372 MR C. TATE REGAN ON THE 



III. No orbitosphenoid ; no opisthotic ; no upturned vertebrae ; mesopterygoids 

 toothed (absent in Salangidse). 



A. A mesocoracoid ; maxillaries dentigerous, entering gape . 4. Osmeridee. 



B. No mesocoracoid ; maxillaries dentigerous, entering gape. 



Head compressed ; mesopterygoid well developed, dentigerous ; ribs ossified. 



5. RetropinnatidfB. 



Head strongly depressed ; no mesopterygoid ; ribs not ossified 6. Salangidx. ' 

 0. No mesocoracoid ; maxillaries toothless, behind prsemaxillaries. 



Prsemaxillaries not extending whole length of maxillaries ; roof of myodome 

 unossified ; no adipose fin . . . . . . 7. Galaxiidse. 



Prsemaxillaries nearly reaching extremities of maxillaries ; roof of myodome 

 ossified ; an adipose fin . . . . . 8. Haplochitonidse. 



The Argentinidse and Microstomidse are inhabitants of rather deep water, but the 

 rest are littoral fishes, many of them entp.ring fresh water and often forming colonies, 

 races, or species confined to fresh water. 



It is of some interest to note that the Galaxiidse and Haplochitonidse are related to, 

 but more specialised than, the Osmericlse, or Smelt family, of northern seas. Retropinna, 

 from the coasts and rivers of Australia and New Zealand, is still nearer to the Galaxiidse 

 and Haplochitonidse ; both these families occur in Australia, Tasmania, New Zealand, 

 South America, and the Falkland Islands, and there are even two species of Galaxias 

 at the Cape of Good Hope. All the species enter fresh water, and the majority seem 

 to be strictly fluviatile or lacustrine, but in a few cases species of Galaxias have been 

 observed in the sea. 



In 1906 (Proc. Zool. Soc., 1905, ii. pp. 36.3-384, pis. x.-xiii.) I published a 

 revision of the Galaxiidse, and then wrote : 



" The occurrence of Galaxias maculatus in the sea has been recorded by VALEN- 

 CIENNES and by PHILIPPI, off the Falkland Islands and off the coast of Chile respectively. 

 The observations of JOHNSTON in Tasmania and of HUTTON and CLARKE in New Zealand 

 are to the effect that Galaxias attenuatus descends to the sea periodically to spawn. 

 Mr RUPERT VALLENTIN has seen shoals of little fishes, which I identify with the 

 Galaxias gracillimus of CANESTRINI, in the sea at the Falkland Islands. Recently 

 Galaxias brevipinnis also has been found to be marine, G. bollansi, described by 

 HUTTON from the Auckland Islands, proving to be identical with this species." 



WAITE (Subantarctic Islands of New Zealand, p. 586) has recently shown that 

 BUTTON'S conclusion as to the marine habit of G. brevipinnis was probably incorrect. 

 EIGENMANN (Rep. Princeton Exped. Patagonia, iii., Zool., 1909, p. 274) says of G. 

 gracillimus : " This is undoubtedly the young of attenuatus " ; and if this opinion, which 

 does not appear to be the result of an examination of specimens, be accepted, the known 

 marine species of Galaxias would be reduced to two only. 



In my revision I distinguished G. gracillimus from G. attenuatus by the more 



(ROY. soc. EDIN. TRANS., VOL. XLIX., 290.) 



