no prootic shelf such as is present in balistids. In 

 triacanthids the interoperculum extends posteriorly to 

 the suboperculum, but in balistids it is shorter and more 

 rodlike, not extending posteriorly beyond the epihyal. In 

 triacanthids the parasphenoid is a relatively narrow 

 shaft in front of the orbit, but it is greatly expanded there 

 into a thick plate in balistids. In triacanthids the vomer 

 is larger and less concave anteriorly than in balistids, 

 which lack the posterolateral wings that in triacanthids 

 suture to the long forward extensions of the prefrontal 

 (with the possible exception of Protacanthodes) . In 

 triacanthids the posttemporal is larger and more super- 

 ficially held to the skull than in balistids. 



The above outline of the similarities and differences 

 between triacanthids and balistids clearly indicates that 

 while balistids are highly differentiated from 

 triacanthids, the latter are the closest ancestral group of 

 the balistids, and that the level of organization of the Re- 

 cent triacanthids (and their close Oligocene relatives) is 

 too specialized in a few ways for their consideration as 

 being ancestral to balistids. Moreover, the same line of 

 triacanthids that gave rise to the balistids also gave rise 

 to the a^acanids, and the few features of both balistids 

 and aracanids that are more generalized than in Recent 

 triacanthids must have been present in the early 

 triacanthids leading to the balistids and aracanids. 

 These few generalized features required of the early 

 triacanthids are: 1) a free parhypural; 2) haemal spine 

 of penultimate vertebra autogenous; 3) caudal peduncle 

 relatively deep and not yet elongate and highly tapered 

 and the caudal fin not deeply forked; 4) soft dorsal fin 

 base probably not yet highly elongate; 5) the uppermost 

 pectoral fin ray only moderately reduced in size and the 

 two halves still of about equal length; 6) the teeth not yet 

 enlarged to massive incisors but somewhat intermediate 

 between conical and incisorlike; 7) the prefrontal not yet 

 with a long forward extension alongside the ethmoid 

 sutured anteriorly with posterolateral wings of the 

 vomer; 8) the second and third basal pterygiophores of 

 the spiny dorsal fin not yet greatly reduced in size. 



Among the fossil triacanthids, the Oligocene Acantho- 

 pleurus and Marosichthys are so like the Recent 

 triacanthids that they are unlikely to be closely related to 

 the line of early triacanthids giving rise to the balistids 

 and aracanids, while the Oligocene Cryptobalistes has 

 certain balistid features in its configuration which shows 

 the possibility of triacanthids being able to give rise to 

 certain balistidlike features, even though Cryptobalistes 

 seems to have been an evolutionary dead end unrelated 

 to the line of triacanthids which were ancestral to the 

 balistids. The Eocene Protacanthodes, which is inter- 

 mediate between the triacanthodids and triacanthids, is 

 the only known example of the early triacanthids, and of 

 those features of its anatomy that are exposed it is 

 generalized enough to encompass some of the critical 

 generalized features required of the triacanthid line 

 leading to the balistids and aracanids as listed above. 



In Protacanthodes the condition of the uppermost pec- 

 toral fin ray, the parhypural and the haemal spine of the 

 penultimate vertebra are not known, but the caudal 



peduncle is relatively deep and of only moderate length 

 and the caudal fin is rounded. These are conditions to be 

 expected in the balistid-aracanid ancestral line, except 

 that the caudal peduncle of Protacanthodes is probably 

 already somewhat too long for that line. The more medial 

 of the teeth in the jaws aie large incisors with con- 

 stricted distal ends, and thus somewhat notched, while 

 more laterally the edges were relatively straighter. While 

 the teeth in Protacanthodes may have been slightly 

 smaller than in Recent species (except Trixiphichthys), 

 they are probably already slightly too enlarged and in- 

 cisorlike to be conveniently ancestral to those of the 

 balistid-aracanid line, even assuming that early balistids 

 had less massive teeth than at present and that those of 

 early aracanids were larger and less conical than at pres- 

 ent. The prefrontal in Protacanthodes is large and has 

 at least a moderate anteroventral extension alongside the 

 ethmoid, although there is no evidence that it formed a 

 long process articulating anteriorly with the vomer, and 

 it is better separated anterodorsally from the ethmoid 

 than in Recent triacanthids. Thus, the prefrontal of 

 Protacanthodes is not known to have been much more 

 specialized than in the balistid-aracanid line, although it 

 may have been slightly more specialized. The soft dorsal 

 fin base in Protacanthodes is not yet greatly elongate, 

 and the second and third basal pterygiophores of the 

 spiny dorsal fin are still moderately large, even though 

 the spiny dorsal fin base is much shorter than in the an- 

 cestral triacanthodids. The second basal pterygiophore 

 articulates ventrally with the tip of the neural spine of 

 the second vertebra and the third basal pterygiophore to 

 that of the third, so that the articulations of both of these 

 pterygiophores has already shifted forward relative to the 

 condition in triacanthodids. In the line ancestral to 

 balistids and aracanids the third basal pterygiophore 

 would be expected to retain an articulation with a more 

 posteriorly located neural spine, as in triacanthodids, 

 and especially as in Eoplectus, as discussed in detail 

 there. Thus, the second and third basal pterygiophores of 

 Protacanthodes are already slightly too specialized to be 

 conveniently ancestral to the condition postulated to be 

 necessary in the balistid-aracanid line. 



While Protacanthodes is by far the most generalized of 

 the triacanthids, a number of its characteristics (caudal 

 peduncle length, tooth size and shape, placement of 

 third basal pterygiophore of spiny dorsal fin) are already 

 slightly too specialized for it to have given rise to the 

 balistids and aracanids. The divergence of the balistid- 

 aracanid line must have taken place among the early 

 triacanthids at a slightly more generalized (that is, more 

 triacanthodidlike) level of organization than as seen in 

 certain anatomical regions of Protacanthodes. In this 

 light the line of early triacanthids derived of a hollardiin- 

 like triacanthodid ancestry diverged into one line leading 

 with little change to Protacanthodes and to another line 

 which after far greater change gave rise to the balistids 

 and aracanids, the latter diverging the furthest from the 

 common triacanthid ancestral stock, and the 

 Protacanthodes line continuing on to give rise to the 

 Triacanthinae and probably Cryptobalistinae as well. 



