174 Mr. G. A. Boulenger on the Suborders and 



2. Pelvic bones not connected with scapular _ arch ; ventrals 

 without spine, with 5 or 6 rays ; snout tubiform ; first ver- 

 tebra very elongate, formed by the fusion of several. 



Isolated dorsal spines ; body scaly 5. Aulostomatidce. 



No dorsal spines ; body naked . .' 6. Fistulariidce. 



B. Mouth toothless ; snout tubiform ; two short dorsal fins, the first 

 with a few spines ; ventral fins with 3 to 5 rays ; anterior ver- 

 tebrae elongate. 



Body covered with bony shields and small 



"rough scales 7. Centriscides. 



Body completely cuira?sed by bony shields, 



which are fused with the endoskeleton . . 8. Amphisilidce. 



III. Praeoperculum absent ; symplectic much elongate ; branchial 

 apparatus more or less reduced ; gill-lamellae reduced in number 

 and enlarged, forming rounded lobes ; post-temporal simple, im- 

 movably attached to the skull ; mouth toothless, at the end of a 

 tubiform snout; body covered with bony plates. (Lopho- 



BKANCHII.) 



Two dorsal fins ; ventral fins present, with 7 

 rays ; gill-openings wide ; exoskeleton of 

 large star-like plates 9- Solenostomida. 



A single dorsal fin ; no ventral fins ; gill- 

 openings very small ; exoskeleton in the 

 form of rings 10. Syngnathidce. 



IV. Pneoperculum and symplectic absent; gills pectinated; mouth 

 inferior, toothless ; body entirely covered with bony plates ; ven- 

 tral fin with 2 or 3 rays. (Hypostomides.) 



11. Pegasidce. 



Suborder VIII. PERCESOCES. 



Air-bladder, if present, without open duct. Parietal bones 

 separated by the supraoccipital. Pectoral arch suspended 

 from the skull ; no mesocoracoid arch. Ventral fins, if 

 present, abdominal, or at least with the pelvic bones not 

 solidly attached to the clavicular arch. 



This group connects the Haplomi with the Acanthopterygii, 

 the Sconibresocidse being somewhat related to the (Jyprino- 

 dontidse *, whilst the Anabantidse show distinct affinity to the 



* Swinnerton (Quart. Journ. Micr. Sci. xlv. 1902, p. 554) has pointed 

 out that the skull of the Scombresoces belongs to what_ he terms the 

 Acrartete type (e. g. in which the attachment of the palatine cartilage or 

 its derivates is confined to the pre-ethmoid cornua), whilst the other 

 Percesoces examined by him, as well as the Cyprinodonts, are Disartete 

 (the attachment being at the par-ethmoid and pre-ethmoid^ cornua) ; but 

 the character is so indistinctly defined in some adult Cyprinodonts, that 

 I feel some diffidence in making use of this character for systematic 

 purposes in the present state of our knowledge. 



