ventral flange of the pterotic in lineatus has a goodly con- 

 tribution from the exoccipital but in mbu next to none. 

 Tetraodon mbu is unique among the tetraodontids 

 studied in completely lacking a pterosphenoid. The eth- 

 moid-vomer-palatine region is essentially similar in both 

 lineatus and mbu, and the prefrontals differ mainly in 

 that of lineatus having an anteromedial arm helping to 

 enclose the olfactory foramen. The skull shape of Ephip- 

 pion is not far removed from that of Tetraodon and both 

 probably have a close common ancestral group, perhaps 

 as close as that which unites Tetraodon with Monotreta 

 and Chelonodon. 



The only two species usually assigned to Chelonodon 

 differ slightly more than do the two of Tetraodon ex- 

 amined. In C. fluviatilis, which has a deeply pitted ol- 

 factory epithelium, the frontals abruptly end anteriorly 

 in a more or less straight transverse line at the rear edge 

 of the prefrontals, the latter being relatively thin and en- 

 tirely separated by the ethmoid. In C. patoca the frontals 

 become narrower anteriorly over the orbit, but they do 

 not stop abruptly at the level of the posterior edge of the 

 prefrontals, continuing anteriorly as rapidly tapering 

 points between the prefrontals, the latter being relatively 

 thick and separated by both the ethmoid and frontals. 

 The sphenotics in patoca are less anterodorsally ex- 

 tended than in fluviatilis. The skull condition in patoca 

 is slightly more generalized than that of fluviatilis, and 

 patoca retains a supraneural which is lost by fluviatilis. 

 Both species lack any remnants of the dorsal roof of the 

 myodome, and there is no interhyal and only one hypo- 

 hyal. The parasphenoid has a dorsal flange in the orbit 

 meeting the frontals and there are minute teeth on the 

 first pharyngobranchial, while fluviatilis has numerous 

 trituration teeth to either side of the midline in both the 

 upper and lower jaws, and patoca has only two or three 

 teeth in each series in the upper jaw and none in the 

 lower jaw. The vertebrae are modally 18 in fluviatilis and 

 19 in patoca. 



On the basis of the differences in the olfactory epithe- 

 lium, and of fluviatilis having the lobes of the olfactory 

 apparatus oriented parallel to the body versus at a right 

 angle to it in patoca, Le Danois (1959) generically 

 separated fluviatilis (as Dichotomy cterus) from patoca 

 (Chelonodon). The validity of such minor nasal dif- 

 ferences alone in distinguishing genera is highly ques- 

 tionable, but should other species assignable to 

 Chelonodon other than the commonly collected 

 fluviatilis and patoca come to light (a few valid species 

 probably lurk among the numerous synonyms usually 

 listed for each of these names) and fall into one or the 

 other of the fluviatilis or patoca skull plans and not into 

 intermediate types, there may be justification for es- 

 tablishing two genera for what is now Chelonodon. 



The precise relationship of Chelonodon to Tetraodon 

 and Monotreta is not clear on the basis of the present 

 data. 



As previously discussed, Chonerhinos and Xenopterus 

 are highly specialized tetraodontids which have second- 

 arily increased the number of vertebrae and of dorsal and 

 anal fin rays, elaborated the lateral line system until 



there are approximately three lines on the body, in- 

 creased the size and amount of folding of the olfactory 

 epithelium in the open cup nasal apparatus, and in- 

 creased the size of at least some of the spines, mostly 

 those of the belly. The greater increase in numbers of 

 vertebrae and dorsal and anal fin rays in Xenopterus in- 

 dicates that it is the more specialized of the two, and this 

 is also borne out in the structure of the skull. While both 

 species are unique among the tetraodontids in the loss of 

 the prefrontals (Fraser-Brunner 1943 said that the pre- 

 frontals are very small, but I find no trace of them at all), 

 the skull of Chonerhinos otherwise is not markedly dif- 

 ferent from that of many of the species of the Mono- 

 treta-Chelonodon-Tetraodon group. The frontals of 

 Chonerhinos are slightly wider over the middle of the or- 

 bit than more posteriorly and are tapered gradually and 

 evenly to gently rounded points anteriorly that broadly 

 overlie the relatively broad ethmoid, which is of 

 moderate length, while the sphenotics are well extended 

 anteroventrally, projecting out beyond the edges of the 

 frontals and forming about the rear half of the upper edge 

 of the orbit. 



In Xenopterus the frontals are much more laterally ex- 

 panded and thickened than in Chonerhinos, forming a 

 large plate over most of the dorsal surface of the skull. In 

 the two smaller specimens studied the frontals are nor- 

 mally articulated by interdigitation in the midline, but 

 in the largest specimen the two frontals are indis- 

 tinguishably fused to one another in about the middle 

 third of their lengths. While the sphenotic oi Xenopterus 

 is about as anterolaterally extended as in Chonerhinos, it 

 is nearly entirely overlain by the frontal and only its ex- 

 treme distal end appears in dorsal view, projecting 

 slightly beyond the edge of the frontal in about the mid- 

 dle of the upper edge of the orbit. The main body of the 

 supraoccipital in Xenopterus is less wide than in Chone- 

 rhinos, but the supraoccipital crest is wider and heavier 

 in the former than in the latter. In Xenopterus the neural 

 and haemal spines of the penultimate vertebra become 

 hyperostotic, but this does not occur in Chonerhinos. 



The ethmoid in Xenopterus is shorter but broader than 

 in Chonerhinos and in both genera the anterodorsal end 

 of the vomer tends to fully fuse with the anteroventral 

 end of the ethmoid. Both genera are also similar in hav- 

 ing the pterosphenoids in contact with the parasphenoid 

 in the rear of the orbit, the pterosphenoid, para- 

 sphenoid, and prootic forming a more elaborate and mas- 

 sive structure there than in any other tetraodontids. This 

 is perhaps functionally similar to the bracing strut of a 

 dorsal flange from the parasphenoid in the middle of the 

 orbit to the under surface of the frontals as found in 

 many other tetraodontids, but not in Chonerhinos and 

 Xenopterus. The only other tetraodontids in which the 

 pterosphenoid meets the parasphenoid are several 

 species of Lagocephalus, but this is associated with the 

 dorsal flange of the parasphenoid in the middle of the or- 

 bit and thus is not analogous to that in Chonerhinos and 

 Xenopterus. In Xenopterus the supraneural is larger and 

 deeper bodied than in Chonerhinos. Both genera lack any 

 remnants of the dorsal roof of the myodome, there is no 



