8 ATHERINID^: SILVERSIDES 



coidean ancestry; moreover, in Nannatherina the dorsals are joined 

 at their base, and in other genera of the Atherinid^ the dorsals are 

 closely approximated; in the Mugilid.e and in certain genera of the 

 Atherinid^ (Nannatherina, Centratherina), the dorsal spines are 

 all pungent. The caudal fin of the Percesoces has been described as 

 diphycercal, but in Sphyr^na, as in many typical percoids, only two of 

 the neural and haemal spines enter into the support of the procurrent 

 caudal rays ; in the Atherinid^, in which the vertebrae are increased 

 in number, more vertebral spines (four in Chirostoma humboldti- 

 anum) are included in the skeletal base of the caudal fin. There remains 

 of the known diagnostic features of the Percesoces but one, namely 

 the abdominal position of the ventral fins, which indicates a close rela- 

 tionship with the soft-rayed fishes. In the Mugilid^e and in certain 

 Atherinid.e, however, the pelvic bones are close to the pectoral arch: 

 the ventral fins are "subabdominal." Moreover the pelvic bones of these 

 forms are attached by ligament to the pectoral girdle, a fact which has 

 led Dollo^ to consider the ventrals of the Atherinid.e to be secondarily 

 abdominal. Regan^ likewise, has recently included the Percesoces among 

 the Percomorphi of his system. The lack of the orbitosphenoid in the 

 Percesoces; the number of caudal rays (15 branched, 17 in all, ex- 

 cluding the procurrent rays) ; the number of anal spines (three in most 

 of the MuGiLiD^ and in Nann.^therina, fewer in the other forms; 

 the constant occurrence of a spine and five soft rays in the ventral fin, 

 point positively to the same conclusion, namely, that the Percesoces are 

 of percoidean ancestry, and that the group is not transitional between 

 the soft-rayed and the spiny-rayed fishes. 



If the Percesoces are of percoidean ancestry, as the foregoing evi- 

 dence leads us to believe, then we have some basis upon which to deter- 

 mine the most probable evolutionary sequence of the genera within the 

 family Atherinid^e. The most evidently perch-like, and consequently 

 very probably the most primitive of the atherine fishes, is Nanna- 

 therina^, a fresh-water fish of Western Australia: its perch-like feat- 

 ures are its moderately robust body ; wide mouth ; eight or nine pungent 

 spines in first dorsal ; the connection of the two dorsals ; three anal 

 spines ; symmetrical, rounded, low pectorals ; anteriorly inserted ventrals ; 

 31 (14+17) vertebrse, fewer than in any other atherine so far as known. 

 Bedotia^ from the mountain streams of Madagascar, resembles Nanna- 

 therina and diflfers from other atherine genera in having the caudal 

 rounded, and in having the premaxillaries little protractile ; as in that 



7 Verh. Zool.-bot. Ges. Wien, 59, 135-140, 1909. 



^Ann. Mag. Nat. Hist. (8), 12, 111, 1913. 



9 Regan, Ann. Mag. Nat. Hist. (7). 18, 451. 1906. 



