•■; . : _ ■:. ■:,: >;'; 157 



a greater number of reduced, reoriented fin rays and supporting elements. Ferraris 

 (1988) went so far as to speculate that internal fertilization does not occur in 

 centromochlids, and he used this argument to partially rationalize family status for 

 the group. I question the assumption that centromochlids lack internal fertilization, 

 in the absence of direct evidence to support the idea. Spermatozeugmata are 

 present in the testes of Centromochlus (personal observation), which is highly 

 suggestive of internal fertilization. Moreover, the unusual anal fin may be an 

 extremely derived structure, which would argue in favor of a more advanced type of 

 gonopodium than is present in other taxa. Regardless of my opinion about the 

 probable mode of fertilization in these species, however, I have little doubt that the 

 taxa included in Ferraris' Centromochlidae are monophyletic. Consideration of the 

 family-group nomenclature rests on the phylogenetic relationships among the 

 remaining genera, and this is where there are the greatest incongruencies in 

 Ferraris' analysis, r, — -, ,^ " •, " • i 'r, ■ .^ \ 



The cladogram of ageneiosids and the remaining auchenipterid genera 

 presented by Ferraris has four unresolved polychotomies (Fig. 33). I have not 

 included the character states supporting Ferraris' hypothesized phylogeny; to do so, 

 and to discuss their distribution, would be very exhaustive. The following comments 

 are provided only for their relevance in assessing a probable sister group of 

 Ageneiosus and Tetranematichthys, and their bearing on the family-group 

 nomenclature. -■ • 



Many of the characters that Ferraris (1988) included on his cladogram as 

 synapomorphies of putatively monophyletic lineages were considered to be 

 homoplasious, both within the ingroup and among other doradoids (centromochlids 

 and doradids). Some of the convergences or reversals were discussed by Ferraris, 

 but several were not explained. Given this, and the number of unresolved nodes, 

 there are probably several equally parsimonious trees to the one presented by 



