anterior boundary ; and all of these i-adiate 

 toward the centre of the plate. Vascular canals are likewise apparent 

 on the antero-ventro-laterals, but are only faintly perceptible on the 

 antero-ventro-median. If the latter plate were articulated or fused with 

 the posterior element, as in D. terrelli, it would be difficult to account 

 for the significant elevation already referred to ; and considering the 

 relative thinness of the plates, such a mode of union could hardly have 

 proved advantageous. It is more natural to suppose that the connection 



