204 



ONTOGENY AND SYSTEMATICS OF FISHES-AHLSTROM SYMPOSIUM 



Salangids have been associated in the past with various mem- 

 bers of the osmeroid assemblage, but even this was questioned 

 by Nelson (1970). Rosen (1974) presented evidence from the 

 caudal skeleton which shows that salangids are osmeroids, but 

 no evidence about their placement within the group has been 

 presented to date. Fink and Weitzman (1982) agreed with Rosen 

 and placed the Salangidae as incertae sedis in the Osmeroidei. 



What little evidence I have been able to find about the rela- 

 tionships of salangids is equivocal. If examined by a standard 

 parsimony procedure, as represented by the Wagner analysis 

 shown in Fig. 107 (bottom), the numerous reductive traits of 

 salangids place them within the "southern smelt" plus galaxiid 

 assemblage. On the other hand, salangids share with Plecoglos- 

 siis and the "osmerids" a complex caudal skeleton character 

 involving fusion of uroneural 1 to a compound centrum made 

 upofPUl, Ul,and U2, followed ontogenetically in some forms 

 by fusion of rudimentary neural arches with the uroneural por- 

 tion of the complex. This latter character is in contrast to the 

 autogenous uroneurals of most galaxiids, the "southern smelts," 

 and other primitive teleosts. Further, when uroneurals and ru- 

 dimentary neural arches are fused in galaxiids, the fusion se- 

 quence is rudimentary neural arch to the compound centrum, 

 followed by fusion with the uroneural, rather than the reverse. 

 The hypothesis that emerges from these observations is illus- 

 trated in Fig. 107 (top), showing salangids, Plecoglossus. and 

 "osmerids" in an unresolved trichotomy. For further discussion 

 of caudal fin morphology, see Greenwood and Rosen (1971), 

 Rosen (1974), and Fink and Weitzman (1982). 



Any choice of these alternate hypotheses of salangid relation- 

 ships would rest on whether or not one wished to accept the 

 numerous reductive traits that unite the salangids with the 

 "southern smelts" and galaxiids as homologues. Such choice is 

 based on criteria which cannot be discussed in detail at this 

 point due to space restrictions, but I have commented elsewhere 

 (Fink, 1982) on hypothesis choice forced by confrontation with 

 apparent paedomorphosis. In this case, for example, some of 

 the general morphological attributes that salangids share with 

 the members of those groups differ when examined in detail. 

 Although this lack of close correspondence in similarity is cer- 

 tainly no guarantee that the reductions are not homologous, it 

 does raise the issue. Further, the highly developed caudal skel- 

 eton of salangids is identical to that of "osmerids," and thus 

 more differentiated than that of either the southern smelts or 

 galaxiids. This incongruity in degree of morphological differ- 

 entiation suggests that in this case, one should be cautious in 

 assuming homology in the reductive process and search for 

 other, non-reductive characters to resolve possible misplace- 

 ments. 



The family Sundasalangidae is not accepted herein because 

 in every case in which Roberts (1981) contrasted sundasalangids 

 and salangids, the character for salangids was primitive. I suggest 

 that recognition of family rank for Sundasalanx^^oviXd probably 

 render the Salangidae paraphyletic and thus defined only by the 

 absence of characters present in Sundasalanx. This is unac- 

 ceptable both because it forces recognition of a group based on 

 characters its members lack and because it artificially breaks up 

 a group all of whose members share a unique evolutionary his- 

 tory. 



Regarding the "southern smelt assemblage" (including gal- 

 axiids, but excluding salangids), I am less pessimistic than 

 McDowall (this volume). I have taken the liberty of using the 

 data he has presented and combined them with my own limited 



survey of specimens and the literature to produce the hypotheses 

 shown in Fig. 107. The group can be diagnosed by presence of 

 a posteroventral deflection of the infraorbital sensory canal (Nel- 

 son, 1972) and 9/9 or fewer principal caudal-fin rays (vs a pos- 

 terodorsal curvature of the canal and 10/9 rays in outgroups). 

 Several characters support the placement of Retropinna and 

 Prototrocles as sister taxa including presence of an abdominal 

 homy keel, loss of the right ovary, and ceratohyal morphology. 

 I have no specimens o( Stokellia on hand, but McDowall's work 

 ( 1979) clearly shows that the genus is diagnosable and that it is 

 related to Retropinna and Prototrocles. Unfortunately, when 

 contrasted with Stokellia, it is not clear that Retropinna is di- 

 agnosable, since the latter is then differentiated by primitive 

 characters present in other taxa. 



Relationship among Aplochiton, Lovettia and the galaxiids is 

 supported by numerous characters, as shown in Fig. 107. I have 

 been unable to find any features that link the former two genera 

 together, however, and more work needs to be done with them. 

 Galaxiids themselves can be shown to be monophyletic based 

 on such characters as basioccipital "pegs" extending lateral to 

 the anterior centrum (McDowall, 1969, Figs. 2B, lOA, but note 

 lack of "pegs" in G. paucispondylus. Fig. 1 OB). 



In summary, it is suggested that the broad outlines of rela- 

 tionships among the osmeroids are beginning to emerge, much 

 as suggested by Gosline (1960a), with a "southern smelt" as- 

 semblage and an "osmerid" assemblage. Interrelationships within 

 these groups remain problematical, the most obvious problems 

 being establishment of the natural groups within the "osmerids" 

 and placement of the salangids. 



Salmontds. — M.onox>\\y\y of this group is based primarily on a 

 single character, apparent polyploidy of the karyotype (Gold, 

 1979). Several investigators have studied interrelationships of 

 salmonids, most notably Behnke (1968) and Norden (1961), but 

 these works were not phylogenetic and changes can be expected. 

 I have examined phylogeny within the group only to establish 

 polarities for characters relevant to relationships with other te- 

 leosts. Regarding the latter relationships, there have been several 

 opinions, with most workers approaching salmonids with an 

 eye to finding ancestors ofother groups (see, e.g., Gosline, 1960, 

 Diagram 2). The only phylogenetic analysis to date is that of 

 Rosen (1974), which was discussed by Fink and Weitzman 

 ( 1 982). The latter authors presented data which they considered 

 suggestive of neoteleostean relationship for salmonids: presence 

 in some members of paired cartilages anterior to the ethmoid 

 region (resembling the median rostral cartilage of neoteleosts) 

 and the exoccipital forming part of the occipital condyle. The 

 anterior cartilages were reported by Fink and Weitzman (1982) 

 to be prominent in Prosoplum, an observation which I can 

 confirm from additional specimens. In addition, examination 

 ofsmall juvenile cichlids shows that the rostral cartilage appears 

 to develop ontogenetically from bilateral cartilage bodies which 

 fuse at the midline; this is suggestive of corroboration of Fink 

 and Weitzman's (1982) hypothesis that the rostral cartilage 

 evolved from paired cartilages anterior to the ethmoid region 

 like those in Prosoplum. More work needs to be done on the 

 homology of "accessory" ethmoid cartilages, using double stain- 

 ing techniques and histology on a wide variety of teleosts. 



1 can also add to what Fink and Weitzman ( 1 982) noted about 

 the occipital condyle. 1 have confirmed that the exoccipital forms 

 part of the condyle in Thymallus and "salmonins." This mor- 

 phology is also present in Prosoplum. but is lacking in other 



