GREGORY: FISH SKULLS 93 



are covered by a system of numerous small plates. These included, according to Stensio 

 (1921, p. 134), the following series: (a) immediately above the premaxilla, a median and 

 a lateral transverse pair of "rostrals"; (b) above (a) two median interrostrals one above 

 the other; (c) above (b) two successive pairs of postrostrals, the more posterior pair articu- 

 lating with the frontals; (d) on each side a longitudinal series of three "nasals" carrying 

 the nasal branch of the supraorbital canal; (e) on each side, two successive antorbitals, 

 between the nostril and the orbit. Thus there were some twenty plates above the rostrum. 

 In the oldest actinopteran fishes the rostral system as described by Watson (1925) was less 

 elaborate than that of the crossopterygian Dictyonosteus but still included one pair of rostrals 

 immediately above the premaxillas, two successive median postrostrals, one pair of large 

 antorbitals lateral to the postrostrals. 



In the teleosts this whole area is taken up by the so-called ethmoid, including the 

 originally cartilaginous olfactory capsule, which is divided into an anterior or ethmoid 

 region and paired lateral regions (variously called prefrontals, ectethmoids, lateral ethmoids 

 or parethmoids). 



Proethmoids are paired bones of uncertain origin covering the anterior corners of the 

 lateral ethmoids in certain fishes; the "mesethmoid" is a median bone articulating in front 

 with the premaxillse and behind with the frontals. Besides these there are usually the 

 paired nasals, sometimes paired dermal prefrontals and adnasals. 



Thus the tracing of the homologies of the surface elements of the rostral and pre- 

 orbital regions through the various families of chondrostean, lepidosteoid and amioid 

 ganoids up into the lower teleosts is a difficult and hazardous undertaking due to various 

 transformations, enlargements and shiftings in the different families. According to Watson 

 (1925), the antorbital (or adnasal) of Amia, which bears a triradiate canal at the meeting- 

 place of the suborbital and supraorbital canals, must be the homologue of the lacrymal of 

 palseoniscids (the probable homologue of the tetrapod lacrymal), which bears a similar tri- 

 radiate canal. But the adnasal or antorbital of Amia lies wholly in front of the circum- 

 orbital series and is closely associated with the anterior nares, the premaxilla and the nasal 

 bones, while the lacrymal of palaeoniscids lies well behind these elements and forms the 

 lower anterior corner of the orbit, like the lower preorbital (our lacrymal) of Amia. More- 

 over, Watson figures a triradiate canal in Coccocephalus (Fig. 12C) in a bone at the lower 

 anterior quarter of the orbit and this bone has every appearance of being homologous on 

 the one hand with the lacrymal of Cheirolepis and on the other with the bone here called 

 lacrymal in Amia (Fig. 28). 



Repeated consideration of the literature of the preorbital and of specimens of early 

 tetrapods, rhipidists, palaeoniscoids, semlonotids, eugnathids, Lepidosteus, Amia, etc., lead 

 to the following conclusions: 



(1) That the so-called mesethmoid (dermethmoid) of teleosts may be traced backward 

 to one of the median rostrals (shown in Amia, Lepidosteus, Eugnathus and Cheirolepis) that 

 originally lay immediately above the premaxillae. 



(2) That the anterior preorbital of teleosts is correctly homologized with Watson's 

 lacrymal of Cheirolepis and of tetrapods. 



(3) That the antorbital or adnasal of Amia (which is absent in teleosts) may be traced 

 back to the antorbital of Eugnathus and of semionotids; this is postero-lateral to the nasals 

 and to the nares. In Perleideus of the palseoniscoids the antorbital seems to have absorbed 

 the lacrymal or driven it out. In Cheirolepis both are present. 



