posterior tips of the third and fourth epibranchials. The infrapharyngobranchials of 

 the third and fourth arch are ossified; that of the third is relatively elongate and 

 club-shaped, while that of the fourth is broad and squarish. The anteromedial tip of 

 the first epibranchial is broadly flared and overlaps the medial tip of the second; 

 Ferraris (1988) considered this to be synapomorphic of ageneiosids. 



Ageneiosids have numerous, relatively large gill rakers on the medial and 

 lateral margins of the ceratobranchials and epibranchials (Fig. 13). In most species 

 the rakers are compressed and have a crenulated medial margin, but in^l. brevifilis, 

 A. marmoratus, and A. polystictus the rakers are conical and relatively short. Britski 

 (1972) surveyed the gill rakers in several auchenipterid and doradid taxa, and found 

 considerable variation in their structure and number. Depending on the taxon, the 

 rakers on the first arch may be in any combination of one or two rows, and relatively 

 long, short, or even absent. Based on Britski's comparisons, and from what is known 

 in other catfishes, the primitive condition for length is one in which the rakers are 

 relatively short; lengthening of the rakers is presumably derived, but has probably 

 occurred independently in a number of taxa. Similarly, arrangement of the rakers 

 into two rows is presumably primitive, since this is the condition in diplomystids and 

 ictalurids (Arratia 1987, Lundberg 1982), and, I assume, many other catfishes. 

 Based on what is available in the literature, I have the impression that the shape, 

 number, and arrangement of gill rakers has been relatively neglected in catfish 

 systematics; from what I have observed in ageneiosids, it seems that this character 

 has considerable systematic value, at least in the alpha-level taxonomy. 



Apart from differences in gill-raker shape, the number of gill rakers varies 

 widely among ageneiosid species. In addition, there is considerable intraspecific 

 variation in gill-raker number. I was unable to find any geographic component to 

 this variation. However, there appeared to be some correlation with size (larger 

 specimens with more gill rakers), although I did not test this hypothesis statistically; 



.11 





