1768 Bulletin //, United States National Museum. 



differ in size (which is usually the case), the supraocular is invariably 

 weaker than the postocular. The depression between the tympanic and 

 postocular is always deep, while between the postocular and supraocular 

 there is frequently a well-marked ridge {chlorostictiis, rliodochloris, ruberri- 

 mus). In leris the true supraocular is usually present; in the skull at 

 hand it was absent, but on one side a blunt knob occupied the position 

 required by the rule of relative distances, and just behind this point, on 

 both sides, there was a depression in the otherwise continuous ridge, 

 marking the depression between the supra- and post-oculars. In the skull 

 of elonyatus, in which 1 of the pairs of spines is normally absent, there 

 is a low, conical, rudimentary spine on the left side, occui^ying the posi- 

 tion of the supraocular, as required by the rule of relative distances. 

 These facts, taken together, seem to establish the conclusion that when 

 1 of the trio of pairs of spines is absent, the supraocular spine has dis- 

 appeared, and the supraocular ridge merged with the postocular. 



Of the changes that take place with increasing age, the following are 

 among the most constant: The bones of the skull grow thicker and in 

 very large specimens become spongy. The processes of the mesethmoid 

 become depressed; and the ventral process of the basisphenoid, when 

 present at all, sometimes suffers complete, and always partial absorption. 

 The interorbital space grows relatively wider, this being one of the most 

 striking and constant variations. The width of this space is always given 

 as measured at its narrowest part (which usually falls immediately behind 

 the preocular spines), and compared with the total length of the base of 

 the skull. In a young vexillans, the ratio of interorbital width into the 

 length of the base of the skull is 5^, in a medium-sized one 4f , and in a 

 large one 4. In a young maliger it is 4|-, in an old one 4^ ; in a young 

 miniatus 3i, in an old one Sn ; in a jonng flavidus 3i^j, in an old one 3. In 

 a very young ruljerrimus it is Of, in one 2 or 3 times as large 5f, in one in 

 which the cranial ridges are almost completely serrated 5, and in a very 

 large old specimen i^ . 



It will be seen from the key given below that in several parts of the 

 group closely related species have the parietals in contact; hut while this 

 serves well as a character of subordinate importance, the mere fact that 

 any 2 species have parietals which meet or overlap is no i>roof of affinity 

 unless it is supported by other agreements. The most reliable cranial 

 characters for the purpose of classification of the species are: The degree 

 of curvature of the base of the skull; the convexity or concavity of the 

 interorbital space and its relative width ; the direction of the mesethmoid 

 processes ; the degree of development of the ventral process of the basi- 

 sphenoid, and the strength or weakness of the cranial ridges. These char- 

 acters are closely correlated, and furnish the only basis for the arrange- 

 ment of the species within the genus. In the paKcispinis, melanops, and 

 pinniger groups (see classification below) the base of the skull is strikingly 

 curved; the interorbital space is always convex (at most flat, never con- 

 cave) and relatively Avide, its width never being more than 3i in the 

 length of the base of the skull; the mesethmoid processes are never 



