GREGORY: FISH SKULLS 97 



Dallia pectoralis as figured by Starks (1926a, p. 204) are perhaps favorable to Dr. AlHs' view, 

 for in this form the proethmoids are large paired elements which extend over the tips of the 

 ascending processes of the premaxillae without, however, any tendency to unite with them. 

 But this fish is assuredly very far off the line of ascent to the percomorphs. Secondly, none 

 of the other teleosts cited by Allis can be considered to be closely related to the direct ances- 

 tors of the percoid and mail-cheeked fishes, while most of them are very widely removed 

 from such relationship. Thirdly, the rostral cartilage now intervenes between the ascend- 

 ing processes of the premaxillae and the dermal plate of the ethmoid. If the ascending 

 processes of the premaxillae are transferred portions of the ethmoid complex, how did the 

 rostral cartilage come to lie between them and the remaining part of the ethmoid, especially 

 if, as Allis rightly concludes (p. 28), "the rostral is quite certainly not a detached portion of 

 the primordial cranium".^ 



Turning now to the more direct evidence, a personal review of the characters of the 

 premaxillae of representatives of most of the group mentioned in the present paper has 

 convinced me that whenever the ascending processes are differentiated from the articular 

 processes they are assuredly a part of the premaxillae and not of the ethmoid complex. Even 

 when the ascending and articular processes appear to be separated from each other at the 

 base, as in the mail-cheeked fishes and in Lophius, a close examination with a pocket-lens 

 will show that there is no suture between the ascending and articular portions of the pre- 

 maxilla but that the apparent separation is due to (a) the different arrangement of the tra- 

 beculae in the two parts, presumably in response to different stresses; and to (b) the occa- 

 sional squeezing together of the ascending and articular processes so that the cleft between 

 them becomes very narrow. 



In Fundulus, which belongs to a group (Microcyprini) that stands more or less on the 

 border between acanthopts and Haplomi, the premaxillae have long ascending processes 

 which have every appearance of being continuous below with the dentigerous part of 

 the bone; posteriorly these tips overlap the mesethmoid (Starks, 1926a, p. 205). 



Percopsis, which above all other living fishes has a claim to be regarded either as a 

 direct descendant of the remote ancestors of the percomorphs or as a descendant of early 

 isospondyls which were progressing in the direction of the percoids, lends no support to 

 Dr. Allis' view that the ascending process was originally separate from the body of the pre- 

 maxilla; for in this fish we see what is apparently an early stage in the differentiation of the 

 ascending and articular processes from each other. Both processes are relatively short and the 

 future cleft between them is represented only by a slight depression; the ascending process 

 is indisputably continuous below with the articular process and with the dentigerous part of 

 the bone, while the whole bone lies well in front of the ethmoid; the future rostral cartilage 

 seems to be represented by the translucent material behind the ascending and articular 

 processes, which has perhaps been secreted by the adjoining articular surfaces of the 

 premaxillae, maxillae, vomer and ethmoids. 



Finally, when we turn to the available palaeontological evidence we find no confirma- 

 tion of Dr. Allis' view. For in the very primitive Cretaceous berycoid Hoplopteryx as 

 figured in detail by Smith Woodward (1902, PI. Ill, Fig. 1), the ascending and articular 

 processes of the premaxillae are both plainly a part of that bone, while the mesethmoid is 

 forked anteriorly much as in modern berycoids and percoids. In Sardinoides of the family 

 Scopelidae, which family Smith Woodward regards as standing near the line of ascent to the 



