FOSSIL AND RECENT 191 



of Nybelin op. cit., producing the lateral rostral ossicle) and the antorbital (producing 

 an additional ossicle). Although there may be a prenasal ossicle incorporated in 

 the dorsal part of the lateral rostral ossicle (Nybelin 1967a : fig. 2), there is never any 

 connection between the supraorbital and ethmoid commissure sensory canals. 



Megalopids differ little from the Elops condition, the most significant difference 

 being the loss of the anterior fragment of the antorbital. According to Gosline 

 (1961 : 22) the loss of this ossicle has been accompanied by the loss of a small section 

 of the canal and thus the ethmoid commissure fails to connect with the infraorbital 

 system. Nybelin (1967a), however, finds a connection in the soft tissue of Megalops 

 and young Tarpon. Following Nybelin's technique of ink injection I find a connec- 

 tion in Megalops and Tarpon (the largest Tarpon examined in this respect was of 

 222 mm S.L.). A difference from the Elops condition is the very close relationship 

 between the lateral rostral ossicle and the premaxilla. In young Megalops and 

 Tarpon the ossicle is clear of the premaxilla (Nybelin 1967a : fig. 2), but in older speci- 

 mens of these genera the rostral ossicle comes to lie in a shallow excavation on the 

 lateral surface of the premaxilla, and is very tightly bound to that bone. 



While reduction in the number of canal bearing bones may be a feature of the 

 megalopids, multiplication is characteristic of the albuloids. Thus Pterothrissus 

 has three ossicles, although it is impossible to be sure which are dermethmoid deriva- 

 tives and which are antorbital in origin. It is suspected that the fossil albuloids 

 had more than three ossicles since the development of the premaxillary canal was 

 probably mediated by a morphological stage such as is seen in Recent halosaurs 

 (Notacanthif ormes) . 



The development of the snout canal system of albuloids is hypothesized to have 

 taken the following course. From a basal albuloid such as Osmeroides (which has 

 an ethmoid commissure very much like that of Elops, other ossicles being unknown), 

 the anterior end of the dermethmoid turned ventrally through 90 ° taking the rostral 

 ossicles with it. The rostral ossicles subsequently fragmented into several smaller 

 ossicles, some of which lay in the skin over the premaxilla. Such a stage may be 

 seen in the halosaurs Halosaurus and Halosauropsis, where there are four to seven 

 ossicles forming a chain across the snout, each ossicle apparently associated with a 

 neuromast (Gosline 1961). In albulids and pterothrissids those ossicles overlying 

 the premaxilla became closely associated with that bone (as in Tarpon) and eventually 

 fused with it, producing the condition seen in the albulids and pterothrissids where 

 the sensory canal appears to pass through the premaxilla. The premaxillary canal 

 is supplied by a branch of the buccal nerve (indicating its relationship with the 

 infraorbital canal) which pierces the premaxilla close to the point of articulation of 

 that bone with the ethmoid. The ethmoid commissure may still be continuous in 

 some halosaurs (Gosline 1961), but it is broken in the mid-line in Recent albuloids, 

 presumably by the loss of a median portion. It appears that in this assemblage 

 (albuloids and halosaurs) the ethmoid commissure has shifted its association from 

 the dermethmoid to the premaxillae. 



The retention of the originally continuous ethmoid commissure-infraorbital canal 

 system in the Elopiformes is probably a consequence of the solid snout of these 

 fishes. In other groups, notably clupeiforms, salmoniforms and cypriniforms, the 



