GREGORY: FISH SKULLS 367 



articular processes of the premaxillae above and the widened vomer below. Thus the 

 consideration of these adaptive or growth changes tends to make intelligible the details 

 of the ethmoid region, which appear so dry and meaningless when considered as detached 

 from their surrounding parts. The derivation of this mesethmoid-prefrontal fossa itself 

 becomes intelligible if we compare the skull of Uranoscopus with that of Ophiocephalus. 

 In the latter fish this fossa is seen in its initial phase as a gentle concavity between the 

 anterior forks of the frontals, its function being to accommodate the ascending processes of 

 the premaxillae. 



The skull roof of Uranoscopus scaher has been carefully studied by Starks (1923), who 

 shows that its broad parietals are beginning to overlap the supraoccipital. In my specimen 

 of Uranoscopus sp. from Japan this process has gone much further, so that the parietals 

 meet in the mid-line above and appear to separate the supraoccipital widely from the 

 frontals (see also Fig. 2455). A better case of an apparent but not real exception to DoUo's 

 "Law of the Irreversibility of Evolution" could hardly be wished for, since the parietals 

 have not regained their original contact along the mid-line but only a secondary one. 



As a whole, the skull of Uranoscopus closely resembles that of Astroscopus as described 

 below, except that it is somewhat less specialized and less simplified secondarily. Thus 

 the bones of the posterior part of the skull-roof maintain their sutural contacts, which may 

 be seen plainly on the sculptured upper surface, while in Astroscopus sutural lines are at 

 best but vaguely suggested. Among additional striking differences from Astroscopus are 

 the following: 



The posttemporal is a very short thick bone with a sculptured dorsal plate. It lacks 

 the long slender inner fork seen in Astroscopus. The supracleithrum bears a long sharp 

 spike directed upward and backward. The preopercular is a broad curved plate with a high 

 and flat, vertically concave anterior border, which bears in all eight radially diverging proc- 

 esses, not extending beyond the web-like outer border but standing in high relief above it; 

 the lowermost three processes end in sharp, horny, downwardly-directed thorns. Obscure 

 traces of this complex armature may be detected in the somewhat degenerate preopercular 

 of Astroscopus (Fig. 247). 



A similar downwardly-directed thorn is borne on the lower end of the subopercular. 

 This spiniferous tendency is also expressed in the presence of pairs of thorns elsewhere, i.e. 

 on the pelvic bones just in front of the pelvic fins, on the postero-superior angles of the small 

 supracleithra, on the antero-inferior projection of the lacrymal. 



Several features in the skull of Uranoscopus are rather suggestive of relationship with 

 the scorpsenoid fishes; the third suborbital is produced backward and downward (although 

 it fails to reach the preopercular border) ; the relations of the radiating ridges and spikes on 

 the preopercular recall the conditions in the scorpaenoids, and so does the presence of spikes 

 on the supracleithra, the peculiar beaded surface of the skull", the radiating trabeculae and 

 zonal growth lines. The pectoral pterygials are more specialized than those of the more 

 primitive scorpaenoids but do not appear to exclude derivation from the scorpaenoid stem. 

 Nevertheless the derivation of the uranoscopoid series from relatives of Trachinus and 

 Notothenia appears far more probable. 



Astroscopus. — An advanced stage of specialization of the family Uranoscopidae is 

 represented by the skull of Astroscopus y-gracum. In this very strange fish (Figs. 247, 248) 

 the eyes are shifted to the dorsal surface; parts of the eye-muscles are greatly hypertrophied 



