On the structure and classification of the Tremataspidae. 1 1 



identical with Rohon's. These plates arc more convex and robust than the one we have just 

 described and more than twice too large to fit into its place. Moreover, there is no corre- 

 spondence in shape or texture between their anterior margins and the edge of the dorsal 

 shield with which they must have articulated, if they occupied the position assigned to them 

 by Rohon. 



In the most perfect specimens, the anterior margin is thickened, of a different color, 

 and with a dull roughened surface, as though it were attached by means of a tough membrane, 

 or by sinews, to some fixed plate. 



A small scale in the University Collection at Petersburg evidently belonging to Tre- 

 mataspis, is interesting in this connection (PI. I. fig. 7j. It is a triangular plate with its 

 inner concave surface exposed (8 mm. long and 4 mm. wide). It at first appeared to be a 

 median plate of symmetrical form, and possibly fitting into the oral region. A second and 

 better plate, belonging to the Dartmouth collection, has been exposed during the prepara- 

 tion of this manuscript. A careful examination indicates that it is not bilaterally symmet- 

 rical and cannot represent an isolated rostral plate. And indeed it is much too large to fit 

 into the place occupied by that plate. 



It is therefore out of the question to place either of these triangular plates in the me- 

 dian oral region, since neither plate has the size, shape or articulating surfaces necessary to 

 fit into such a place. Moreover if the convex triangular plate belongs in the place assigned 

 to it by Rohon, it must have been articulated to the anterior rim of the dorsal shield, and 

 thus completely closed the place where Rohon supposed the mouth was situated. 



It is clear, therefore, that neither of the two triangular plates that have been found 

 separately can be identified with the median oral plate we have found in situ. If they belong 

 to Trematasjns, and there is little room to doubt that, they must have belonged to some 

 other part of the body, most-likely to a pair cf movable appendages like those of Pterich- 

 thys. These plates will he described more fully in a subsequent publication. 



There is no trace whatever of the large semilunar space figured by Rohon '93 in PI, I. 

 fig. 8, and which he regards as the mouth. If he had worked out more carefully the folded 

 anterior margin of the dorsal shield, he would have hardly felt justified in placing the mouth 

 in this position. It is true that in the cast we are describing, the anterior margin of the 

 ventral shield is not certainly visible, but in front of the anterior row of oral plates is a row of 

 four or five small quadrangular plates that Rohon failed to figure or describe, These plates 

 are either the crushed margin of the dorsal shield, or else a row of independent plates, their 

 regularity in form favoring the latter view. But in any case no space is left here for such 

 an oral opening as Rohon describes. Moreover there is an insuperable objection to placing 

 the mouth in a position where it must always be closed. For the upper jaw being a part of 

 the dorsal shield must be stationary and the supposed lower jaw could not be moved without 

 dislocating the remaining oral plates. 



Where then is the mouth of Trcmataspis? It seems to me that the whole arrangement 



