The species of the Hypostomus cochliodon group (Siluriformes: Loricariidae). Jonathan W. Armbruster Zootaxa 2003 249 1 60 3NZ6W urn:lsid:zoobank.org:act:7520B3CD-59C7-4425-A30E-1336A0B2AADD Loricariidae Hypostomus CoL Animalia Hypostomus cochliodon Kner 1854:265 [15] Siluriformes 54 Chordata species cochliodon  Discussion Delimiting the species of the Hypostomus cochliodongroup is made difficult by the fact that the species vary little in morphometrics or meristics, and the species have a well developed ability to alter color pattern. However, the species can be divided into four groups and from those groups it is possible to distinguish the species. The four groups are the intermediate group ( H. hemicochliodonand H. sculpodon), H. cochliodonwith a unique color pattern, the odontodeless opercle group ( H. ericius, H. levis, H. oculeus, H. pyrineusi, and H. taphorni), and an undifferentiated northern group ( H. hondae, H. pagei, and H. plecostomoides). The odontodeless opercle group can be further subdivided into a highly keeled group ( H. ericiusand H. oculeus) and a weakly to non-keeled group ( H. levis, H. pyrineusi, and H. taphorni). Once in these groups, it is fairly simple to separate the species. Principal components analysis is provided for each of the groups although H. levisis excluded from the analysis because it lacks many of the measurements because of its absence of an adipose fin. The PCA of the intermediate group showed a 100% separation of the species on principal component 2 (Fig. 10A). PC2 is most strongly and negatively effected by dorsal-anal length and anal-fin width and most strongly and positively effected by anal-fin length and folded dorsal-fin length. A ratio of dorsal-anal length to anal-fin length completely separates the two species from one another: average = 111.1 ± 8.4%, 92.5-132.6% in Hypostomus hemicochliodonand average = 79.2 ± 7.6%, 69.6-89.7% in H. sculpodon). The PCA of the odontodeless groups showed two trends, a complete or almost complete separation of H. taphornifrom H. ericius, H. oculeus, and H. pyrineusiand a complete separation of H. ericiusand H. oculeuson a graph of PC2 vs. PC3 (Fig. 10B). PC2 is most strongly and negatively effected by adipose-caudal length and mouth width and most strongly and positively effected by interdorsal length, eye-nare length and orbit length. PC3 is most strongly and negatively effected by base of anal-fin length, anal-fin length, and adipose-spine length and most strongly and positively effected by interdorsal length, adipose-caudal length, and postanal length. Although there is no morphometric evidence suggesting that H. ericiusand H. oculeusare distinct from H. pyrineusithere are morphological differences (well-developed keels in H. ericiusand H. oculeusvs. no or weak keels in H. pyrineusi). The PCA of the northern group showed some general trends separating H. hondaeand H. pageifrom H. plecostomoidesand a complete separation of H. hondaefrom H. pageion PC2 vs. PC3 (Fig. 10C). PC2 is most strongly and negatively effected by eye-nare length and interorbital width and most strongly and positively effected by anal-fin length and interdorsal length. PC3 is most strongly and negatively effected by anal-fin length and folded dorsal-fin length and most strongly and positively effected by adipose-caudal length and interdorsal length. For the most part, the species of the Hypostomus cochliodongroup are allopatric (Figs. 8, 11, 17). It is only in the Amazon basin that several species coexist. Panaque, the other wood-eating genus of loricariid catfishes, also reaches its peak of diversity in the Amazon basin (Schaefer & Stewart 1993; Schaefer & Stewart 2002). In the upper Rio Maranondrainage of Peru, there are at least nine species of wood-eaters: H. ericius, H. hemicochliodon, H. oculeus, H. pyrineusi, P. albopunctatus, P. dentex, P. gnomus, P. nocturnus, and probably at least one species of the P. nigrolineatusclade (distributions based on this study and Schaefer & Stewart 1993). How these species subdivide the wood-eating niche is unknown, and ecological studies on wood-eating by fishes in the upper Rio Maranonwould represent a very interesting study. Weber and Montoya-Burgos (2002) recently described Hypostomus fonchiiand suggested that the species was derived from the H. cochliodongroup. The teeth described for H. fonchiiare elongate and unicuspid and it is suggested that the unicuspid teeth represent a synapomorphy for H. fonchiiand the H. cochliodongroup. This is problematic in that I have not observed truly unicuspid teeth in any members of the H. cochliodongroup. In all specimens of the H. cochliodongroup that I have examined, the medial cusp remains present, but is generally fused with the lateral cusp. The mesial cusp remains visible as a slightly darker, thicker ridge on the medial side of the tooth. I have not examined any specimens of H. fonchii, and I cannot determine whether or not this is the case with this species. The only truly unicuspid teeth that have been reported for the Hypostominaeare the teeth found in nuptial male Aphanotorulus(Armbruster & Page 1996). I do not consider H. fonchiias a member of the H. cochliodongroup as its body shape is that of a generalized, fast-water dwelling Hypostomusand not the derived shape of H. cochliodon. Exact placement of H. fonchiiawaits an analysis of its diet, osteology, and external anatomy for the synapomorphies of the H. cochliodongroup. Recent expeditions into poorly collected regions where no members of the Hypostomus cochliodongroup have been reported from in the past are finding more species of the H. cochliodongroup. Two undescribed species are now known from the Essequibo and Takutu River drainages of Guyana (pers. obs.) and one species from the Rio Tocantins in Brazil (Reis, pers. comm.). These species and a revised key will be presented in future publications. Phylogeny Character 1: Teeth - 0: elongate (Fig. 1A); 1: intermediate throughout life (Fig. 1B); 2: spoon-shaped at least in adults (Fig. 1C). Character 2: Maxilla - 0: straight to moderately curved (Fig. 3A); 1: greatly curved, almost forming right angle (Fig. 3B). Character 3: Odontodes on opercle - 0: 11+ (Fig. 2A); 1: 0-10 (Fig. 2B). Character 4: Longitudinal ridge formed from bone and slightly enlarged odontodes on pterotic-supracleithrum - 0: present (Fig. 5A); 1: absent (Fig. 5B). Character 5: Nuptial body odontodes - 0: absent (Fig. 2A); 1: present (Fig. 2B). Character 6: Notch between the hyomandibula and metapterygoid - 0: present (Fig. 3A); 1: absent (Fig. 3B). Character 7: Buccal papilla - 0: present (Fig. 4A); 1: absent or extremely small (Fig. 4B). Character 8: Dentary angle - 0: averaging greater than 90°; 1: averaging less than 80°. Character 9: Sharp keel odontodes - 0: present; 1: absent. Character 10: Body shape deep and narrow with the head taller than wide - 0: absent; 1: present. Very few morphological characters are useful in elucidating the relationships of the species of the Hypostomus cochliodongroup. Based on the ten characters found, a phylogeny is produced for the species of the H. cochliodongroup (Fig. 23). The single tree found has 12 steps and CI = 1.00. Character state data is in Table 8. No skeletal material is available for H. sculpodonor H. ericius; however, external evidence does provide information as to the potential relationships of these species with other members of the H. cochliodongroup. Considering there are so few characters available for analysis, the phylogeny should be considered tentative. Also providing support for the basal position of Hypostomus hemicochliodonand H. sculpodonis diet. In all of the specimens of the H. cochliodongroup examined, almost the entire intestine is filled with small wood chips suggesting that the fishes consume little other than wood. In H. hemicochliodonand H. sculpodonthe intestine mostly contains wood, but other organic matter makes up a major portion of the diet (not greater than 50%). It is apparent that H. hemicochliodonand H. sculpodonare wood eaters but not wood specialists.