GEOLOGICAL SURVEY IN SILURIAN ROCKS OF SOUTH OF SCOTLAND. 847 



in unyielding bony plates. In fact, the shark-like Ccelolepidse seem to have specialised 

 themselves into a form, which in former days would certainly have been called a verit- 

 able " Placoderm." 



But if Drepanaspis points backwards to the Ccelolepidse, it also points forwards to the 

 Pteraspida?, but the consideration of this question I shall defer till we come to treat of 

 the last-named group itself. But before doing so we must look into the question of the 

 affinities of the Psammosteidse, a family upon whose hitherto somewhat doubtful position 

 the structure of Drepanaspis throws an unexpected light. 



The Psammosteidse. 



The plates known as Psammosteus (Agassiz), which occur usually in a very fragmentary 

 condition in the Devonian (Old Red Sandstone) rocks of Great Britain and Russia, have 

 long been a puzzle to palaeontologists. By Agassiz, to whom only small fragments were 

 known, Psammosteus was classed as a "Coelacanth" (ii. p. 61); by Trautschold, 

 plates apparently belonging to the same genus were interpreted as swimming paddles of 

 Coccosteus (xxxvii., PI. VI., PI. VII., fig. 2) ; but the most prevalent opinion at present 

 is that these remains are elasmobranch in their nature, and belonged to some extinct 

 group of " armoured sharks." 



The remains of Psammosteus consist in the first place of large oblong plates, gently 

 hollowed in boat-like fashion, obtusely pointed at one extremity (anterior), and truncated 

 or obtusely notched at the other. Internally these plates are smooth, externally they 

 are covered with minute closely-set tubercles, which are brilliantly ganoid and have 

 beautifully crimped edges. In many instances these tubercles are arranged in polygonal 

 areas, which in worn specimens are often removed, leaving shallow polygonal depres- 

 sions behind, so as to give the surface of the plate something of a honeycombed 

 appearance. 



The inner layer of these plates is formed by a dense laminated substance perforated 

 by vessels ; the middle one is thicker, and shows a close network of vascular canals, the 

 intermediate substance displaying numerous minute tubules, so that, as Agassiz already 

 remarked, it appears more related to dentine than to bone. The outer layer consists of 

 the tubercles themselves, which show a radiating arrangement of dentine tubules pre- 

 cisely similar to those figured by Rohon in the scales of Thelodus (xxviii. p. 33), while 

 the external resemblance of these tubercles to certain Thelodus scales, especially to 

 those of Thelodus Pagei from the Forfarshire Old Red (xxxvi. fig. 2), is obvious enough. 



It seems, therefore, pretty clear that, as I have already remarked in my Extinct 

 Vertebrata of the Moray Firth Area (xxxiv. p. 262), the stellate tubercles of Psam- 

 mosteus are shagreen-granules which have coalesced, and have also become united to a 

 plate formed in a deeper layer of the skin. Here an analogy seems to be afforded 

 by the condition of the dermal covering in Ateleaspis, on the anterior part of which 

 we have also reason to believe that minute scales have run together into polygonal 



