matrix of the biting edge of the premaxillaries and den- 

 taries, as found in the triodontids, the most generalized 

 gymnodonts, and in their ancestral eoplectin triacan- 

 thodids (as well as in the moderately specialized diodon- 

 tids). Of the two known eoplectins, the biting edges of 

 the jaws are exposed only in Eoplectus, which has small 

 rounded dental units, but this can also be expected to be 

 the condition in the related Zignoichthys. The eoplec- 

 tins undoubtedly evolved from a group of early 

 triacanthodids with a generalized dentition of numerous 

 conical teeth in an outer series, internal to which there 

 were a smaller number of conical inner series teeth. 



Conversion of this generalized dentition into that as 

 found in eoplectins and triodontids probably involved a 

 great increase in the number of outer series teeth and 

 their placement in more than one row concomitant with 

 a great reduction in size of the individual teeth, which 

 eventually protruded less and less to the exterior beyond 

 their basal sockets in the premaxillaries and dentaries 

 and finally became entirely nonprotrudant, being fully 

 surrounded by the matrix of the bone. It is not known 

 whether eoplectins had trituration plates internal to the 

 biting edges of the jaws, but it is probable that they did, 

 for trituration plates are found in all triodontids, diodon- 

 tids, molids, and in many tetraodontids, and the jaws of 

 eoplectins are just as massive as in most gymnodonts and 

 the eoplectin diet probably consisted of hard shelled in- 

 vertebrates for which trituration plates would be useful 

 in crushing and grinding. These trituration plates are ob- 

 viously formed from modified internal series teeth as 

 found in the more generalized triacanthodids, the teeth 

 becoming greatly increased in number, and sometimes in 

 size, to form a series of rows which variously retained 

 much of their individual identity or became incor- 

 porated into a plate in which some of the teeth lost their 

 individual identity, especially anteriorly in the plate 

 away from the posterior region of tooth replacement. 

 However, the development of trituration teeth, whether 

 as plates or not, is probably a highly labile feature in 

 various lines of gymnodonts closely correlated with 

 dietary changes as species evolved in differing habitats. 



The generalized gymnodont dentition of small rounded 

 units in the biting edge has been retained by triodontids 

 and diodontids, but tetraodontids and molids have great- 

 ly modified the dentition of the biting edge. In molids 

 discrete dental units are no longer present, at least as ob- 

 served at 30 magnifications, the ancestral small rounded 

 units apparently indistinguishably fused to the bony 

 matrix. In tetraodontids the ancestral small rounded un- 

 its have been modified into less numerous long rods lying 

 in series approximately parallel to the biting edge. The 

 tetraodontid dental units of the biting edge obviously did 

 not become specialized long rods until after the division 

 of the joint tetraodontid-diodontid ancestral line, for 

 diodontids retained small rounded units from this line. 



It seems reasonable to assume that the tetraodont jaw 

 type (but not tooth type), with both the premaxillaries 

 and dentaries closely articulated but not fused to their 

 opposite members, was the primitive condition of early 

 gymnodonts, and that both the triodont jaw type, with 



only the dentaries fused to their opposite members, and 

 the diodont jaw type, with both the dentaries and pre- 

 maxillaries fused to their opposite members, are derived 

 from a tetraodont jaw type ancestry, either in- 

 dependently or with the diodont type perhaps having 

 been preceded by a triodont type ancestral group. 



It is not known whether the premaxillaries and den- 

 taries were fused or articulated to their opposite mem- 

 bers in the Eocene triacanthodid Eoplectus, but in the 

 apparently closely related Zignoichthys from the same 

 strata, at least one of the jaws, probably the lower, had 

 the two halves (presumedly dentaries) fully fused 

 together. 



If these two Eocene triacanthodids, representing the 

 subfamily Eoplectinae, are indeed representatives of the 

 ancestral line leading to Triodon and the other gym- 

 nodonts, then at least Zignoichthys is already too 

 specialized in its jaw structure (which must be either 

 triodont or diodont type) to be considered as ancestral to 

 the tetraodontids. It seems to me extremely unlikely that 

 once the full fusion of the dentaries and/or premaxil- 

 laries was established in a line of gymnodont evolution 

 that consistent reversal to an unfused condition could 

 take place, especially in light of the intricate inter- 

 locking interdigitations that are almost always present 

 when the premaxillaries or dentaries are articulated to 

 one another rather than fused. 



It is assumed here that the immediate ancestry of the 

 few known eoplectins and thus of the triodontids had the 

 premaxillaries and dentaries separate from their op- 

 posite members, and that the tetraodontids diverged 

 from this ancestral line at a time in the Eocene when at 

 least some members of that line retained separate pre- 

 maxillaries and dentaries. This line was probably 

 TViodon-like except in the retention of separate den- 

 taries and also probably before the elongate tapered 

 caudal peduncle as seen in the Recent Triodon had been 

 fully established. Since it is here considered that the 

 diodontids are more closely related to tetraodontids than 

 to molids, an additional implication of the above reason- 

 ing is that the fusion of the premaxillaries and dentaries 

 in molids and diodontids has taken place in- 

 dependently, that in diodontids from a tetraodontid an- 

 cestral group, and that in molids from the same eoplec- 

 tin-triodontid line retaining separate premaxillaries and 

 dentaries from which tetraodontids are evolved. 



The features in the conversion of an eoplectinlike fish 

 into a triodontidlike one are discussed under Eoplectus 

 in the account of the Triacanthodidae. The conversion of 

 a triodontidlike fish into a fish like that of any of the 

 other derived families of gymnodonts involves primarily 

 the loss of most of its more generalized or triacanthodid- 

 like features. 



The features of Triodon that are lost by all other gym- 

 nodonts are: 1) the rudimentary spiny dorsal fin and all 

 but one of the basal pterygiophores and supraneural 

 elements supporting it; 2) the ribs and epipleurals; 3) the 

 pelvis; 4) uroneurals and hypurapophysis, and the con- 

 solidation of the hypurals so that only one free element is 

 present and the conversion from an autogenous to fused 



