372 



TRANSACTIONS OF THE AMERICAN PHILOSOPHICAL SOCIETY 



quadrate, by the broadened hyomandibular and backwardly produced preopercular. 

 Starks (1905, p. 295) shows that in Caularchus, a member of this family, the pterygoid is 

 reduced to a vestige adhering to the front edge of the quadrate, while the meso- and meta- 

 pterygoids are absent. The palatine is a narrow rod, connected with the vestigial pterygoid 



pfiiot 



sym '"y 



pop sop 



Gobiesox 



G. cephalus 



Fig. 249. Gobiesox. Side View. 



by a ligament but hooking over the maxilla in the normal way in front. The only bony 

 contact between the hyomandibular and the quadrate is by means of the small but strong 

 symplectic which is received into a notch in the middle of the quadrate, whereas in most 

 fishes this notch is in the back part of the quadrate. A prominent fenestra lies between 

 the preopercular and the quadrate, recalling the conditions in the goby Eleotris. Possibly 

 the presence of the great pelvic sucking-disc may have somehow caused the strange back- 

 ward growth of the opercular and subopercular and of the parts attached to them, which 

 together form a sharp lower rim projecting backward on either side of the disc. They may 

 serve for the origin of the disc muscles. The skull is broad and low, with long lateral 

 processes from the prefrontal (lateral ethmoid) and sphenotic. In Caularchus the vomerine 

 region is broadly notched in front (doubtless to receive the enlarged premaxillse), and the 

 flat, long ascending rami of the premaxillae lie in a broad depression on the upper surface 

 of the rostrum and between the orbits. These features are more emphasized in Caularchus 

 than in Gobiesox (Starks, 1905, pp. 292, 293). The parasphenoid has broad lateral wings 

 in front. Three occipital condyles are present, one on the basioccipital, the other two on 

 the exoccipitals, so that the three parts of the condyle are in a horizontal line (Starks). 



Many other strange cranial characters are recorded by Starks, who also made a pains- 

 taking comparison of skeletons of cottoid, blennioid and gobioid fishes in the effort to deter- 

 mine the affinities of the Gobiesocidae, but with "small results." The families Batrachi- 

 didae and Callionymidae offer some slight indications of relationship to the Gobiesocidse, and 

 the weight of evidence is thrown toward the former family by the young of some or all of 

 them having a ventral sucking-disc just behind the base of the pectorals. The family 

 Batrachididae further resembles the Gobiesocidse in having the suborbital ring reduced to a 

 small preorbital bone, only very small parapophyses present posteriorly, no myodome, and 



