dM d M ..>" juiy '^f^'SIS - \ cl>Tn i 'tfi :\ r^'^ )"' eainidae (i eieostei 4^% . J. . FIELDIANA Zoology NEW SERIES, NO. 84 A Phylogenetic Analysis of the Asian Catfish Families Sisoridae, Akysidae, and Amblycipitidae, with a Hypothesis on the Relationships of the Neotropical Aspredinidae (Teleostei, Ostariophysi) Mario C. C. de Pinna Department of Zoology University ofSdo Paulo Caixa Postal 1 1461. SOo Paulo, SP 05422-970 Brazil Department of Geology Field Museum of Natural History Roosevelt Road at Lake Shore Drive Chicago. Illinois 60605-2496 U.S.A. Accepted September 13, 1995 Published July 31, 1996 Publication 1478 PUBLISHED BY HELD MUSEUM OF NATURAL HISTORY © 1 996 Field Museum of Natural History ISSN 0015-0754 PRINTED IN THE UNITED STATES OF AMERICA Table of Contents Abstract 1 Resumo 1 Introduction 2 Materials and Methods 3 PhylogenetiQ Methodology 3 Terminal E>elimitation 4 Abbreviations 5 Material Examined 6 Preparation Techniques and Nomenclature 8 Character DESCRiFnoN and Discussion . . 9 Jaws and Related Structures 9 Suspensorium and Opercular Apparatus . . 13 Neurocranium 15 Branchial Arches 21 Hyoid Bar 25 Weberian Apparatus 27 Vertebrae and Ribs (Other Than Weberian Apparatus) 33 Pectoral Fins and Girdle 35 Dorsal Fin and Support 39 Laterosensory System 42 Pelvic Fins and Girdle 45 Miscellaneous 47 Phylogenetic Analysis 55 Classihcatory Scheme 57 Dl\gnoses for Clades and Synonymies FOR Named Taxa 59 Identification Key to Families Through subtribes 65 Discussion and Comparison with Pre- vious Classifications 66 Comments on Miniaturization and Gigan- tism IN SiSOROIDEA 70 Evolution of the Sound-producing Ap- paratus IN SiSOROIDS 71 Evolution of the Thoraoc Adhesive Ap- paratus 73 Relationships of the Neotropical Aspre- dinidae 75 Acknowledgments 77 Literature Cited 77 List of Illustrations 1 . Dysichthys sp., right palatine, lateral and dorsal views 9 2. Pseudexostoma yunnanensis, right max- illa, dorsoposterior view 10 3. Gagata gagata. premaxillae, anterior view 11 4. Akysis sp. 2 and Parakysis, right pala- tines, ventromesial view 1 1 5. Outline of premaxillae in Glyptothorax major and Euchiloglanis kishinouyei ... 1 2 6. Bagarius yarrelli, right premaxilla, dor- sal view 12 7. Dentary teeth in Pseudecheneis sulcatus and Pseudexostoma yunnanensis 13 8. Amblyceps mangois, metapterygoid and endopterygoid, ventromesial view 13 9. Laguvia ribeiroi, right opercle, lateral view 14 10. Breitensteinia insignis, neurocranium and associated structures 16 1 1 . Posterior skull, lateral view, in Conta conta and Erethistes pusillus 18 12. Bagarius yarrelli, posterior temporal re- gion, dorsolateral view 19 13. Glyptothorax major, anterior skull, dor- sal view 20 14. Glyptothorax sinensis, anterior neuro- cranium, ventral view 20 15. Gagata gagata, ethmoid region, dorsal view 21 16. Erethistoides montana, basal gill arches, dorsal view 22 17. Acrochordonichthys ichnosoma, ventral gill arches, dorsal view 22 18. Breitensteinia insignis, ventral gill arch- es, dorsal view 23 19. Amblyceps mangois, basal gill arches, dorsal view 24 20. Sisor rhabdophorus, basal gill arches, dorsal view 24 2 1 . Urohyals, ventral view, in Glyptothorax sinensis, Exostoma sp., and Pseudeche- neis sulcatus 26 22. Breitensteinia insignis, hyoid bar, dorsal view 27 23. Hara hara, hyoid bar, dorsal view 27 24. Gagata gagata, swim bladder capsule, ventrolateral view (SEM) 28 25. Dysichthys coracoideus, Weberian com- plex, dorsal view 29 26. Weberian complex, ventral view, in Glyptothorax major and Laguvia ribei- roi 29 27. Gagata gagata, tripus, lateral view in- side capsule (SEM) 30 28. Akysis sp. 1, anterior vertebral column, ventral view 31 29. Amaralia hypsiura, Weberian complex and other structures, dorsal view 31 30. Breitensteinia insignis, vertebral col- umn, ventral view 34 3 1 . Parakysis anomalopteryx, pectoral fin ray support, ventral view 35 32. Pectoral spines in Glyptothorax pectin- opterus and Akysis sp. 1 36 33. Psendexostoma yunnanensis, left pecto- ral fin ray support, ventral view 37 34. Conta conta, pectoral girdle, ventral view 38 35. Hara hara, nuchal plates, dorsal view . . 39 36. Conta conta, dorsal fin spines, lateral view 40 37. Parakysis anomalopteryx, putatively ju- venile condition of dorsal spines and nuchal plates, dorsal view 41 38. Erethistes pusillns, right lacrimal, dorsal view 42 39. Akysis leucorhynchus, left nasal, dorsal view 43 40. Left lacrimals, dorsal view, in Dysich- thys sp. and Acrochordonichthys rugosus 44 4 1 . Nangra nangra, preopercular branch of laterosensory canal system 44 42. Laguvia ribeiroi, right basipterygium, ventral view 45 43. Exostoma labiatum, right basiptery- gium, ventral view 46 44. Gagata gagata, hemal spines, anterolat- eral view (SEM) 47 45. Gagata gagata, hemal spines and stridu- latory apparatus (SEM) 48 46. Conta conta, posterior neurocranium, Weberian complex, and dorsal fin sup- port, dorsal view 49 47. Gagata gagata, dorsal fin support and vertebrae, lateral view 50 48. Psendecheneis sulcatus, adhesive appa- ratus (SEM) 51 49. Pseudecheneis sulcatus. adhesive appa- ratus (SEM) 52 50. Conta conta, adhesive apparatus (SEM) 53 5 1 . Laguvia ribeiroi, adhesive apparatus (SEM) 54 52. Most parsimonious hypothesis of siso- roid relationships 55 53. Famihal cladogram based on hypothe- sized relationships 58 54. Familial and subfamilial classification superimposed on most parsimonious cladogram 58 55. Nangra nangra, right palatine, lateral and ventral views 69 List of Tables 1. Data matrix for superfamily Sisoroidea ..56 IV A Phylogenetic Analysis of the Asian Catfish Families Sisoridae, Akysidae, and Amblycipitidae, with a Hypothesis on the Relationships of the Neotropical Aspredinidae (Teleostei, Ostariophysi) Mario C. C. de Pinna Abstract A cladistic analysis based on 1 1 2 morphological characters was conducted for the Asiatic catfish families Sisoridae, Akysidae, and Amblycipitidae and the Neotropical family Aspredi- nidae. Sisoridae, Akysidae, and Amblycipitidae were divided into nearly all their currently recognized genera for the analysis, but Aspredinidae was treated as a single taxon, for a total of 20 terminal taxa. The single, fully resolved most parsimonious cladogram requires numerous modifications of the previous suprageneric classification of these fishes. The Sisoridae of pre- vious authors was found to be a paraphyletic assemblage because a subunit of it is more closely related to the Aspredinidae than to remaining taxa so far allocated in Sisoridae. Therefore, the current Sisoridae is split into the monophyletic Sisoridae (sensu stricto) and Erethistidae, with Erethistidae hypothesized as the sister group to Aspredinidae. Akysidae (including Parakysis) is demonstrated to form a monophyletic group and to be the sister group to the clade Sisoridae {sensu stricto) + Erethistidae + Aspredinidae. Within Akysidae, relationships hypothesized are (Akysis (Parakysis (Breitensteinia, Acrochordonichthys))). Amblycipitidae is also shown to be monophyletic and is the sister group to all other families treated in the analysis. Sisoridae is composed of two major lineages, one represented by (Bagarius (Sisor (Nangra, Gagata))) and the other by {Glyptothorax (Pseudecheneis, "glyptostemoids")). Relationships among the genera of Erethistidae are hypothesized as (Conta ((Laguvia, Pseudolaguvia) (Erethistoides (Erethistes, Hard)))). On the basis of the hypothesized relationships, a new classification of the whole group, named superfamily Sisoroidea, is proposed, from the level of family down to tribe and subtribe in some cases. Formal diagnoses are provided for all suprageneric clades proposed, with synon- ymies added for all named taxa. Evidence supporting the monophyly of some genera is also offered, along with a discussion of the possible paraphyly of some others, such as Akysis and Glyptothorax. The evolution of two functional complexes, namely the stridulatory mechanism and the adhesive apparatus, is interpreted in light of the hypothesized relationships. Events of minia- turization and gigantism in Erethistidae and Sisoridae, respectively, are also discussed and compared with similar cases in other ostariophysans. The placement of the Neotropical As- predinidae within an otherwise exclusively Asian clade indicates a trans-Pacific biogeographic component. The implications of this hypothesis are discussed. Resumo Uma analise cladistica baseada em 1 1 2 caracteres morfologicos e apresentada para as familias asiaticas Sisoridae, Akysidae, Amblycipitidae e para a familia Neotropical Aspredinidae. Os FIELDIANA: ZOOLOGY, N.S., NO. 84, JULY 31, 1996, PP. 1-83 sisorideos, akysideos e amblycipitideos foram subdivididos em praticamente todos os seus generos atualmente reconhecidos para fins de analise, enquanto que os aspredinideos foram tratados como um unico taxon, perfazendo um total de 20 terminals. O unico cladograma mais parcimonioso encontrado para os dados requer varias modificafoes na atual classificagao destes peixes. Os Sisoridae de autores previos formam um grupo parafiletico, uma vez que um de seus subgrupos esta mais proximo a Aspredinidae que a outros taxa tradicionalmente alocados em Sisoridae. Em fun^ao disto, os atuais Sisoridae sao divididos nas familias monofileticas Sisoridae (sensu strictd) e Erethistidae, sendo a ultima proposta como grupo irmao de Aspre- dinidae. Os Akysidae (incluindo Parakysis) sao demonstrados formar um grupo monofiletico, e formam o grupo irmao de Sisoridae {sensu strictd) + Erethistidae + Aspredinidae. Dentro de Akysidae, a hipotese de relafoes e {Akysis {Parakysis {Breitensteinia, Acrochordonichthys))). A familia Amblycipitidae, tambem demonstrada monofiletica, forma o grupo irmao de todas as outras familias tratadas nesta analise. Dentro de Sisoridae, duas grandes linhagens sao reconhecidas, uma representada por {Bagarius {Sisor {Nangra, Gagata))) e a outra por {Glyp- tothorax {Pseudecheneis, "glyptostemoids")). A hipotese de relafoes dentro de Erethistidae e {Conta {{Laguvia, Pseudolaguvia) {Erethistoides {Erethistes, Hara)))). Com base na hipotese filogenetica proposta, uma nova classifica9ao do grupo inteiro, chamado superfamilia Sisoroidea, e formulada do nivel de familia ao de tribo e subtribo em alguns casos. Diagnoses formais sao oferecidas para todos os taxa supragenericos, com sinonimias para todos OS taxa nomeados. Sinapomorfias corroborando monofiletismo de alguns generos tambem sao propostas, assim como uma discussao sobre o possivel parafiletismo de alguns outros, como Akysis e Glyptothorax. A evolu9ao de dois complexos funcionais, a saber, mecanismo estridulatorio e aparelho adesivo, e interpretada a luz da nova hipotese de relafoes filogeneticas. Eventos de miniaturi- za9ao e gigantismo em Erethistidae e Sisoridae, respectivamente, tambem sao discutidos e comparados com casos similares em outros Ostariophysi. A inclusao da familia Neotropical Aspredinidae em um grupo que de outra forma e exclusivamente asiatico indica um componente biogeografico transpacifico. As implica9oes desta hipotese sao discutidas. Introduction The families Sisoridae, Akysidae, and Ambly- cipitidae comprise a major portion of the catfish fauna in South and Southeast Asia. Members of these three families have been informally consid- ered to be closely related for a long time (e.g.. Day, 1871; Bridge & Haddon, 1893; Chaudhuri, 1919), and in several aspects they represent the most typ- ical exclusively Asian siluriform assemblage. Else- where within the order Siluriformes, a greater de- gree of morphological and ecological diversifica- tion is found only in the Neotropics. Sisoridae is the largest and most diverse of the three families, containing 25 genera and approximately 120 spe- cies, with new species being discovered frequently (e.g., Kottelat, 1 983; Mo & Chu, 1 986; Ding et al., 1991; Zhou & Chu, 1992). Sisorids are also the most widely distributed of the three families, oc- curring throughout nearly the whole of South and Southeast Asia, from Iran and Turkey in the west (Coad, 1981; Coad & Delmastro, 1985) to Sun- daland, including the Indian subcontinent, Malay Peninsula, Java, Borneo, Sumatra, and part of China. Akysidae is a highly distinctive but obscure family of only four genera and about 1 5 species, virtually all of which are poorly known and usually are represented by few specimens in museum col- lections. An overview of their most speciose genus, Akysis, was done by He and Chen (1981). Akysids occur in the Malay Peninsula and adjacent areas, Java, Borneo, and Sumatra. The genus Parakysis, containing only two rather similar species from Borneo, has been traditionally included in Aky- sidae but was recently assigned its own family by Roberts (1989). A separate Parakysidae has been accepted by some authors (e.g., Eschmeyer, 1990; Nelson, 1994) but not others (e.g., Mo, 1991), a controversy that stems at least in part from lack of knowledge about the anatomy and phylogenetic relationships of these fishes. The Amblycipitidae, with two genera and about 10 species, occurs in FIELDIANA: ZOOLOGY the Indian subcontinent, East Asia, Malay Pen- insula and adjacent areas, Borneo, Java, and Su- matra. Sisorids have been the object of a relatively ex- tensive taxonomic and anatomical study. Never- theless, the family has never been subjected to an investigation using strictly phylogenetic method- ology. With the exception of one isolated effort with the genus Pseudecheneis (Chu, 1982), cladis- tics has never been applied to any level of the taxonomy of the Sisoridae. Its various genera and subfamilies, and even the family as a whole, re- main recognized on the basis of informal criteria, and their monophyly has not yet been rigorously proposed or tested. The same situation holds for amblycipitids (but see a recent contribution by Chen & Lundberg, 1995) and akysids. In view of the wide distribution of the group, a phylogenetic framework for sisorids and their rel- atives would be an important source of data rel- evant for understanding the biogeography of a large portion of Asia. Also, the great morphological variation in the group is certainly a rich ground for studies on the evolutionary significance of clade diversification. However, just as with biogeogra- phy, this area of research awaits investigation of the higher-level systematics of the Sisoridae, Aky- sidae, and Amblycipitidae. A number of obstacles have hindered cladistic analyses of the Sisoridae and their closest relatives. First, many members of these groups are extreme- ly rare in collections, and gathering representative material for anatomical studies is difficult. The internal anatomy of some genera, such as Breiten- steinia and Acrochordonichthys (Akysidae), was virtually unknown prior to this study. Even for genera with several nominal species, it is some- times impossible to obtain skeletal preparations for more than one or two individuals. Not unex- pectedly, ontogenetic data are almost nonexistent for the group. Also, the absence of obvious out- groups closely related to sisorids and their relatives has prevented anatomical comparisons from being conclusive phylogenetically. This study started as an analysis of Sisoridae; however, preliminary investigations revealed rel- atively weak support for sisorid monophyly and very strong support for a group composed of Si- soridae and Akysidae, plus some evidence for a more inclusive clade with Amblycipitidae. The most effective way to circumscribe and address the problem was an analysis encompassing the three families. The scope of the analysis was further expanded by the inclusion of Aspredinidae. This Neotropical family was first explicitly proposed as closely re- lated to Asiatic taxa (in particular Akysidae) by Ferraris (1989). Later, a number of derived sim- ilarities between aspredinids and various akysids and sisorids were recorded by de Pinna (1993). The Aspredinidae, with 10 genera and ca. 35 spe- cies widely distributed in South America, have long been a family of uncertain phylogenetic re- lationships. These fishes have been previously proposed as relatives of Neotropical loricarioids (Chardon, 1967, 1968), a hypothesis subsequently refuted (Baskin, 1973; Howes, 1983; Schaefer & Lauder, 1986) but not replaced by an alternative. Evidence previously reported plus additional data disclosed in a preliminary phase of this work were combined to support the working hypothesis that Amblycipitidae, Akysidae, Sisoridae, and Aspre- dinidae form a monophyletic ingroup. Although it was clear at that point that aspredinids were related to some level of the sisorid-akysid-ambly- cipitid problem, their exact position remained un- certain. One point that seemed certain, though, was that a phylogeny of the three Asian families without Aspredinidae would be incomplete and that the only consistent way to approach the prob- lem would be to include aspredinids as a terminal taxon in the study. Accordingly, the objectives of this work were to provide a cladistic framework for Sisoridae and its closest relatives, including aspredinids, and to propose a new higher-level classification reflecting the hypothesized relationships. This work is not intended to be a generic-level revision but instead is a first attempt at a phylogenetic understanding of Sisoridae, Amblycipitidae, Akysidae, and As- predinidae, with a hypothesis on the monophyly of the whole group and a definition of its major subunits. The investigation focused on Sisoridae and Akysidae, which accordingly were more finely subdivided and better represented in the material examined. More in-depth studies of amblycipitids and aspredinids are currently being undertaken by X. Chen (Chen, 1994) and J. Friel (Friel, 1994), respectively. Materials and Methods Phylogenetic Methodology The systematic methodology employed for pro- posing hypotheses of relationships was cladistic, or phylogenetic, as first proposed by Hennig (1950, 1 966, 1 968). Cladistics has undergone substantial DE PINNA: ASIAN CATFISH FAMILIES developments since Hennig's original formulation and is presently considered the standard approach for studying interrelationships among organisms. Detailed explanations of cladistic principles and foundations are currently available in a number of textbooks (e.g., Eldredge & Cracraft, 1980; Nel- son & Platnick, 1981; Wiley, 1981;Schoch, 1986; Rieppel, 1988). Recent general works focusing on more technical methodological aspects include Wiley et al. ( 1 99 1 ) and SwofTord and Olsen ( 1 990). Parsimony was employed to find the hypothesis best supported by the data set. Wagner (strict or simple) parsimony was used, in which character reversals and convergences are permitted and counted equally (Farris, 1983; Swofford & Mad- dison, 1987). Parsimony analyses were performed with the Hennig86 computer program (Farris, 1988). This program was chosen mainly because of its efficiency in calculation time. Specific infor- mation on the options and capabilities of the pro- gram were provided by Fitzhugh (1989). (See the "Phylogenetic Analysis" section for specific Hennig86 commands employed in this study.) Tree manipulations and diagnostics were done with the help of the Clados computer program (Nixon, 1992). Trees longer than the most parsimonious trees were obtained and examined with the pro- gram PAUP (SwofFord, 1 989), which was also used to check the most parsimonious result found with Hennig86. Multistate characters were ordered according to similarity among character states whenever pos- sible. Outgroup comparison provided the ple- siomorphic end of the series, and other states were ordered sequentially according to their divergence from that putatively primitive condition. There- fore, more extreme expressions of a derived con- dition were considered evidence of monophyly for progressively less inclusive groups. This procedure is considered appropriate because similarity is the basis for any homology propositions, and the pro- cess of ordering character states is equivalent to the recognition of primary homologues. Polarity for character states was determined by outgroup comparison. For the most part, the out- groups considered were all other catfish families available for examination. In cases where the state in outgroups varied, the algorithm of Maddison et al. (1984) was employed to decide about an optimal assignment for the plesiomorphic condi- tion for the ingroup. The siluriform phylogeny pre- sented by de Pinna (1993) served as a baseline for such determinations. In a few instances, ontoge- netic sequences provided some information on the direction of character- state transitions. Such cases were few, though, because juvenile specimens of the taxa studied are rarely available for study. Terminal Delimitation It was impossible, for practical reasons, to sep- arately analyze representatives of all known sisor- id, akysid, and amblycipitid species. Therefore, some a priori decisions had to be made about the relevant terminal taxa to be considered in the anal- ysis. Because no prior information was available on the phylogeny of sisorids and their relatives, a conservative overdivisive approach was adopted to decrease the chances of having nonmonophy- letic terminals. This subdivision was also intended to decrease the number of missing entries due to terminal polymorphism, a procedure demonstra- bly problematic in current implementations of parsimony analysis (Nixon & Davis, 1991). The usual bottom line for subdivision was the presence of some conspicuous, obviously apomorphic trait(s) indicating that a certain assemblage of spe- cies likely composed a monophyletic group. For the most part, the sisorid genera were considered satisfactory as a starting point. Most genera in the family either are monotypic or include few species, and they are usually relatively well circumscribed by one or more putatively synapomorphic traits.. The most striking exception is the genus Glypto- thorax, with ca. 40 species (Li, 1986), which has served as a taxonomic wastebasket for species lacking the evident diagnostic characters of other genera. At present, there is little reason to believe Glyptothorax is monophyletic, but the taxonomic representation available for this study was not suf- ficient to tackle this problem. Some evidence dis- closed in this work suggests that the species of Glyptothorax examined form a monophyletic group. Whether these characters are actual syna- pomorphies for the genus or only for a subgroup thereof is a question that will have to await a more complete investigation of the genus. Laguvia and Pseudolaguvia, occasionally considered as syn- onyms oT Glyptothorax (e.g., Menon, 1954), were treated as separate terminals. At the other extreme are the genera and species usually grouped together as the "glyptostemoid" sisorids (e.g., Hora & Silas, 1952a,b; Tilak, 1963, 1 976; Chu, 1 979; equivalent in composition to the subtribe Glyptostemina proposed in this study), which were found early in the study to form a highly corroborated monophyletic clade and were therefore treated as a single terminal. HELDIANA: ZOOLOGY As with Sisoridae, monophyly of Akysidae and Amblycipitidae has not previously been demon- strated or tested, and these two famiHes were ac- cordingly split into their presently recognized gen- era. Characters supporting monophyly of Aspre- dinidae were proposed by de Pinna (1993), and the family was therefore treated as a single ter- minal. In summary, the following terminals (numbered sequentially) were considered. Akysidae 1. Acrochordonichthys 2. Akysis 3. Breitensteinia 4. Parakysis Amblycipitidae 5. Amblyceps 6. Liobagrus Sisoridae 7. Bagarius 8. Conta 9. E ret hist es 10. Erethistoides 1 1 . Gagata 12. glyptostemoids 13. Glyptothorax 14. Hara 15. Laguvia 16. Nangra 17. Pseudecheneis 18. Psendolaguvia 19. Sisor Aspredinidae 20. Aspredinidae al., 1986; Carpenter, 1988) is not considered un- desirable or "artificial" (Yeates, 1992). Abbreviations The following abbreviations are used for insti- tutions and anatomical terminology. AMNH = American Museum of Natural History, New York ANSP = Academy of Natural Sciences of Philadelphia BMNH = The Natural History Museum, London CAS, CAS-SU, CAS-iu = California Academy of Sciences, San Francisco CMK = Collection Maurice Kottelat, for- merly housed at ZSM, Munich Duvc = Duke University Vertebrate Col- lection, Department of Zoology, Duke University, Durham, N.C. FMNH = Field Museum of Natural His- tory, Chicago LACM = Los Angeles County Museum of Natural History Mzusp = Museu de Zoologia da Univer- sidade de Sao Paulo, Sao Paulo NRM = Naturhistoriska Riksmuseet (Swedish Museum of Natural History), Stockholm UMMZ = Museum of Zoology, University of Michigan, Ann Arbor USNM = National Museum of Natural History, Smithsonian Institu- tion, Washington, D.C. ZRC = Zoological Reference Collection, National University of Singapore ZSM = Zoologische Staatssammlung, Munich AC (The exact composition of each of the above ter- ALP minals is presented in the list of material exam- ined.) ANP Autapomorphies for the terminals were actively BAS searched for and included in the analysis. These BB,^ autapomorphies should help define smaller mono- C,., phyletic subunits of the family, therefore provid- CC ing important background for future taxonomic CF and revisionary works. The inflation of the con- CL sistency index that is associated with inclusion of CO autapomorphies in parsimony analyses (Brooks et DC anterior ceratohyal anterolateral process of third hypobran- chial anterior nuchal plate basioccipital basibranchials 1-4 ceratobranchials 1-5 complex centrum cranial fontanel cleithrum coracoid distal condyle of first basal radial DE PINNA: ASL\N CATFISH FAMILIES DEN = dentary DF,_2 = dorsal fin spine (1,2) DR = distal pectoral radials E,^ = epibranchials 1-4 EA = external arm of basipterygium END = endopterygoid F = frontal FR = fin rays H,_3 = hypobranchials 1-3 HS = hemal spine HVL = humerovertebral ligament HY = hyomandibula HYP = hypohyal lA = internal arm of basipterygium lO = infraorbital LAC = lacrimal LAT = lateral ethmoid LIG = ligament LLO = lateral-line ossicle ME = mesethmoid MET = metapterygoid MNP = medial nuchal plate N = nasal NS = neural spine OP = opercle ORB = orbitosphenoid OS = OS suspensorium PAL = palatine PAR = parasphenoid PC = posterior ceratohyal PCF = paired cranial fontanel PDR = proximal dorsal fin radial PLR = pleural rib PMX = premaxilla PNP = posterior nuchal plate POP = preopercle PR = proximal pectoral radial PS = pectoral spine PTE = pterotic QUA = quadrate SBE = supernumerary basibranchial element SC = sensory canal SPH = sphenotic SU = supracleithrum SOC = supraoccipital SPO = suprapreopercle STF = supratemporal fossa TP = tooth plate TPR = transverse process TRI = tripus V = vomer VC„ = vertebral centrum n VCB = ventral crest of basipterygium VP„ = vertebral parapophysis n VPD = ventral process of first dorsal fin spine WL = Weberian lamina Material Examined Material representative of ingroup taxa directly examined for this study is listed below. Family, genus, and species names are ordered alphabeti- cally. The number of specimens indicated (ex) re- fers to those examined for this study, not neces- sarily the total number in the respective lot. The material was cleared and stained for bone and car- tilage (BC), cleared and stained for bone only (B), radiographed (R), or dry skeleton (SK). If no des- ignation is applied to the number of specimens, it refers to alcohol lots from which no special prep- aration was made. If the number of specimens (ex) exceeds that indicated for skeletal preparations (BC, R, etc.), the difference represents alcohol-pre- served specimens examined. Holotypes (h), para- types (p), and syntypes (s) are indicated immedi- ately following the catalogue number. Because of the highly heterogeneous sources of material, it was impossible to provide standard- ized locality citations. Whenever available, infor- . mation on the country is provided first, followed by any additional locality data, including the ma- jor drainage system. Any additional information bearing on special conditions of certain specimens is also provided (e.g., disarticulated, uncalcified, etc.). Published accounts provided data for some outgroup and ingroup taxa not directly examined. As sources of data for outgroups, the most im- portant papers consulted were by Arratia (1987), Ferraris (1988), Grande (1987), Lundberg (1970, 1982), Mago-Leccia (1978), Schaefer (1990), Sie- bert (1987), Skelton et al. (1984), Vari (1979), and Weitzman (1954, 1962). Publications on ingroup taxa provided information on species not avail- able for direct examination and were also used to verify the consistency of some characters in species from which only one or few specimens were ex- amined. The most important papers consulted were Chu (1982), Gauba (1962, 1966, 1967, 1968, 1969, 1970a,b), Mahajan (1966a, b, 1967a,c), Tilak (1963), and Zhou and Chu (1992). Material ex- amined for outgroups included representatives of all nominal catfish families and some other ostar- iophysan groups (for a complete list of compara- tive outgroup material, see de Pinna, 1993). FIELDIANA: ZOOLOGY Akysidae Acrochordonichthys rugosus (Bleeker)— zrc 2361a, 1 ex; Malaysia, Pahang, Taman Negara, Kuala Tahan. zrc 4415, 1 ex, 1 BC; no data. Acrochordonichthys ichnosoma Bleeker— mzusp 42468, 3 ex, 1 BC; Indonesia, Kalimantan Tengah, Barito River drainage. Akysis leucorhynchus Fowler— usnm 109636, 5 ex, 1 B, 1 BC; Thailand, north, Meping River at Chiengmai. Akysis macronemus Bleeker— bmnh 1980. 10. 10: 188-192, 1 ex, 1 BC; no data. Akysis pseudobagarius Roberts— ummz 155702 (p), 1 ex; Indonesia, Sumatra, Moesi River, Moeara Klingi. Akysis sp. 1-ummz 214907, 10 ex, 3 BC, 1 B; Vietnam, Chau Doc, Chau Doc River, 7 km up- stream from Chau Doc, CD- 7, Mekong River drainage. Akysis %^. 2— cmk4296, 1 ex, 1 B; Kampuchea, km 1 17, route de Sianoukville. Akysis sp. 3— usnm 232930, 1 ex, 1 BC (juve- nile); Thailand, northeast, Nakhon Ratchasima Province, Lam Nam Mun, 1 km below dam and ca. 2 km downstream of Phimai. Breitensteinia insignis Steindachner— amnh 58378, 4 ex, 1 BC, 1 SK; aquarium material, no data. Parakysis anomalopteryx Roberts— CAS 49421 (p), 2 ex, 2 BC; Indonesia, Borneo, small stream into Kapuas River, NE of Gunung Setunggul. Parakysis verrucosa Herre- cmk 7915, 2 ex, 2 BC; Indonesia, Borneo, Kalimantan Barat, Ka- puas River basin. AMBLVaPITIDAE Amblyceps mangois (Hamilton)— ansp 59316, 1 ex, 1 BC; Thailand, usnm 208906, 10 ex, 1 BC. AMNH uncat., 1 ex, 1 BC (disarticulated); no data. Liobagrus anguillicauda Nichols— amnh 1 1069, 1 ex, 1 B; China, Fukian, Chungan, Hsien. Liobagrus reini Hilgendorf— amnh 26744, 2 ex, 1 B; Japan, Lake Biwa. usnm 089370, 4 ex, 4 B; China, Yachow Szechuan. SiSORIDAE Bagarius bagarius (Hamilton-Buchanan) — UMMZ 186793, 3 ex, 1 BC; Thailand, market at Ubol (= Ubon Ratchathani). Bagarius yarrelli (Sykes)— amnh uncat. 2 ex, 2 BC; aquarium material, no data, iacm 38294-55, 1 ex, SK; Pakistan, Karachi, West Wharf Fish Market; complete skeleton of large specimen, ca. 260 mm skull length. Conta conta (Hamilton)— ummz 208632, 10 ex, 2 BC; Bangladesh, Sylhet, Piyain Gang River (Danki River) at Songram Punji, 1/4 mi down- stream from Indian border of the Khasi and Jainta Hill region. Erethistes pusillus Muller & Troschel— ummz 208697, 10 ex, 2 BC; Bangladesh, N Sylhet, Chick- nagul Bazaar, 1 4 mi NE of Sylhet-Shillong High- way, Surma drainage. Erethistes sp.— usnm 044759, 3 ex, 1 BC; Bur- ma, Meetan (probable locality). Erethistoides montana Hora— ummz 208745, 3 ex, 1 BC; Bangladesh, N Sylhet, Gowain River and Khal at Gowainghat, Surma drainage. Euchiloglanis kishinouyei Kimura — usnm 120365, 3 ex, 1 BC; China, Szechuan Ts'Ao Po near Kuanshien. amnh 15261, 1 ex, 1 BC; China. Exostoma labiatum (McClelland)— nrm 25 105, 8 ex, 1 BC; Burma, Kachin State, Kambaiti. Exostoma sp.— usnm 130183, 3 ex, 1 BC; Chi- na. Gagata gagata (Hamilton)— amnh 8358, 1 ex, 1 B (disarticulated); Burma. Gagata cenia (Hamilton)— amnh uncat., 1 ex, 1 BC; no data. Glyptothorax major (Boulenger)— amnh uncat., 2 ex, 2 BC; no data. Glyptothorax pectinopterus McClelland— bmnh 17.10.81, 2 ex, 2 BC; India, Aglar River. Glyptothorax platypogon (Valenciennes)— usnm 087431, 6 ex, 1 BC; China, Szechuan, Suifu. Glyptothorax sinensis {Ke^n)—MXHH 10265, 2 ex, 2 BC; China, Hunan, Huping, Tungting Lake. Glyptothorax sp.— usnm 288474, 5 ex, 1 BC; Thailand, Kwai Noi River (Mekong basin), be- tween Kanchanaburi and Sai Yok. Glyptothorax telchitta (Hamilton)— ummz 208876, 1 ex; Bangladesh, Rangpur, Dharla River at Kurigram, 1/4 mi upstream from ferry ghat (dock). Glyptothorax trilineatus Blyth— bmnh 45.1.9: 850, 1 ex, 1 BC; Nepal, ummz 186849, 1 ex, 1 B; Thailand, Khlong Phra Phai, 45 km NNE Chan- taburi. ummz 208358, 1 ex; Bangladesh, Comilla, Meghna River, upstream from Chandpur, just downstream from Gumti River mouth at Kanudi. UMMZ 208458, 1 ex; Bangladesh, Dacca, Maghna River, just downstream from Gazaria, Meghna drainage. DE PINNA: ASIAN CATFISH FAMILIES Hara hara (Hamilton)— ummz 208748, 10 ex, 3 BC; Bangladesh, N Sylhet, Gowain River and Khal at Gowainghat, Surma (Meghna) drainage. FMNH 93597, 3 ex; Bangladesh, Sylhet, Gowain River Khalat Gowainghat. Hara jerdoni (Day)— amnh uncat. 4 ex, 4 BC; no data. Laguvia ribeiroi Hora— ummz 208955, 1 5 ex, 5 BC; Bangladesh, Dinajpur, Maharanda River at Tetulia, near location of Dak Bungalow, Mahar- anda (Gangetic) drainage. Laguvia shawi (Hora)— ummz 208633, 7 ex, 1 BC; Bangladesh, Sylhet, Piyain Gang River (Danki River) at Songram Punji, 1/4 mi downstream from Indian border of the KJiasi and Jainta Hill region. ummz 208788, 1 ex; Bangladesh, Sylhet, Piyain Gang River at Songram Punji, 1 .5 mi downstream from India border. Surma (Meghna) drainage. Nangra nangra (Hamilton)— cmk 6369, 5 ex, 1 BC; India, Assam, Dibru River, fmnh 93598, 1 ex, 1 BC; Bangladesh, Comilla, Meghna River, just above Chandpurat Jahasmarcihar. Nangra viridescens (Hamilton)— ummz 208725, 3 ex, 1 BC; Bangladesh, Sylhet, Sharigat Bazaar, 22 mi NE of Sylhet on Sylhet-Shillong Highway, Surma (Meghna) drainage. Oreoglanis siamensis SvaiXh—CMYi A3>5\, 1 ex, 1 BC; Thailand, Doi Inthanon. usnm 1 1 8430, 3 ex, 1 BC; Thailand, Doi Angka. Pseudecheneis sulcatus (McClelland)— fmnh 99630, 4 ex, 1 BC; Nepal, Shebu, Sabhaya Khola, Arun River, bmnh 1985.9.16:52, 1 ex, 1 BC; Ne- pal, Naryani River. Pseudexostoma yunnanensis (Tchang)— nrm 25124, 5 ex, 1 BC; China, Yunnan, Irrawaddy River drainage on road to Tengchong-Burma bor- der. Pseudolaguvia tuberculata (Prashad & Muker- ji)-UMMZ 209010, 5 ex, 1 BC; Bangladesh, Chit- tagong, Koilla Khal (creek), 6 mi E of Feni-Chi- tagong Highway on road to Ramgarh. SisorrhabdophonisWdiVoxXxon—hMtfH 1970.6.25: 2, 1 ex, 1 BC; India, Kalinadi, M. Nagar. ASPREDINIDAE Agmus sp.— AMNH uncat., 2 ex, 2 BC; no data. AMNH uncat., 1 ex, 1 SK; no data. Amaralia hypsiura (Kner)— amnh uncat., 1 ex, 1 BC; no data. Aspredinichthys filamentosus (Cuvier &. Valen- ciennes)—usnm 207452, 2 ex, 2 B; Brazil, offshore, 02°34'N, 49°18'W. Aspredinichthys tibicen (Temminck)— amnh 7094, 1 ex, 1 B; Guyana, Georgetown. Aspredo aspredo (Linnaeus)— usnm 226072, 2 ex, 2 BC; Suriname. Bunocephalus coracoideus (Cope)— amnh 21815, 1 ex, 1 BC; no data. Ernstichthys sp.—MZUSP 37814, 5 ex, 1 BC; Bra- zil, State of Mato Grosso, Aripuana River, down- stream from Dardanelos. Hoplomyzon sexpapilostoma Taphom & Mar- rero— amnh uncat., 1 ex, 1 B; no data. Xiliphius melanopterus—D\JVcT-3\-19, 1 ex, 1 BC; Venezuela. Preparation Techniques and Nomenclature Specimens were prepared for skeletal observa- tion by the clearing and staining procedures of Dingerkus and Uhler (1977) and Taylor and Van Dyke (1985), which allow visualization of both bone and cartilage. Several of the specimens ex- amined were prepared prior to this study and were stained for bone only. The outline of cartilaginous structures in these specimens could usually be traced by their slightly different diffraction index in glycerol under transmitted light. In some cases, for large and/or old specimens, some modifica- tions of the methods mentioned above had to be applied to accomplish good preparations (see de Pinna, 1993). In view of the rarity of some nom- inal forms, special dissecting techniques were em- ployed to expose all relevant anatomical structures while preserving intact as many articulations as possible. In these cases, the procedure described by Weitzman (1974) was followed, with some mi- nor alterations to fit the peculiarities of catfish anatomy. Anatomical illustrations were prepared with the aid of a camera lucida attached to a Wild stereo- microscope. Addition and checking of details were done freehand under higher magnifications and different angles of observation. All illustrations were prepared by the author. Nomenclatural decisions follow the rules and recommendations of the International Code on Zoological Nomenclature (ICZN, 1985). The grammar of formation of family-group names was guided by the suggestions of Steyskal (1980). FIELDIANA: ZOOLOGY Character Description and Discussion The characters Hsted below are separated ac- cording to the main region of the body where they are observed, arranged roughly in the order they are found from anterior to posterior. The account on each character begins with a sentence-like ab- breviated description of the derived condition, fol- lowed by CI and RI (consistency and retention indices, respectively) values for the character un- der the most parsimonious hypothesis. A list of the terminal taxa with the apomorphic condition is presented in the line immediately below, fol- lowing the order presented in the subsection "Ter- minal Delimitation." A more detailed description and discussion are presented in a separate para- graph, where the plesiomorphic condition and the justification for the hypothesized polarity are ex- plained, along with any other p)ertinent observa- tions or interpretations regarding the character. Taxa assigned missing entries are shown within brackets, with the indication "= ?" immediately following the name (e.g., [Sisor = ?]). Missing en- tries may be assigned for a number of reasons (Nixon & Davis, 1991; Platnick et al., 1991), and each case is explained in the discussion of the re- spective character. Most characters are skeletal, the only system that could be studied in the majority of the ter- minal taxa using the material currently available. Preliminary data indicate that a significant amount of phylogenetically informative variation also ex- ists in myology (cf Saxena & Chandy, 1 966). How- ever, it is presently impossible to adequately study the muscular system in many members of the in- group, because of the unavailability of study ma- terial for many forms. With present knowledge, including these potential characters in the analysis would add such a large number of missing entries in the data matrix as to possibly obscure the phy- logenetic signal from other characters. Incorpo- ration of data from the muscular system in an analysis of the Sisoridae and their relatives will only be possible after additional material of their several rare forms becomes available. B Fig. 1. Dysichthys sp., amnh uncat., right palatine, anterior to right. A, lateral view. B, dorsal view. Larger stippling represents cartilage. Scale bar = 1 mm. Acrochordonichthys, Akysis, Parakysis, Ambly- ceps, Liobagrus, Bagarius, Conta, Erethistes, Er- ethistoides, Gagata, glyptostemoids, Glyptothor- ax, Hara, Laguvia, Nangra, Pseudecheneis, Pseu- dolaguvia, Sisor, Aspredinidae. The plesiomorphic morphology of the posterior part of the palatine in siluriforms is that of a cy- lindrical ossification, roughly round in cross-sec- tion, with a short, cylindrical cartilage at the pos- terior tip. This is the condition seen in diplomys- tids, pimelodids, bagrids, and the vast majority of remaining siluriforms. In the apomorphic condi- tion in the taxa above, the posterior portion of the palatine is laterally compressed and dorsoven- trally expanded. The distal cartilage is propor- tionally enlarged and covers the posterior margin of the bone (Fig. 1 ). The expansion of the posterior portion of the palatine involves a modification in the shape of the ossification center of the bone, and not an enlargement of peripheral bone alone. This detail distinguishes the condition in the taxa above from the superficially similar state in the Amphiliidae and some Auchenoglanidinae (e.g., Notoglanidium), where the compressed shape is the result only of the vertical expansion of pe- ripheral bone, with the cartilaginous center re- maining cylindrical. Jaws and Related Structures 1 . Posterior center of ossification of palatine com- pressed and expanded vertically (CI = 50, RI = 0). 2. Mandibular canal entering dentary at its mid- section (CI = 100, RI = 100). Acrochordonichthys, Breitensteinia, Parakysis, [Conta = ?], [glyptostemoids = ?], [Sisor = ?], [As- predinidae = ?]. DE PINNA: ASIAN CATFISH FAMILIES Fig. 2. Pseudexostoma yunnanensis, nrm 25 1 24, right maxilla, dorsoposterior view. Scale bar = 1 mm. The mandibular laterosensory canal in catfishes primitively enters the dentary through its ventral margin, which coincides with the anterior portion of the ventral margin of the lower jaw. In the de- rived condition, the canal enters the dentary at a point about the mid-depth of the posterior part of the lower jaw. The canal gradually approaches the ventral margin of the dentary as it extends ante- riorly, becoming similar to the plesiomorphic con- dition in siluriforms along the anterior half of the lower jaw. Taxa assigned missing entries are those lacking the mandibular branch of the laterosensory canal system. 3. Mandibular laterosensory canal absent (CI = 25, RI = 0). Conta, glyptostemoids, Sisor, Aspredinidae. The mandibular portion of the laterosensory ca- nal system, primitively present in most siluriforms and remaining ostariophysans, is apomorphically absent in the taxa above. In these taxa, the ventral branch of the preopercular laterosensory canal ter- minates ventrally at a pore near the posterior end of the lower jaw, not penetrating the dentary or anguloarticular. A peculiar situation is seen in as- predinids, where the dentary lacks a sensory canal but there is a branch in the soft tissue along the lower jaw. This branch often makes strong curves that do not match the profile of the dentary, in- dicating that its development is independent of the limits of the bone. Also, the dentary in aspre- dinids is a large and well-ossified bone, indicating that the canal alongside its ventral margin is not free because of bone reduction. Therefore, this ca- nal probably represents a secondary extension of the ventral portion of the preopercular canal rather than a primitive mandibular branch. Examination of ontogenetic series of aspredinid species, not available for this study, will be the only way to test that assumption. 4. Distal tip of maxilla pointed (CI = 50, RI = 50). Bagarius, Gagata. Nangra. The distal end of the maxilla in siluriforms primitively has a transverse, round, or oblique edge, with a laminar morphology similar to that of the rest of the bone. In the taxa listed above, the distal portion of the maxilla is modified into a nonlamillate solid rod, with a sharp pointed tip. In some forms with the derived condition (e.g., Bagarius, Gagata cenia, Nangra nangra), the non- laminar portion of the maxilla is also extremely elongate, and the longest axis of the bone is some- times as long as the head length. The variation in length of the bone is also possibly informative about relationships but could not be adequately quantified for this analysis. 5 . Maxilla dorso ventrally flat and club-shaped (CI = 100). Glyptostemoids. The plesiomorphic morphology of the maxilla in siluriforms is rather complex, composed of a bifurcated proximal end articulating with the an- terior cartilage of the palatine and a distal portion (frequently laminar, but with shelves in three di- mensions) that expands slightly at its midlength and then narrows again toward the distal tip. In all glyptostemines, the distal portion of the maxilla is markedly flat and expanded evenly toward the relatively blunt distal end (Fig. 2). 6. Interopercle narrow and elongate, its lamillate portion reduced or absent (CI = 100). Glyptostemoids. The interopercle in most siluriforms and other ostariophysans is a roughly triangular bone, with its posterior region much deeper than its anterior end. Other than for its anteriormost part, the bone is lamillate. In glyptostemoids, the interopercle is modified into a narrow and horizontally elongate structure, and most or all of its area is thick and nonlamillate. 7. Premaxillary dentition reduced or absent in adults (CI = 50, RI = 0). Gagata. Sisor, [Aspredinidae = ?]. The premaxilla in catfishes primitively bears bands of numerous teeth over its ventral surface. 10 FIELDIANA: ZOOLCXjY Fio. 3. Gagata gagata, amnh 8358, premaxillae, an- terior view. Scale bar = 1 mm. In the derived condition, premaxillary teeth are absent in the adult stages (Fig. 3). It has been re- ported that young specimens oiSisor have at least some premaxillary teeth (Mahajan, 1 966b; p. 377), which are lost later in ontogeny. Whether or not this observation applies to other taxa remains to be determined. In aspredinids, the premaxilla is edentulous in bunocephalines and toothed in as- predinines, hence the missing entry for the taxon. 8. Premaxillae with well-defined elevated facet for articulation with mesethmoid (CI = 100, RI = 100). Gagata, Nangra, Sisor. The dorsal surface of the premaxilla in siluri- forms is primitively connected to the mesethmoid without well-defined processes or facets. In the apomorphic state, the premaxilla has a large fac- eted process dorsally. This structure displays a large degree of variation among the three taxa. In Sisor, the facet is broad and short, taking up most of the dorsal margin of the premaxilla. In Gagata, the facet is at the posterior surface of a large ascending process (Fig. 3) and is connected to the meseth- moid by a long cartilage. The facet is well devel- oped in Nangra viridescens but is totally absent in N. nangra, a condition assumed to be a result of the marked reduction in the size and degree of ossification of the premaxilla in that species and not representative of the primitive condition in the genus. A superficially similar ascending process and facet are present in the erethistid Conta. However, in that genus the facet is located on the anterior surface of the ascending process instead of in the posterior surface, as in all other taxa with the de- rived condition. The condition in Conta is there- fore considered nonhomologous. B Fig. 4. Right palatines, ventromesial view. A, Akysis sp. 2, CMK 4296. B, Parakysis verrucosa, cmk 79 1 5. Larg- er stippling represents cartilage. Scale bar = 1 mm. 9. Posterior half of palatine bent laterally (CI = 100, RI = 100). [Akysis = ?], Parakysis. In the plesiomorphic condition, the palatine in siluriforms is mostly straight from its anterior to its posterior ends as seen in dorsal view. In diplo- mystids, the bone can be slightly and gradually curved medially (Arratia, 1992). In the apomor- phic state, the posterior portion of the palatine (posterior to its articulation with the lateral eth- moid) is abruptly bent laterally relative to the an- terior portion (Fig. 4). Among the species of Akysis examined, the character is absent in A. leucorhyn- chus and Akysis sp. 1 but present in remaining representatives of the genus. This spotty distri- bution makes the primitive condition of the char- acter uncertain for this taxon, which is therefore assigned a missing entry. 10. Premaxilla subdivided (CI = 50, RI = 0). Bagarius, [glyptostemoids = ?], Glyptothorax. The premaxilla in siluriforms and other teleosts is primitively composed of a single bony piece. In Glyptothorax and Bagarius, the distal portion of the premaxilla (usually directed posteriorly) is a DE PINNA: ASIAN CATFISH FAMILIES 11 Fio. 5. Outline of premaxillae, teeth not represented. A, Glyptothorax major, amnh 58410. B, Euchiloglanis kishinouyei, ltsnm 120365. Scale bars = 1 mm. separate plate (Figs. 5, 6) connected to the re- mainder of the bone by connective tissue (this characteristic was first noted in Bagarius by Gauba [ 1 962] and in both Bagarius and Glyptothorax by Tilak [1963]). The anterior margin of the distal piece is usually convex and fits into a correspond- ing concavity along the posterior margin of the proximal piece. Despite the arrangement, the two portions of the premaxilla do not seem to form a movable articulation and are firmly united to each other by rigid connective tissue. The covering of teeth is uniform and uninterrupted across the two pieces. In Bagarius. the two pieces overlap (but do not ankylose) in adults (Fig. 6), but in Glypto- thorax they do not (Fig. 5A). Only one glypto- stemoid examined {Euchiloglanis kishinouyei) shows the derived condition for this character (the species also uniquely displays further fragmenta- tion of the lateral piece, which results in a tripartite or multipartite premaxilla; Fig. 5B). Elsewhere in glyptostemoids, a subdivided premaxilla has also been reported in Glyptosternum reticulatum (Gau- ba, 1969). 1 1 . Presence of incisiform teeth on jaws (CI = 100, RI = 100). Glyptostemoids, Pseudecheneis. The primitive condition for jaw dentition in cat- fishes is to have simple conical teeth. In a few taxa there are deviations from that plesiomorphic mor- phology, such as in Neotropical loricarioids and most African mochokids, which may have bifur- Fio. 6. Bagarius yarrelli. lacm 38294-55, right premaxilla, dorsal view. 12 FIELDIANA: ZOOLOGY MET Fig. 8. Amblyceps mangois. ansp 59316, metapter- ygoid and endopterygoid, ventromesial view. Wavy lines represent ligaments. Scale bar = 1 mm. Fto. 7. Dentary teeth. A, Pseudecheneis sulcatns, BMNH 1985.9.16:50. B, Pseudexostoma yunnanensis, nrm 25 1 24. Scale bar = 1 mm. the maxilla points almost directly ventrally when partly abducted. Gated or spatulate teeth. Within the clade under study, a conspicuous derived dental morphology is seen in Pseudecheneis and most glyptostemoids. The jaw teeth in these taxa are strongly incisiform and usually concave in one surface (Fig. 7). Also, their distal edge is frequently straight. The teeth are conical in a few glyptostemoids, but even in these the enamel cap at the tip of each tooth is slightly larger and more prominent than in the primitive condition. These cases are postulated as partial reversals within the glyptostemoid clade. Comparative tooth morphology in some glypto- stemoids was illustrated by Chu (1979, p. 75). 1 2. Direction of main axis of maxilla, when partly abducted, with strong ventral component (CI = 100, RI = 100). Gagata, Nangra. The plane of movement of the maxilla in silu- riforms is mostly horizontal. As a consequence, when the bone is abducted (i.e., erect), its main axis points either laterally or laterally and slightly ventrally. In the two genera above, the maxilla is attached to the palatine in such a way that when it is abducted to a nearly 90° angle with the pal- atine, its main axis points strongly ventrally. The situation is most marked in Gagata cenia, where Suspensorium and Opercular Apparatus 13. Metapterygoid with long anterior process (CI = 50, RI = 50). Acrochordonichthys, Akysis, Breitensteinia. The metapterygoid in siluriforms other than di- plomystids is primitively an irregular, roughly round plate of bone. Although it may have well- marked angles along its borders, it usually lacks large processes. In the taxa above, the metapter- ygoid has an elongate process at its anterior comer. This process, when long enough, runs alongside the ventrolateral border of the posterior part of the palatine and parallels the ventral margin of the ectopterygoid. The process is very small but is still quite evident in Akysis macronemus. 14. Metapterygoid with well-defined thickening for ligamentous connection with endoptery- goid (CI = 100, RI = 100). Amblyceps, Liobagrus. The metapterygoid and endopterygoid in cat- fishes, when both elements are present, are con- nected by a sheet-like but usually poorly differ- entiated ligament. In amblycipitids, the ligament is narrow and very well defined, and its attachment site on the metapterygoid is delimited by an os- DE PINNA: ASIAN CATFISH FAMILIES 13 Fio. 9. Laguvia ribeiroi, ummz 208955, right opercle, lateral view. Scale bar = 1 mm. seous thickening that forms a small globular pro- cess on the mesial margin of the bone (Fig. 8). 15. Quadrate with anterior spur-like process (CI = 50, RI = 0). Parakysis, Amblyceps. The anterior portion of the quadrate in catfishes primitively has no expansions extending anteri- orly significantly beyond the vertical through its articulation with the lower jaw. In the taxa above, there is a large laminar extension of bone extend- ing anteriorly from the dorsal margin of the an- terior portion of the quadrate. This elongate lam- inar expansion extends alongside most of the lat- eral margin of the metapterygoid, and a conspic- uous ligament connects it either with the lateral ethmoid or with the posterior region of the pre- maxilla (the exact point of attachment is impos- sible to determine from the material available). 16. Endopterygoid surrounding metapterygoid dorsally (CI = 100). Pseudecheneis. The meta- and endopterygoid in siluriforms primitively do not encircle one another. In Pseud- echeneis, the endopterygoid is strongly curved (ventrally concave) to closely surround the dorsal margin of the metapterygoid. 17. Metapterygoid reduced to its center of ossi- fication (CI = 100). Parakysis. The metapterygoid of siluriforms is primitively composed of a small ossification center adjacent to its articulation with the quadrate, plus a more or less extensive area of laminar bone. The laminar region usually forms the majority of the area of the metapterygoid. In Parakysis, the laminar por- tion is extremely reduced, and the metapterygoid is effectively represented only by its small, cone- shaped ossification center. 18. Very large subpreopercle (CI = 100). Parakysis. The subpreopercle of siluriforms is typically a tubule of bone lining the branch of the laterosen- sory canal between the preopercle dorsally and the lower jaw ventrally. It is commonly found in cat- fishes as a poorly ossified and thin-walled struc- ture. In Parakysis, the subpreopercle is apomorph- ically modified into a large and thick-walled bone, still canal-like in morphology but far more heavily ossified than is usual in other siluriforms. 19. Opercle very narrow (CI = 100, RI = 100). Laguvia, Pseudolaguvia, [Aspredinidae = ?]. The overall morphology of the opercle in silu- riforms is that of a broad, roughly triangular bony lamina. Although there is wide variation in the shape of the bone, it is almost invariably a rela- tively wide plate as in other ostariophysans and teleosts in general. In the derived state, most of the laminar portion of the opercle is absent, and its remaining part is represented mostly by the regions corresponding to the dorsal and antero- ventral margins of the bone in the primitive con- dition. The opercle thus has the form of a long thin arc with the concave portion located ventro- posteriorly (Fig. 9). This apomorphic morphology 14 FIELDIANA: ZOOLOGY is the result of a reduction of the ossification of the central lamina of the opercle. This reduction, and the fact that some of the taxa showing this opercle modification (Laguvia) have a small body size, indicate that it may be a paedomorphic trait. However, another paedomorphic sisorid, Hara jerdoni, has a plesiomorphic condition for the opercle. The opercle in most aspredinids corre- sponds to the derived condition; however, at least one (Xyliphius) has a large opercular lamina, cast- ing doubt on the exact coding for this character in that taxon. 20. Hyomandibula with convex anterior expan- sion (CI = 100, RI = 100). Amblyceps, Liobagrus. The posterior margin of the hyomandibula in catfishes primitively forms a simple arc not ex- tending significantly posterior to the articular con- dyle for the opercle. In both genera of Amblycip- itidae, the hyomandibula has a conspicuous lam- inar expansion posteriorly, immediately ventral to its articulation with the neurocranium. This ex- pansion has a roughly round profile and extends posteriorly considerably beyond the articulation with the opercle. 21. Endopterygoid with spur-like process antero- laterally (CI = 100). Bagarius. The catfish ectopterygoid, when present, is quite irregular in shape and may have a number of an- gles around its margin, some of which can assume the aspect of poorly defined processes. The large ectopterygoid in Bagarius uniquely has a well-de- fined pointed process anterolaterally. This spur- like nonlaminar process extends dorsally along part of the posterior segment of the premaxilla and is ligamentously attached to it. 22. Endopterygoid bifurcated anteriorly (O = 100, RI = 100). Laguvia, Pseudolaguvia. The anterior end of the endopterygoid in La- guvia and Pseudolaguvia is divided into two broad laminar arms with pointed tips. This condition contrasts with the plesiomorphic morphology of the bone seen in other catfishes (cf. character 21). 23. Condyle for opercle on hyomandibula direct- ed ventrally (CI = 100). Aspredinidae. The opercular condyle of the hyomandibula is primitively directed posteriorly or slightly ventro- posteriorly in siluriforms. The derived condition shows the condyle directed strongly ventrally. A similar situation is also present in most members of Clariidae. However, the number of nodes sep- arating aspredinids and clariids in catfish phylog- eny and the fact that the primitive clariid Hora- glanis shows the plesiomorphic condition for this character (de Pinna, 1993) indicate that the two occurrences of the derived state are convergent. Neurocranium 24. Supratemporal fossa present (CI = 33, RI = 60). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Bagarius. Conta, Erethistes, Erethistoides, Gagata, Glyptothorax, Hara, Laguvia, Nangra, Si- sor, Aspredinidae. These taxa share a well-defined depression, usu- ally round or triangular in shape, between the su- praoccipital, pterotic, and supracleithrum bones (Fig. 1 0). The depression is often filled with what appears to be adipose tissue and does not serve as a site for muscle insertion. Mahajan (1966b, p. 367) referred to this structure as a "temporal groove" when noting its presence in Sisor rabdo- phorus. The degree of development of the fossa varies widely among the examined taxa that have it, reaching its largest size in the akysid Acrochor- donichthys rugosus. It tends to be less well marked in species with heavy skull ossification and or- namentation, such as Erethistes, Glyptothorax, and similar forms. In these cases the fossa can be partly attenuated by additional ossification but is usually still evident from an angled dorsolateral view. The fossa is reduced in adult specimens of Parakysis examined but is still present as a small space ad- jacent to the anterior margin of the supracleith- rum. It is more evident in smaller specimens of Parakysis anomalopteryx. The fossa is occasion- ally absent in some specimens of Laguvia shawi, but it is definitely present, albeit extremely small, in L. ribeiroi. The structure is also absent in the largest Bagarius examined (a representative of B. yarrelli), presumably as a result of additional os- DE PINNA: ASIAN CATFISH FAMILIES 15 soc PTE Fig. 10. Breitensteinia insignis, amnh 58378, neuro- cranium and associated structures, dorsal view. Dotted lines outline eyes. Scale bar = 1 mm. sification that develops in large individuals. Small- er specimens of the species have the fossa. The structure is clearly visible in aspredinids, except in Xyliphius, where it is largely obscured by heavy bony growth. 25. Paired fontanels among supraoccipital, fron- tals, and sphenotics (CI = 50, RI = 75). Acrochordonichthys, Breitensteinia, Gagata, Nan- gra, Sisor. Siluriforms primitively have two median fon- tanels in the neurocranium, similar to those in other otophysans. The two fontanels are usually separated by a bony strut called the epiphyseal bar. In the taxa above, the neurocranium possesses an additional paired fontanel, positioned between the frontal and supraoccipital (Fig. 1 0), sometimes also occupying a small area of the sphenotic. The paired fontanels are usually located at about the same transverse level as the posterior median fontanel and do not seem to serve as passage for nerves or blood vessels. They are actual fontanels, opening to the interior of the braincase, and not just de- pressions in bone, such as with the supratemporal fossa. The paired fontanels reach their greatest de- velopment in Nangra nangra, as a result of which the skull nearly lacks a roof. In Nangra viridescens, the paired fontanel is small and strongly displaced posteriorly, and its margins are sometimes partly coalescent with the supratemporal fossa. The paired fontanels are smallest in Breitensteinia and Ac- rochordonichthys. The actual distribution of the paired fontanels may be far broader than recorded here, and it may be obliterated by additional ossification in adults of many taxa. Although absent in male aspredinids examined, for example, small fontanels are present in juvenile specimens of Aspredo aspredo (usnm 226072). This condition may occur in several oth- er terminal taxa whose juveniles are currently un- known but that have heavily ossified skulls as adults. Until more information on juveniles of several other taxa becomes available, this char- acter is defined as the presence of paired fontanels in the adult stage only. The presence of these fon- tanels in any stage of development may represent a character for a considerably larger clade. 26. Supracleithrum strongly attached to skull (CI = 100, RI = 100). Acrochordonichthys. Akysis, Breitensteinia, Par- akysis, Bagarius, Conta, Erethistes, Erethistoides, Gagata, glyptostemoids, Glyptothorax, Hara, La- guvia, Nangra, Pseudecheneis, Pseudolaguvia, Si- sor, Aspredinidae. The complex structure forming the upper por- tion of the shoulder girdle in catfishes (formed by 16 FIELDIANA: ZOOLOGY the supracleithrum, posttemporal, and arguably the extrascapular and ossified Baudelot's ligament; see Lundberg [1975] and Fink & Fink [1981] for a discussion on this issue; for convenience, the whole structure is herein referred to as the supra- cleithrum) is primitively loosely connected with the posterolateral portion of the skull roof The articulation of the supracleithrum with the skull is movable to a certain extent, and although the bony surfaces are closely set, the link is by soft connective tissue only. In the derived condition, the supracleithrum is firmly attached or ankylosed to the neurocranium, and the two elements are immovably connected either by strong connective tissue (aspredinids) or direct bone contact (all oth- er terminal taxa above). 27. Posterior portion of supracleithrum anky- losed to margin of Weberian lamina (CI = 66, RI = 91). State 1 : Bagarius, Conta, Gagata, Glyptothor- ax, Nangra, Sisor, Aspredinidae. State 2: Erethistes, Erethistoides, Hara, La- guvia, Pseudolaguvia. The supracleithrum in catfishes has a complex three-dimensional structure that may play a va- riety of structural roles in the region of the back of the skull, Weberian complex, and shoulder gir- dle. It is difficult to make a generalization about the primitive condition of the supracleithrum as a whole for siluriforms because it includes a num- ber of possibly independent characters that can only be appropriately described and understood separately. In the case in question, the taxa enu- merated above share a derived condition unique among siluriforms, in which the supracleithrum has a large posterior process originating about the same horizontal level as the lateral margin of the Weberian lamina. This process joins the margin of the Weberian lamina and ankyloses with it, forming a firm bony strut between the back of the skull (to which the supracleithrum is also firmly attached or ankylosed in the above taxa) and the Weberian complex (Fig. 1 1 ). Among the taxa ex- amined, only Glyptothorax pectinopterus has a doubtful condition for this character because its posterior supracleithral process is poorly devel- oped, a condition assumed to be a result of the small size of the specimens examined for that spe- cies. This character is subdivided into two ordered states. In state 1 , the posterior supracleithral pro- cess extends to midway along the lateral margin of the Weberian lamina (Fig. 1 1 A). In state 2, more extreme, the process extends for at least two thirds of the total lateral margin of the lamina (Fig. 1 IB). 28. Pterotic with shelf-like expansion laterally (CI = 100). Aspredinidae. The lateral margin of the pterotic in catfishes is primitively devoid of conspicuous expansions be- yond the general profile of the neurocranium in dorsal view. In aspredinids, the pterotic has a large shelf-like expansion of laminar bone extending lat- erally well beyond the remainder of the profile of the skull. At its site of attachment, this shelf is about half the width of the skull. 29. Epioccipital with slender posterior process (CI = 50, RI = 50). Bagarius, Gagata, Nangra. The plesiomorphic condition of the epioccipital in siluriforms is a simple convex posterior surface without conspicuous expansions or processes. In these three taxa, the epioccipital has a well-de- fined, slender posterior process that is somewhat rounded or flattened in cross-section and oriented perpendicular to the main surface of the bone (Fig. 12). The process is pointed at its distal tip and seems to serve as an attachment for a ligament whose other end of attachment could not be pre- cisely located in the material available. It attains its largest development in Bagarius, where it ap- parently becomes more heavily ossified with growth. The process is smallest but still clearly present in Gagata cenia. In some specimens of Glyptothorax, there is a slight prominence at the posterior surface of the epioccipital that may rep- resent an incipient stage of the process. If this ob- servation is confirmed in more detailed morpho- logical studies, then the level of this character will have to be modified and perhaps be divided in two ordered derived states, state 1 applying to Glyptothorax and state 2 assigned to the assem- blage above. In the meantime, Glyptothorax is coded as having the primitive condition. A few other catfishes also have an epioccipital process, namely Neotropical doradoids (Chardon, 1968; Ferraris, 1988). In these taxa, however, the structure is much larger, laminar, and expanded at the tip. These characteristics distinguish it from the condition seen in the taxa above and indicates DE PINNA: ASIAN CATFISH FAMILIES 17 WL dj ez^ PIE soc WL ^-rj PTE B Fk>. 1 1 . Posterior portion of skull, Weberian lamina and lateral-line canal, lateral view. A, Conta conta, ummz 208632. B, Erethistes pusillus. ummz 208697. Scale bar = 1 mm. nonhomology between the two structures, a con- clusion also corroborated by the distant relative phylogenetic position of sisorids and doradoids within siluriforms (de Pinna, 1993). 30. Lateral ethmoid with narrow lateral exten- sions directed posteriorly alongside lateral margin of frontals (CI = 100, RI = 100). Bagarius. [Gagata = ?], glyptostemoids, Glypto- thorax, [Nangra = ?], Pseudecheneis, [Sisor = ?]. The dorsolateral portion of the lateral ethmoid in catiishes is frequently somewhat extended pos- teriorly underneath the lateral margin of the fron- tal. This extension is broad and not markedly dif- ferentiated from the body of the lateral ethmoid. Also, it normally does not extend along the level of the dorsal surface of the skull but instead is located ventral to the frontal. In the derived con- dition, the lateral ethmoids have long, narrow, posterior extensions running lateral to and at the same level as the frontals (Fig. 1 3). In many cases, the lateral ethmoid extensions form part or the whole of an antorbital process, and in glyptoster- noids the extension can be considerably wider pos- teriorly. Although this was not observed in any specimens examined, it seems that the distal por- tion of the lateral ethmoid extensions may become detached in Glyptosternum, as observed by Tilak (1963, p. 289, Figs. 41, 42) and Gauba (1969, p. 3, Figs. 1, 14). Both authors considered the de- tached posterior piece an independent bone of un- known homologies. In view of the condition seen in related taxa, the indef)endent structure probably originates by fragmentation from a portion of the lateral ethmoid. Some of the taxa above are coded as missing because of the extensive additional os- 18 HELDIANA: ZOOLCXjY Fig. 12. Bagarius yarrelli, lacm 38294-55, posterior part of temporal region, dorsolateral view. Arrow indicates pKjsterior process on epioccipital. Scale bar = 1 mm. siiication in the ethmoid region, which makes it difficult to observe the limit between certain regions of the frontal and lateral ethmoid. 31. Lateral processes of vomer long and thin, ex- tending underneath entire length of articular processes of lateral ethmoid (CI = 100). Glyptothorax. The posterolaterally directed processes of the vomer give the bone its characteristic arrow shape in catfishes. Normally the processes are relatively short, not extending beyond the width of the eth- moidal region of the neurocranium. In all exam- ined representatives of Glyptothorax, the vomer's processes are elongate and reach or nearly reach the tip of the articular process (which is itself quite elongated laterally, cf character 32) of the lateral ethmoid (Fig. 14). 32. Articular region of lateral ethmoid an elongate process, with articular facet for palatine at tip (CI = 100, RI = 100). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Amblyceps, Liobagrus, Bagarius, Conta, Erethistes, Erethistoides, Gagata, glyptostemoids, Glyptothorax, Hara, Laguvia, Nangra, Pseudeche- neis, Pseudolaguvia, Sisor, Aspredinidae. The articulation for the palatine on the lateral ethmoid in siluriforms shows a well-defined artic- ular facet provided with a cartilage lining contin- uous with the ethmoidal cartilage. Although the articular facet is well differentiated from the re- mainder of the lateral ethmoid, it does not extend markedly beyond the lateral margin of the bone. In the taxa above, which include all terminals treated in this analysis, in the articular region of the lateral ethmoid there is a long process extend- ing directly lateral from the bone (Figs. 10, 13, 14). The articular facet for the palatine is located at the tip of that process. The plug of cartilage lining that facet is typically completely independent from the median ethmoidal cartilage in most taxa. The only taxa in which they are continuous are Pseudeche- neis and perhaps Gagata (not clear in the material available). In Nangra, the two cartilages are almost continuous, separated only by a small gap. Based on the results of this analysis, the separation of cartilages is at the same phylogenetic level as the DE PINNA: ASIAN CATFISH FAMILIES 19 PMX NfE LAC Fig. 13. Glyptothorax major, amnh 58410, anterior part of skull and associated structures, dorsal view. Scale bar = 1 mm. elongation of the articular process, and conditions such as in Pseudecheneis are hypothesized to be reversals. Possibly, the elongation of the process and the separation of its distal cartilage from the main ethmoidal cartilage are two separate char- acters. However, until more information is avail- able on the development of this complex in siso- roids and other catfishes, they are here conser- vatively treated as a single trait. 33. Mesethmoid comua with dorsal subdivision (CI = 100, RI = 100). Acrochordonichthys. Breitensteinia. Primitively, the mesethmoid comua in siluri- forms are simple rod-like structures lacking note- worthy irregularities or deviations from their even surface. In these two genera, the mesethmoid cor- nua have a short subdivision stemming from their dorsal surface (Fig. 10). This dorsal "arm" is very PAR lAT ORB Fig. 14. Glyptothorax sinensis, amnh 10265, ante- rior part of neurocranium, ventral view. Larger stippling represents cartilage. Scale bar = 1 mm. short and in some cases appears just as a swelling of the distal portion of the dorsal surface of the comua. 34. Lateral ethmoid with well-defined dorsal facet contacting ventral surface of frontal (CI = 100, RI = 100). Gagata, Nangra. In the plesiomorphic condition, the contact be- tween the lateral ethmoid and the frontal in silu- riforms is made by simple flat surfaces on both bones. Uniquely, in Gagata and Nangra, the pos- terior region of the dorsal surface of the lateral ethmoid has a well-defined arc-shaped facet (Fig. 1 5) abutting the ventral surface of the frontal. This facet (herein called the dorsal facet of lateral eth- moid) is provided with a well-developed cartilage, clearly visible through the transparent frontal in counterstained specimens of thinly ossified species such as Nangra nangra and Gagata cenia. In spe- cies with heavily ossified skulls, the facet and its cartilage can be exposed by removing the anterior 20 FIELDL\NA: ZOOLCXJY ME LAT PAL Fig. 1 S. Gagata gagata, amnh 8358, ethmoid region, frontals removed, dorsal view. Arrow indicates dorsal facet on lateral ethmoid. Scale bar = 1 mm. portion of the frontals. The cartilage of the facet is completely independent from the more ventrally located ethmoid cartilage. The frontal facet of the lateral ethmoid is an extremely peculiar structure, without parallel in any other siluriform and probably unique among ostariophysans. Its functional significance is as yet unsolved; there seem to be no associated modifi- cations in the frontal and no special bone move- ment in the region. 36. Sphenotic not contacting supraoccipital (CI = 100). Pseudecheneis. In most siluriforms and other ostariophysans, the sphenotic contacts posteromesially the antero- lateral margin of the supraoccipital. Uniquely in Pseudecheneis, the two bones are separated by the frontal, which has a large posterior extension in that genus (Gauba, 1968, p. 233). 35. Lateral ethmoid cartilage for palatine facet with extension for lacrimal (CI = 100, RI = 100). Gagata, Nangra. The cartilage on the lateral ethmoid that con- nects it with the palatine is primitively restricted to this articulation. In the two genera above, this cartilage has a lateral extension of its dorsal margin that abuts the lateral surface of the lacrimal. Branchial Arches 37. Last hypobranchial with anterolateral exten- sion (CI = 100, RI = 100). Conta, Erethistes, Erethistoides, Hara, Laguvia, Pseudolaguvia. The anterolateral comer of the last (third) hy- pobranchial in catfishes is either similar in form DE PINNA: ASIAN CATFISH FAMILIES 21 Fio. 1 6. Erethistoides montana, ummz 208745, basal portion of gill arches, dorsal view. Larger stippling rep- resents cartilage. Scale bar = 1 mm. Fig. 17. Acrochordonichthys ichnosoma, mzusp 42468, ventral portion of gill arches, dorsal view, gill rakers not represented. Larger stippling represents car- tilage. Scale bar = 1 mm. to the other comers of the structure or protrudes only slightly anteroiaterally. In the taxa above, the third hypobranchial is strongly defined anteroiat- erally, and its anterolateral comer is modified into a large process (Fig. 1 6). In some cases the exten- sion appears as an almost or distinctly separate cartilaginous element, as in Erethistoides. 38. Second hypobranchial unossified (CI = 50, RI = 88). Acrochordonichthys, Breitensteinia, Parakysis, Conta, Erethistes, Erethistoides, Hara, Laguvia, Pseudolaguvia, Aspredinidae. Primitively in catfishes, the first and second hy- pobranchials are ossified and the third is cartilag- inous. The degree of ossification varies, and there is always a remnant cartilaginous rim posteriorly or at the tips of the ossified hypobranchials. In some instances, such as in some trichomycterids, the third hypobranchial is also ossified, a condi- tion that approaches the state seen in nonsiluri- form outgroups but that is derived within the or- der. In the taxa above, there is an apomorphic modification in the opposite direction, with only the first hypobranchial being ossified (Figs. 16, 17). Both the second and third hypobranchial are en- tirely cartilaginous, even in the adults examined. In some specimens of Laguvia ribeiroi, there is a small ossified area on the second hypobranchial, which is interpreted here as an abnormality. 39. First hypobranchial unossified (CI = 100.) Parakysis. The first hypobranchial in catfishes is primi- tively ossified (see character 38). In Parakysis, the structure (as well as the other hypobranchials) re- mains cartilaginous in adults. Hypobranchials also fail to ossify in hoplomyzontin aspredinids. Al- though a phylogeny of aspredinids is not yet avail- able, the condition in hoplomyzontins is tenta- tively interpreted as a result of paedomorphosis found exclusively in that tribe and thus is not rep- resentative of the general condition in aspredinids. All other members of the family retain the ossified condition for the first hypobranchial. 40. All basibranchials unossified (CI = 100, RI = 100). Parakysis, [glyptostemoids = ?]. There is a large variation in the size and extent of basibranchial ossifications in siluriforms. Near- ly always, however, there is some degree of ossi- 22 HELDIANA: ZOOLOGY fication of the first and second basibranchial ele- ments, a condition that can be considered ple- siomorphic for the order. The third basibranchial is invariably cartilaginous in catfishes. In Para- kysis, all basibranchial elements are present, but none are ossified. A similar condition is seen else- where only in Oreoglanis (in no other glyptoster- noids). Therefore, glyptostemoids are coded as missing for this character. The conditions in Para- kysis and Oreoglanis are quite different structur- ally. In Parakysis the cartilaginous basibranchials are of normal shape, whereas in Oreoglanis they are extremely narrow and elongate, nearly thread- like. Within aspredinids, basibranchial ossifica- tions are very reduced or absent in hoplomyzon- tins. These reductions are postulated to be the result of the paedomorphic nature of that tribe and probably are not representative of the general con- dition in aspredinids. This assumption of non- homology is supported by the overall phylogeny. 41. Additional cartilaginous element anterior to last basibranchial (CI = 20, RI = 50). Acrochordonichthys, Breitensteinia, Parakysis, Bagarius, Glyptothorax, Laguvia, Pseudecheneis, Psendolaguvia, Sisor. The last basibranchial in siluriforms (the third from the anterior, homologous to the fourth in other ostariophysans) is a single cartilaginous structure that is usually much wider than the first and second basibranchials and is normally located immediately posterior to the latter. In the taxa above, there is a distinct separate cartilaginous element between the second and third basibran- chials (Figs. 17, 18). This structure is most de- veloped in Acrochordonichthys rugosus and Brei- tensteinia insignis, where it has the shape of a nar- row rectangle closely adpressed to the anterior margin of the third basibranchial. This supernu- merary cartilage probably is formed as an anterior segmentation of the third basibranchial, insofar as it is usually more closely connected with the third rather than second basibranchial. Also, its width normally matches that of the broad third basi- branchial instead of the much narrower second. In some cases, the extra cartilage is present in an extremely reduced condition, represented by a small globule of cartilage somewhat loosely situ- ated between the second and third basibranchials. Such is the case in Sisor and Laguvia. The cartilage is present in most Glyptothorax except G. pectin- opterus. Until more details are available on the development and structure of this cartilage, its Cj+TP 4f I pm Fig. 1 8 . Breitensteinia insignis, amnh 58378, ventral portion of gill arches, dorsal view, gill rakers not rep- resented. Larger stippling represents cartilage. Scale bar = 1 mm. variation in shape and size cannot be utilized as primary evidence for relationships. Its presence, however, can justifiably be hypothesized as de- rived. 42. First epibranchial expanded anteriorly (CI = 100). Pseudecheneis. The first epibranchial in siluriforms is primi- tively round or flat in cross-section and not mark- edly wider in dorsal view than the more posterior second and third epibranchials. Uniquely, in Pseu- decheneis, the first epibranchial is broad, roughly three times as wide as the second epibranchial. 43. Second hypobranchials with longest axis near- ly perpendicular to first (CI = 100, RI = 100). Amblyceps, Liobagrus. The margins of first and second hypobranchials in catfishes are usually parallel or slightly divergent DE PINNA: ASIAN CATFISH FAMILIES 23 Fig. 1 9. Amblyceps mangois, ansp 593 1 6, basal por- tion of gill arches, dorsal view. Larger stippling repre- sents cartilage. Scale bar = 1 mm. mesiad. In amblycipitids, the two bones (i.e., the posterior margin of first and anterior margin of second hypobranchials) diverge strongly toward the midline, forming a nearly 90° angle mesially (Fig. 1 9). This peculiar relative position seems to be the result of a shift in the orientation of the second rather than the first hypobranchial. 44. Last basibranchial narrow (CI = 50, RI = 0). Gagata, Sisor. The last basibranchial in catfishes is always unossified and primitively is considerably wider and rounder in dorsal view than are the two more anterior basibranchials. In Sisor and Gagata, the last basibranchial is apomorphically modified into a long and thin cartilage, as narrow as or more so than the more anterior basibranchials (Fig. 20). The narrowing does not affect the vertical com- ponent of the cartilage, which remains as deep (in lateral view) as in the plesiomorphic condition. The condition is most extreme in Sisor. 45. Anterior cartilage of anteriormost basibran- chial enlarged (CI = 50, RI = 50). Acrochordonichthys, Akysis, Breitensteinia. The first basibranchial of other ostariophysans is absent in siluriforms. Their anteriormost basi- FiG. 20. Sisor rhabdophorus, bmnh 1 970.6.25. 1 , bas- al portion of gill arches, dorsal view. Larger stippling represents cartilage. Scale bar = 1 mm. branchial (corresponding to the second in other otophysans) has an anterior rim of cartilage that articulates with the mesial ends of the first pair of hypobranchials. In the taxa above, the anterior cartilage of the second basibranchial is elongated anteriorly, reaching beyond the anterior margin of the hyoid bar (Figs. 1 7, 1 8). The enlarged cartilage sometimes shows some irregular areas of ossifi- cation, such as in Breitensteinia (Fig. 18). It is unclear if this cartilage represents an atavistic ex- pression of the first basibranchial of nonsiluriform ostariophysans. 45. Fourth ceratobranchial with tooth plates (CI = 100). Bagarius. In siluriforms, only the fifth ceratobranchial typ- ically bears tooth plates. Almost uniquely within the order, Bagarius possesses narrow tooth plates attached also along the dorsal border of its fourth ceratobranchial. These tooth plates are absent in 24 FIELDIANA: ZOOLOGY the fourth ceratobranchials of the largest specimen examined of the genus (a member of B. yarrelli) but are present in smaller specimens of the same species. The tooth plates may have been lost in that specimen (which is a dry skeleton and there- fore more prone to tooth plate detachment), or alternatively they may disappear with growth. Gauba (1962) did not describe or illustrate tooth plates on the fourth ceratobranchial of Bagarius bagarius. It is unclear whether their absence there- in was due to large size because the size of spec- imens examined is not indicated in the paper. 47. Lateral cartilage of first hypobranchial en- larged (CI = 100). Aspredinidae. The lateral cartilage of the first hypobranchial in siluriforms is primitively far shorter than the bony portion of the bone. In the derived condition, the cartilage is at least 50% of the length of the bony portion of the first hypobranchial. This char- acter could not be observed in the specimens of hoplomyzontins examined because their first hy- pobranchials are unossified and reduced or absent. 48. Last basibranchial bifurcated anteriorly (CI = 100). Pseudolaguvia. The totally cartilaginous last basibranchial in catfishes is undivided anteriorly in the plesiomor- phic state. In Pseudolaguvia, the basibranchial is anteriorly split into two short arms, with their tips slightly expanded and curved laterally. 49. Proximal portion of first and/or second epi- branchial curved posteriorly (CI = 50, RI = 66). Glyptostemoids, Glyptothorax, [Laguvia = ?], Pseudecheneis, Pseudolaguvia. The first and second epibranchials in siluriforms are roughly straight in the plesiomorphic condi- tion. In the taxa above, the proximal portions of the first or second epibranchials (or more fre- quently both) are strongly curved posteriorly. The change in direction may be a distinctly angled twist, as in Glyptothorax, or a more gentle curve, as in Pseudecheneis and glyptostemoids. The condition in Laguvia is weakly developed and in some spec- imens is similar to the plesiomorphic condition. 50. Proximal portions of epibranchials 1 and 2 adpressed to one another (CI = 100). Bagarius. The first and second epibranchials in catfishes are primitively not particularly close to each other, being connected by connective tissue in a way comparable to that between other epibranchials. In Bagarius, the first and second epibranchials are closely adpressed along their proximal margins. The two bones are furthermore anteroposteriorly flattened at the area of connection, increasing the surface of contact. Hyoid Bar 51. Branchiostegal rays forming a closed round arrangement (CI = 100, RI = 100). Amblyceps, Liobagrus. The angular difference between the bases of the first and last branchiostegal rays in catfishes is primitively about 90°. In the taxa above, the bases of the first and last branchiostegal rays are oriented approximately at a 135° angle to one another. This orientation results in a peculiar arrangement in which the set of branchiostegal rays in amblycip- itids surrounds a major portion of the posterior margin of the hyoid bar. 52. Lateral wings of urohyal extremely reduced or absent (CI = 100, RI = 100). Glyptostemoids, Glyptothorax, Pseudecheneis. The urohyal in siluriforms primitively has bi- lateral laminar expansions in the form of lateral "wings" extending from the margins of a relatively small central body (Arratia & Schultze, 1 990). In the taxa above, the urohyal is composed mostly of the narrow median portion, with the wing-like extensions small or totally absent (Fig. 21). 53. Proximal end of anterior ceratohyal with nar- row, anterolaterally directed extension (CI = 100, RI = 100). Acrochordonichthys, Breitensteinia. The mesial end of the anterior ceratohyal in catfishes is primitively a simple stmcture, roughly rounded in cross-section at its articulation with the hypohyal and without major expansions or DE PINNA: ASIAN CATHSH FAMILIES 25 Fig. 21. Urohyals in ventral view. A, Glyptothorax sinensis, amnh 10265. B, Exostoma sp., usnm 130183. C, Pseudecheneis sukatus, bmnh 1985.9. 16:50. Scale bar = 1 mm. processes. In the taxa above, the anterior margin of the proximal end of the anterior ceratohyal is developed into a narrow, anterolaterally oriented process (Fig. 22). The process does not involve any portion of the articular surface for the hypo- hyal and is devoid of cartilage. The hypohyal in some species has an equivalent process laterally, which is ankylosed to the anterior process of the ceratohyal. A similar structure also occurs in a single glyptostemoid examined, Pseudexostoma. In that genus, however, the condition differs in that the process includes part of the articular sur- face for the hypohyal and is accordingly lined with cartilage along its anterior margin. This morpho- logical difference, plus the fact that all other glyp- tostemoids lack the derived condition, indicates that the condition in Pseudexostoma is not ho- mologous to that of Breitensteinia and Acrochor- donichthys. 54. Posterior margin of anterior ceratohyal with laminar expansion forming laterally directed process (CI = 100, RI = 100). Conta, Erethistes, Erethistoides. Hara, Laguvia, Pseudolaguvia. The plesiomorphic condition for the anterior ceratohyal in catfishes is to have the form of a simple bar somewhat expanded vertically in its anterior and posterior portions and roughly round or triangular in cross-section. In the taxa above, the anterior ceratohyal has a posterior laminar shelf with irregular margins (Fig. 23). The extent of the shelf varies, and in extreme cases it extends along the entire posterior margin of the bone. Laterally, the shelf has a large concavity that delimits a broad, laterally directed process. The structure can be markedly reduced in some species of the genus Laguvia but is well developed in the closely related Pseudolaguvia. 55. Posterior ceratohyal with dorsal process near its connection with suspensorium (CI = 1 00, RI = 100). Acrochordonichthys, Breitensteinia, Parakysis. The proximal tip of the posterior ceratohyal in siluriforms primitively has an irregular surface, with small depressions at the sites of ligament at- tachment. The region next to the dorsal tip of the bone lacks any well-differentiated processes. In the derived condition, there is a prominent process on the dorsal margin of the posterior ceratohyal im- mediately dorsal to its distal tip (Fig. 22). This process, which is apparently not an attachment site for the ligamentous connections between the hyoid bar and the mesial surface of the suspen- sorium, is most developed in Acrochordonichthys and Breitensteinia. 26 FIELDIANA: ZOOLOGY HYP Flo. 22. Breitensteinia insignis, amnh 58378, hyoid bar, dorsal view. Arrow 1 indicates dorsal process on posterior ceratohyal; arrow 2 indicates anterior expansion of anterior ceratohyal. Larger stippling represents cartilage. Scale bar = 1 mm. Weberian Apparatus 56. Weberian capsule complete, without apparent lines of suture (CI = 100, RI = 100). Gagata, Nangra, Sisor. The Weberian complex in siluriforms has a number of different forms of swim bladder encap- sulation. They all usually involve some degree of wrapping of the swim bladder by the parapophyses of the complex centrum (Alexander, 1965). In sev- eral taxa, there is also a contribution to the ventral wall of the capsule by neomorph offshoots of the ventral surface of the complex centrum. In any case, the walls of the bony capsule contain gaps or lines of suture between its different components. In the taxa above, in contrast, the capsule is con- tinuous and surrounds the whole swim bladder (except its lateral surface) without spaces or trace of sutures (Fig. 24). Elsewhere in siluriforms, a similar capsule morphology is found in Neotrop- ical loricarioids and in Leptoglanis and Zaireich- thys of the African Amphiliidae. In view of the number of nodes separating loricarioids from si- sorids with encapsulated swim bladders in the hy- pothesized phylogeny of siluriforms (de Pinna, 1993), the conditions in loricarioids and in some sisorids are assumed to be independent. A similar HYP Fig. 23. Hara hara, ummz 208748, hyoid bar, dorsal view. Larger stippling represents cartilage. Scale bar = 1 mm. DE PINNA: ASIAN CATFISH FAMILIES 27 FiQ. 24. Gagata gagata, amnh 8358, SEM photograph, ventrolateral view of swim bladder capsule. reasoning applies to Leptoglanis and Zaireichthys, which are demonstrably closer to other amphiliids than to any sisorids (de Pinna, 1993). 57. Parapophysis of fifth vertebra strongly flat- tened and expanded (CI = 50, RI = 87). Bagarius, Conta, Erethistes, Erethistoides, Gaga- ta, Glyptothorax, Hara, Laguvia, Nangra, Pseu- dolaguvia, Sisor, Aspredinidae. The parapophysis of the fifth vertebra in cat- fishes is primitively a narrow structure, often gent- ly curved anteriorly, that tapers toward a pointed tip. In the taxa above, the parapophysis is modi- fied into a stout shelf-like structure, often mark- edly expanded horizontally toward its distal end (Fig. 25, 26). In Gagata, the parapophysis is partly incorporated into the swim bladder capsule and is somewhat reduced. However, it still clearly cor- responds to the derived condition. 58. Parapophysis of fifth vertebra branched or notched distally (CI = 100, RI = 100). Erethistes, Erethistoides, Hara, Laguvia. Pseudo- laguvia. [Aspredinidae = ?]. In the taxa above, the fifth parapophysis, besides being expanded laterally, is notched one or more times along its distal margin (Fig. 26B). In the more extreme forms, the notching is deep and multiple, forming a coarse branching of the flat distal portion of the parapophysis. Within aspre- dinids, the process is notched in Bunocephalini and some members of Hoplomyzontini, but not in Aspredininae. Because the family is treated as a single terminal for this analysis, this character is coded as missing for this family. 59. Complex centrum as short as or shorter than free centra (CI = 100, RI = 100). Glyptostemoids, Pseudecheneis. The complex centrum of siluriforms, composed of fused primitive centra 2—4, is considerably shorter than the summed lengths of three free cen- tra, indicating that considerable shortening has oc- curred along with the fusion process (Alexander, 1965). The shortened complex centrum, however, is primitively still longer than a single free cen- trum. In the derived condition observed for Pseu- decheneis and glyptostemoids, the complex cen- trum is shortened to an extreme degree, being as short as or shorter than the individual free centra immediately posterior to it. 28 FIELDIANA: ZOOLOGY Fig. 25. Dysichthys coracoideus, amnh 21815, dorsal (partial) view of Weberian complex, supracleithrum, and part of lateral line; anterior on top. Scale bar = 5 mm. VC, TRI Fig. 26. Ventral view of Weberian complex. A, Glyp- tothorax major, amnh 58410. B, Laguvia ribeiroi, ummz 208955. Scale bar = 1 mm. 60. Distal tip of os suspensorium closely linked and/or adpressed to ventral process of com- plex centrum (CI = 100, RI = 100). Bagarius, Gagata, glyptostemoids, Glyptothorax, Nangra, Pseudecheneis, Sisor. The distal tip of the os suspensorium in siluri- forms is primitively free from any direct contact or strong ligamentous connections with a bony surface other than for the transformator process of the tripus. In the taxa above, the os suspenso- rium is firmly linked to the independent process that underlies part of the ventral wall of the swim bladder (Fig. 26A). In some taxa, the two struc- tures almost form a single shelf, although actual fusion is not observed except in taxa with a com- plete swim bladder capsule, namely Nangra, Ga- gata, and Sisor. Here the ventral process of the compound centrum is indistinguishably fused to the Weberian lamina, forming a complete swim bladder capsule. The os suspensorium in these cases seems to be modified or incorporated into a thick strut of bone fused dorsally to the internal wall of the capsule (Fig. 27). This condition is assumed to be a modification of the connection observed in the remaining taxa listed above. 6 1 . Parapophysis of fifth vertebra strong and at- tached to ventral side of centrum, directed directly perpendicular to vertebral axis (CI = 100). All terminal taxa. The parapophysis of the fifth vertebra in cat- fishes is primitively a narrow laminar structure attached to the dorsal half of its centrum and usu- ally gently curved anteriorly. In the derived con- dition, the base of the parapophysis is located on the ventral half of the centrum, is not laminar, and is directed straight laterally. There is a significant amount of phylogenetically informative variation in the morphology of this parapophysis, and its size ranges from very short (as in amblycipitids and akysids) to very long so as to reach the body wall laterally (as in aspredinidsand many sisorids). These similarities indicate, however, that some of the modifications are homologous throughout all terminal taxa included in this analysis. This par- DE PINNA: ASIAN CATFISH FAMILIES 29 Fio. 27. Gagata gagata, amnh 8358, SEM photograph, tripus, lateral view inside capsule shown in Figure 24. apophysis is sometimes referred to as a transverse process (e.g., Chardon, 1968). However, trans- verse process has been used in this and other works to refer to a different vertebral structure (cf. char- acter 70). The structiu-e here called the parapo- physis of the fifth vertebra appears to be serially homologous to a vertebral parapophysis. In sup- port of that conclusion, half-grown specimens of Liobagrus anguillicaudatus (amnh 11069) have small pleural ribs associated with the modified parapophyses, in the position expected for normal vertebral parapophyses. 62. Parapophysis of fifth vertebra long, almost or quite reaching lateral surface of body wall (CI = 100, RI = 100). Bagarius, Conta, Erethistes. Erethistoides, Gaga- ta, glyptostemoids, Glyptothorax, Hara, Laguvia, Nangra, Pseudecheneis, Pseudolaguvia, Sisor, As- predinidae. The parapophysis of the fifth vertebra in all ter- minals of this analysis is apomorphically modified from the condition observed in other siluriforms, as described for character 6 1 . In all terminal taxa except akysids and amblycipitids, the parapo- physis undergoes an additional modification in its length and extends laterally well beyond the limit of the parapophyses of the complex centrum (Figs. 25, 26). In most taxa, it extends laterally all the way to the body wall. In glyptostemoids, the com- plex centrum and associated structures are highly modified. Here, the parapophysis of the fifth cen- trum is markedly smaller and narrower than in other members of this clade and seems to be an extreme condition of the modifications seen in Pseudecheneis, the closest relative. 63. Posterodorsal comer of Weberian lamina with pointed process directed posteriorly (CI = 100, RI = 100). Bagarius, Gagata, Nangra, Sisor. The posterolateral comer of the Weberian lam- ina (modified parapophysis of complex centmm) in siluriforms is primitively a simple angle without conspicuous processes or expansions. Sometimes there is a small thickening of bone that does not form a well-differentiated structure. In the taxa above, there is a thick and well-defined process on the posterolateral comer of the lamina, directed posteriorly and frequently ornamented on its lat- eral surface (Fig. 24). In all the taxa above except Bagarius, the lamina is modified to form the pos- teroventral portion of a swim bladder capsule, but the process remains well differentiated and departs markedly from the lateral margin of the capsule. In Bagarius, the process is well developed in small to medium-sized individuals but seems to get less defined with growth. 30 FIELDIANA: ZOOLOGY Fio. 28. Akysis sp. 1 , ummz 2 1 4907, anterior portion of vertebral column, right humerovertebral ligament and tip of right humeral process, ventral view. Wavy lines represent ligament. Scale bar = 1 mm. 64. Humeral region connected with anterior por- tion of vertebral column by well-defined lig- ament (humerovertebral ligament) (CI = 60, RI = 60). State 1: Amblyceps, Liobagrus. State 2: Acrochordonichthys, Akysis, Breiten- steinia. State 3: Bagarius, Conta, Erethistes, Erethis- toides, Gagata, Glyptothorax, Hara, La- guvia, Nangra, Pseudecheneis, Pseudola- guvia, Sisor, Aspredinidae. [Parakysis = ?]. The humeral process in catfishes is a stout, pointed process directed posteriorly on the upper limb of the cleithrum. It primitively lacks well- defined ligamentous connections with surround- ing skeletal elements. In the derived condition dis- played by the taxa above, the tip of the humeral process (or the soft tissue in the vicinity of the humeral region) is linked to the vertebral column by a conspicuous ligament (Figs. 28, 29). This lig- ament, here named the humerovertebral ligament, is most evident in aspredinids. There are three well-defined conditions of the attachment of the ligament on the vertebral column, distinguished by a number of traits. In the first (state 1), the ligament attaches posteriorly on either the para- pophysis or the rib of the sixth vertebra and an- teriorly ends in soft tissue (seemingly associated with the wall of the axillary gland, but this asso- ciation could not be verified with certainty in the preparations available) without actually reaching the humeral process or any other bone. In the second condition (state 2), the ligament is ante- FiG. 29. Amaralia hypsiura, amnh uncat., part of Weberian complex, right humerovertebral ligament, su- pracleithrum, and part of pectoral girdle, dorsal view. Wavy lines represent ligament. Scale bar = 1 mm. riorly firmly attached to the mesial surface of the tip of the humeral process. Posteriorly it inserts on the lateral side of the sixth vertebral parapo- physis and proximal portion of its corresponding rib (Fig. 28). In a few taxa (e.g., Akysis sp. 2), part of the ligament also attaches on the fifth parapo- physis. In the third condition (state 3), the liga- ment attaches anteriorly on the humeral process, as in state 2, but its posterior attachment is exclu- sively on the ventral surface of the distal portion of the highly modified parapophysis of the fifth vertebra (Fig. 29). The attachment site on the par- apophysis is usually delimited by some bony in- dentation on the anterior margin of the parapo- physis. The humerovertebral ligament varies widely in length, according to the distance between the tip of the humeral process and the vertebral attach- ment. At one extreme is the situation in Bagarius, where the two attachment sites are nearly juxta- posed and the ligament is consequently very short. The other extreme is exhibited in Akysis, where the ligament can be slightly longer than the longest pleural rib. A humerovertebral ligament was not directly observed in the specimen of Sisor available be- cause its preparation did not preserve ligaments in good condition. However, the anterior margin of its fifth vertebral parapophysis has a slight notch corresponding to the attachment site of the liga- ment in other taxa. On the basis of this, Sisor is DE PINNA: ASIAN CATFISH FAMILIES 31 tentatively assigned state 3 for this character. This assumption should be verified in additional spec- imens of the genus when they become available. Pseudecheneis, which lacks a humeral process, has a humerovertebral ligament attaching anteri- orly onto a small cleithral splint immediately ven- tral to the articulation between the cleithrum and the skull. This splint is probably a remnant of the humeral process, which apparently in this taxon serves only as an attachment for the ligament. All other sisorid and akysid taxa that lack the hu- merovertebral ligament also lack the humeral pro- cess (e.g., glyptostemoids, Parakysis). This char- acter is coded as missing for Parakysis because of the presence of a conspicuous vertebral ligament that cannot be unambiguously homologized with the humerovertebral ligament. This ligament unites the first pleural rib with the neural prezygapo- physis of its own supporting vertebra, and it may arguably be a highly modified humerovertebral ligament. Resolution of this question will have to await examination of additional material and an ontogenetic series of Parakysis, currently unavail- able. It was not possible to determine the ordering of the three states defined above because there is no evident sequence from a less extreme to a more extreme condition. Also, no ontogenetic infor- mation relevant to the question is available. Therefore, the three states of the humerovertebral ligament are considered unordered (nonadditive) for the purpose of parsimony analysis. Thus, the transitions of the character are ambiguous on the most parsimonious topology, and in one possible optimization the three states are considered in- dependently derived. This alternative is unreal- istic because the presence of the ligament itself, being unique among catfishes, is obviously evi- dence for the monophyly of a large assemblage composed of all taxa above. Therefore, although the ordering of this character is presently un- known, it constitutes valid evidence for the mono- phyly of the whole group under study (see diag- noses for clades below), with two reversals to total absence (in Parakysis and glyptostemoids). The very conspicuous humerovertebral liga- ment in Aspredo was first reported by Bridge and Haddon (1893, p. 185, Fig. 73). These authors, however, did not note its presence in their speci- mens of Akysidae and Sisoridae. Tilak (1963, p. 298) observed the presence of the humerovertebral ligament in the sisorids Nangra, Glyptothorax, and Glyptosternum but stated that it was absent in Ga- gata and Bagarius. This "absence" is likely an artifact of preservation, because a well-developed ligament is present in all representatives ofGagata and Bagarius examined. Chardon (1968) also de- scribed the humerovertebral ligament in sisorids (p. 127)andaspredinids(pp. 171, 173) and clearly illustrated the structure in representatives of As- predinidae (cf. Chardon, 1968, Figs. 157-160). Chardon, however, did not comment on the pos- sible significance of the sharing of this trait by the two taxa. The functional significance of the humerover- tebral ligament is rather obscure. It seems to re- strict lateral movement of the anterior vertebrae relative to the pectoral girdle because the liga- ments are disposed bilaterally as antagonists rel- ative to the horizontal plane of the vertebral col- umn. Movement in that plane, however, is already severely limited in catfishes in general by the na- ture of their pectoral suspension and the fusion involved in the formation of the complex centrum. Also, in several members of the ingroup, move- ment of the skull relative to the complex centrum is impossible because of the firm bony bridge formed by the supracleithrum between the back of the skull and the lateral margin of the Weberian lamina (cf. character 27). All these modifications seem to be part of a general trend toward reduction of movement of the anterior part of the body, including the Weberian complex, relative to the free portion of the vertebral column. The hume- rovertebral ligament may play an important role in the functioning of the stridulatory apparatus (cf character 109 and discussion below on the evo- lution of the sound-producing apparatus). 65. Superficial ossification on ventral surface of complex centrum combined with part or all of OS suspensorium, forming elongated an- teroventral flange, anteriorly reaching basi- occipital (CI = 100, RI = 100). Erethistes, Hara. In the plesiomorphic condition for siluriforms, the OS suspensorium is a relatively narrow, flat, bony arc that extends anteroventrally in the space circumvented by the curvature of the transfor- mator process of the tripus. Further, the os sus- pensorium has a mostly even width from base to tip, and its anteriormost point is free and distant from the basioccipital. In the derived condition, the distal portion of the os suspensorium fuses with the ventral superficial ossification of the com- plex centrum, and the resulting structure is ex- panded into a leaf-like flange distally adjoined but 32 FIELDIANA: ZOOLOGY not fused to the ventrolateral surface of the com- plex centrum. The anterior tip of the shelf reaches the basioccipital anteriorly. The degree to which the OS suspensorium and the superficial ossifica- tion contribute to the final compound structure is unclear, although the superficial ossification seems to form a larger area of the leaf-like expansion. In some cases, a small protuberance of bone remains at the primitive site of the distal tip of the os suspensorium. Ossa suspensoria show compara- tively modest morphological variation in catfish- es, and the form seen in these taxa is one of the most remarkable modifications of this structure observed in the order. 66. Posterior part of Weberian lamina extensively contacting parapophysis of fifth vertebra (CI = 100, RI = 100). Bagarius, Conta, Erethistes, Erethistoides, Gaga- ta, glyptostemoids, Glyptothorax, Hara, Laguvia, Nangra, Pseudecheneis, Pseudolaguvia, Sisor, As- predinidae. The posterior region of the Weberian lamina in siluriforms is primitively independent of the par- apophysis of the fifth vertebra. It may approach the parapophysis and even contact its margin for a very short area near the vertebral centrum, but this contact is never extensive or firm. In the de- rived condition observed in all terminal taxa ex- cept amblycipitids and akysids, the Weberian lam- ina overlaps and broadly contacts the dorsal sur- face of the enlarged fifth parapophysis, and the two structures are either strongly connected or an- kylosed. 67. Posterior portion of Weberian lamina with lateral extension along anterior margin of par- apophysis of fifth vertebra (CI = 100, RI = 100). Bagarius, Gagata, glyptostemoids, Glyptothorax, Nangra, Pseudecheneis, Sisor. As described for character 66, the posterior part of the Weberian lamina is tightly and extensively connected to the dorsal surface of the parapophysis of the fifth vertebra, an apomorphic condition shared by all ingroup taxa except amblycipitids and akysids. In the subset listed above, there is a further modification of that derived condition, in which the posterior part of the Weberian lamina is laterally expanded to cover part or all of the anterior margin of the parapophysis of the fifth vertebra. In Gagata, the basal portion of the fifth vertebral parapophysis is completely incorporated into the ventral wall of the swim bladder, with only its distal portion still independent (see char- acter 69). A long suture is visible in this free por- tion, indicating the limit between the parapophysis itself and the apomorphic elongate lateral process of the Weberian lamina in other taxa. 68. Parapophysis of fifth vertebra with vertical lamina (CI = 100, RI = 100). Erethistes, Hara. The parapophysis of the fifth vertebra, which is enlarged in several members of the ingroup, is primitively a dorsoventrally flat horizontal shelf In Hara and Erethistes, the parapophysis has a well-developed vertical wall dorsally continuous with and ankylosed to the posterior margin of the Weberian lamina. The condition is somewhat re- duced, but still present, in the miniature Hara jerdoni. 69. Parapophysis of fifth vertebra incorporated into ventral wall of swim bladder capsule (CI = 100). Gagata. The parapophysis of the fifth vertebra is tightly united with the Weberian lamina in numerous members of the ingroup. The anterior margin of the parapophysis is adpressed to the posterior wall of the swim bladder capsule when this capsule is present, such as in Nangra and Sisor. In these cases, however, the parapophysis remains well dif- ferentiated from the capsule itself and is located posterior to it. Uniquely in Gagata, the parapo- physis of the fifth vertebra in incorporated into the capsule to an extreme degree. Its proximal portion is indistinguishably fused to the ventral wall of the capsule. Vertebrae and Ribs (Other Than Weberian Apparatus) 70. Dorsal expansion on first pleural rib (CI = 100). Pseudecheneis. The pleural ribs in siluriforms are primitively a simple splint of bone, round or flat in cross-sec- tion. In Pseudecheneis, the flat first pleural rib has a large laminar expansion dorsally. The expansion DE PINNA: ASIAN CATFISH FAMILIES 33 TPR Fig. 30. Breitensteinia insignis. amnh 58378, part of vertebral column comprising free vertebrae 10-17, ven- tral view. Scale bar = 1 mm. is broad-based and slightly hooked at its lateral tip. 7 1 . Vertebral parapophyses sutured to one anoth- er (CI = 100). Breitensteinia. The vertebral parapophyses in catfishes, which are fused to their respective centra, are primitively separated from each other along the vertebral col- umn. In Breitensteinia, the parapophyses are su- tured to their neighbors at their anterior and pos- terior borders. These sutures seem to markedly limit the lateral movements of the vertebral col- umn. 72. Vertebrae with transverse processes (CI = 33, RI = 50). Acrochordonichthys, Breitensteinia, Parakysis, Si- sor. Aspredinidae. Vertebral transverse processes are lacking in most siluriforms, a plesiomorphic condition for the order. In the taxa above, a transverse process extends horizontally from the centrum of at least some caudal vertebrae (Fig. 30). These processes are considered neomorphs, not modified par- apophyses, because some vertebrae have both structures. The maximum development of the processes is observed in Breitensteinia, where they are present in most vertebrae, with those along the caudal region extending laterally as far as the skin surface. In the posterior portion of the caudal re- gion in that genus, the processes are as broad as the length of their vertebral centra, forming a near- ly continuous horizontal bony shelf alongside the vertebral column. The structures are also well de- veloped in aspredinids. The least developed con- dition is in Parakysis, where vestigial transverse processes are present only the posteriormost cau- dal vertebrae. In Sisor, most of the transverse pro- cesses are associated with the lateral bony plates formed from modified lateral-line ossicles (see character 111). The distal margin of each process contacts the mesial surface of its corresponding bony plate, sometimes forming a well-defined ar- ticulation. 73. Reduced number of pleural ribs (CI = 50, RI = 81). State 1 (five ribs): Conta. State 2 (three or four ribs): Parakysis, Er- ethistes, Erethistoides, Hara, Laguvia, Psendolaguvia, [Aspredinidae = ?]. Siluriforms primitively have six or more pairs of fully formed pleural ribs. In the taxa above, there is an apomorphic reduction in the number of pleural ribs to five or fewer. The numbers are consistent enough to allow recognition of two stales in this character, one of which (state 2) represents a more extreme reduction. Conta usually has five normal pairs of ribs and is the only taxon assigned state 1 , although a sixth vestigial rib is present on one side of one specimen examined. The closest condition outside the assemblage above is found in Sisor, but here all six pairs are well developed, which distinguishes it from the variable condition observed in Conta. All other members of Sisori- 34 FIELDIANA: ZOOLOGY dae, Akysidae, and Amblycipitidae have seven or more pleural ribs. All aspredinids have state 2, except for Xiliphius, which shows the plesio- morphic condition; hence the uncertain coding for that taxon. 74. First two pleural ribs flattened and expanded (CI = 100). Conta. The pleural ribs in catfishes are primitively sim- ple rods of bone, sometimes slightly flattened, elongated, and tapering only slightly from base to tip. The first rib is usually somewhat thicker than the remaining ones but retains the same overall morphology. In Conta, the first and second ribs are flat and expanded, and laminar for most of their length. The modification is evident in both ribs but is more marked in the first, which has an elongate leaf-like shape. The condition differs from that observed in Pseudecheneis (cf character 70) because in that taxon the dorsal expansion is not part of a general flattening of the rib. 75. Parapophysis of first rib-bearing vertebra di- rected anteriorly (CI = 100). Gagata. The parapophyses of free vertebrae in most members of the ingroup and in catfishes in general are primitively directed ventrolaterally or poster- oventrolaterally. Uniquely in Gagata, the direc- tion of the parapophysis of the first rib-bearing vertebra (probably the seventh in this taxon) has a strong anterior component, and the main axis of the parapophysis is directed anteroventrolater- ally (see Fig. 47). Pectoral Fins and Girdle 76. First proximal radial an elongated cartilagi- nous rod (CI = 50, RI = 50). Parakysis, Amblyceps, Liobagrus. The first proximal radial in siluriforms is ple- siomorphically a compact and roughly round car- tilaginous structure, often with an anterior process that extends somewhat into the base of the pectoral spine. In the taxa above, the first proximal radial is modified into an elongate rod of even width and roughly rectangular profile (Fig. 3 1 ). A nearly iden- FiG. 3 1 . Parakysis anomalopteryx, cas 4942 1 , pec- toral fin ray support, ventral view. Larger stippling rep- resents cartilage. Scale bar = 1 mm. tical configuration of the first proximal radial is seen in Chacidae (Brown & Ferraris, 1988; pers. obs.), an occurrence postulated as homoplastic in view of the large number of nodes separating chac- ids from amblycipitids and Parakysis in siluriform phylogeny (de Pinna, 1993). 77. Anterior half of segments of pectoral fin spine elongate, almost parallel to axis of spine (CI = 33, RI = 50). Acrochordonichthys, Akysis, Parakysis, Ambly- ceps, Liobagrus. The peculiar structure of catfish pectoral spines has been studied in detail by Reed (1924a). In the plesiomorphic condition for the order, the various segments are disposed in an imbricated pattern that forms the shaft of the spine. Each of the seg- ments that form the spine has roughly a short "V" morphology (Fig. 32A), with the two halves ap- proximately the same size. The segmentation pat- tern is usually still evident in the fully ossified DE PINNA: ASIAN CATHSH FAMILIES 35 ,V,i7 ft-'v {.i >. ;> i.. <■ j I .i,i."i yi i^'ij >> li ■ ; ! I • 1 1 iV'ii <"' ! i '■' V " I-)' -f.- B Fig. 32. Pectoral spines. A, Glyptothorax pectinoptenis, bmnh 15445, left spine, ventral view (inverted). B, Akysis sp. 1, UMMZ 214907, left spine, dorsal view. First free segment in A partially ossified. Larger stippling represents unossified areas. Scale bars = 1 mm. adult stage, and each segment extends only for a small portion of the total length of the spine. In the derived condition, the anterior half of each segment is elongated to such an extent that the actual pattern of formation of the spine becomes partly obscured (Fig. 32B). The long arms of ad- jacent segments are nearly parallel to each other and to the main axis of the spine. These modifi- cations seem to be associated with an extremely fine tip of the spine in the fishes with the derived condition. 78. Proximal pectoral radials 2 and 3 flared at distal tip (CI = 100). Glyptostemoids. The ossified proximal pectoral radials in siluri- forms primitively expand evenly and gently from base to tip. In glyptostemoids, these radials ex- pand enormously near their distal end, assuming a flared morphology (Fig. 33). The greatly ex- panded distal end of the radials is apparently as- sociated with the large number of pectoral fin rays in those fishes (character 80), which are disposed in a round fan-like arrangement. 79. Coracoid with ventral process (CI = 33, RI = 75). Bagarius, Conta, Erethistes, Erethistoides, Glyp- tothorax, Hara, Laguvia, [Nangra = ?], Pseude- cheneis, Pseudolaguvia, Sisor, Aspredinidae. The coracoid in siluriforms primitively lacks major processes on its ventral surface. In the taxa above, there is a well-defined process extending directly posteriorly from the ventral surface of the bone (Fig. 34). The length of this process varies widely among taxa, ranging from shorter than the anteroposterior width of the cleithrum to long enough to reach or pass the anterior tip of the basipterygium. In some cases, the shaft of the pro- cess is extended anteriorly and ankyloses or fuses with the ventral surface of the cleithrum. Within Nangra, the process is absent in A^. nangra and present, albeit small, in N. viridescens, a distri- bution that casts doubt on the precise coding of this character in the genus. In Bagarius, there is a 36 FIELDIANA: ZOOLOGY well-defined flange ventrally in the coracoid, in the position of and with an orientation corresponding to the basal portion of the process in other taxa. This flange is considered homologous to the pro- cess, representing either a very reduced or an in- cipient condition in its development. Mo and Chu (1986, p. 343) illustrated this process in some spe- cies of Glyptothorax that were not available for direct examination in this study. Elsewhere within siluriforms, a cleithral process similar to that described above is present in scol- oplacids (Bailey & Baskin, 1976, who also noted its presence in doradids; see also Schaefer, 1990), many if not all loricariids (Schaefer, 1987), dor- adoids, and probably also callichthyids (where the putative process is obscured by large bony expan- sions). In all cases, however, these outgroups have been shown to be part of larger assemblages whose other members lack the derived condition (cf Schaefer & Lauder, 1986; Schaefer, 1987; de Pin- na, 1993). 80. Eleven or more pectoral fin rays (CI = 100, RI = 100). Glyptostemoids, Pseudecheneis. There is a large variation in the number of pec- toral fin rays in siluriforms. For the group under study, the plesiomorphic count of 8-10 rays seen in most catfish outgroups is also postulated here to be plesiomorphic. In Pseudecheneis and glyp- tostemoids, pectoral fin rays range from 1 1 {Ex- ostomd) to 1 7 (Pseudexostoma). Fig. 33. Pseudexostoma yunnanensis, nrm 25124, left pectoral fin ray support, ventral view. Larger stip- pling represents cartilage. Scale bar = 1 mm. 8 1 . Humeral process absent (CI = 50, RI = 50). Parakysis, glyptostemoids, Pseudecheneis. The humeral process of the pectoral girdle, a roughly triangular, pointed process directed pos- teriorly on the upper limb of the cleithrum, is present in the vast majority of catfishes. It is al- most always present in the group under study, although in some taxa it is poorly developed, as in amblycipitids. The only ingroup taxa where the humeral process is completely absent are Para- kysis, the glyptostemoids, and Pseudecheneis. In Pseudecheneis, a small splint of bone marking the attachment of the humerovertebral ligament (cf character 64) may represent a remnant of the hu- meral process. If so, the character is still valid as evidence at the level of generality proposed be- cause the condition in Pseudecheneis is still the most reduced in the ingroup other than that in glyptostemoids and Parakysis. 82. Upper limb of cleithrum with mesially di- rected laminar expansion (CI = 100, RI = 100). Glyptostemoids, Pseudecheneis. The plesiomorphic upper limb of the catfish cleithrum does not extend markedly toward the midline. The space between the two upper limbs of the cleithra are thus filled with unconstrained soft tissue. In glyptostemoids and Pseudecheneis, the upper limb has a large mesially directed lam- inar expansion, which in some cases (as in Pseu- dexostoma) forms almost a complete bony wall separating the head from the trunk regions, leaving only a narrow passage for the esophagus. DE PINNA: ASIAN CATHSH FAMILIES 37 Fig. 34. Conta conta, ummz 208632, pectoral girdle, ventral view. Arrow indicates jwDsterior process on coracoids. Scale bar = 1 mm. 83. Anterior margin of pectoral fin spine with notch visible in dorsal view (CI = 100). Akysis. The fully developed pectoral fin spine in cat- fishes, despite its rather complex morphogenesis (Reed, 1924a), is primitively a cylinder of bone, roughly round or ovoid in cross-section. Although serial structures such as dentations or serrations are frequently present, there are no significant al- terations of the shape of the shaft of the spine. As seen for character 87, some ingroup taxa have pouch-like depressions in the dorsal- and some- times also in the pectoral fin spines. Uniquely in all Akysis species examined, the pectoral fin spine additionally has a well-defined notch located ap- proximately at the midlength of its anterior margin (Fig. 32B). This notch is easily visible in dorsal aspect in cleared-and-stained prepar^ations and is associated with the pouch-like modifications in the same spine of those fishes. A similar notch is sometimes also observed in the anterior margin of the dorsal fin spine, but there it is usually less well defined. The pectoral spine notch in Akysis prashadi was illustrated by Hora (1936, p. 201, Fig. 1). 84. Anterior margin of pectoral spine with ser- rations (CI = 50, RI = 83). Conta, Erethistes, Erethistoides, Hara, Laguvia, Pseudolaguvia, Sisor, [Aspredinidae = ?]. The posterior margin of the pectoral spine in catfishes primitively has a series of pointed den- ticulations, usually slightly curved toward the base of the spine. The anterior margin of the spine, contrastingly, lacks such dentations in the primi- tive condition. In the derived condition, serrations similar to those along the posterior margin are also present along the anterior margin of the spine. In aspredinids, the derived condition is present in all species examined except for Amaralia, Xyliphius, and hoplomyzontins, hence the doubtful condi- tion for Aspredinidae. 85. Internal support for pectoral fin rays small in size (CI = 50, RI = 75). Conta, Erethistes, Erethistoides, Hara, Aspredi- nidae. The internal support for pectoral fin rays in silu- riforms, composed of distal and proximal radials, primitively forms a complex whose lateral (ray- bearing) surface is at least twice as long as the 38 FIELDIANA: ZOOLOGY diameter of the base of the pectoral spine. In the derived condition, the internal ray supports are small in overall size, their ray-bearing surface be- ing shorter or only slightly longer than the basal diameter of the pectoral spine. Dorsal Fin and Support 86. Extensive contact between anterior and pos- terior nuchal plate elements (CI = 100, RI = 100). Conta, Erethistes, Erethistoides, Hara. Laguvia, Pseudolaguvia. In many siluriforms, the first and second pter- ygiophores of the dorsal fin have dorsal expansions on the body surface. These expansions, which are often exposed on the surface of the skin, often appear as plates and indeed are referred to as nu- chal plates. Primitively in catfishes, there are three nuchal plates: anterior, median, and posterior (Royero, 1987). In the ingroup in question, the primitive anterior and median plates are repre- sented by a single ossification (it is still unclear whether this happens by fusion or loss of one of the elements), a trait the ingroup shares with a variety of other siluriforms. There is much vari- ation of detail in the morphology of the nuchal plates, but when more than one is present their contact is narrower than the anteroposterior width of each individual plate. In the taxa above, the pterygiophore expansions are adpressed to one an- other along their common surface and the length of contact is larger than the anteroposterior width of each individual plate (Fig. 35). 87. Second dorsal fin spine with medial ridge along its anterior surface, forming bilateral longi- tudinal pouches (CI = 33, RI = 50). Acrochordonichthys, Akysis, Parakysis, Ambly- ceps, Liobagrus. The anterior surface of the catfish second dorsal fin spine is primitively a simple convex structure from base to tip. In the derived condition, the anterior surface of the spine has two deep lateral longitudinal grooves, which leave a medial ante- rior keel-like ridge between them. The soft tissues associated with the anterior surface of the spine form a pair of bilateral pouches along the grooves. The whole complex in these taxa may serve as storage for some substance associated with the dorsal spine. The integument around the spine in ANP+MNP PNP Fig. 35. Hara hara, ummz 208748, nuchal plates, dorsal view, dorsal fin spines removed. Arrow indicates contact between anterior and posterior nuchal plate el- ements. Scale bar = 1 mm. some of the taxa above is easily retracted toward the base of the spine shaft, revealing an extremely sharp tip and in the process squeezing out what- ever content is inside the pouches. This arrange- ment suggests the presence of a toxic substance associated with function of the dorsal spine as a defense mechanism. The presence of gland cells associated with the dorsal and pectoral spines in some siluriforms has been well documented by Reed (1924b, pp. 257- 258), who stated that the spines [in catfishes] are covered with integument, so that when inflicting a wound it is necessary for the spine to first puncture the skin which covers it. In addition to the skin wound thus produced, the growing end of the spine is multilated [sic], since the devel- oping caps are soft and not firmly attached to the next older cap. Reed (1924b, p. 257) contrasted this condition with that seen in other teleosts with poisonous secretions associated with fin spines where the spines are naked in the distal portion, and are so related to the gland and surrounding integument that there is formed a mobile sheet, which easily slips away from the af>ex of the spine, and at the same time forms a reservoir in which secretion may be stored against later needs. The derived condition observed for the taxa above is intermediate between the two conditions de- scribed by Reed. The skin covering the spine is retractable and forms a reservoir, but it is obvi- ously punctured by the spine in action, which also loses its soft tip in the process. A detailed histo- DE PINNA: ASIAN CATFISH FAMILIES 39 Fig. 36. Conta conta, ummz 208632, dorsal fin spines, lateral view. Scale bar = 1 mm. logical study of the dorsal fin spine in akysids and amblycipitids is necessary to understand the func- tion of the structure in those taxa. 88. Ventral arms of first dorsal fin spine with pos- terior processes attached dorsal to their tip (CI = 100, RI = 100). Bagarius, Conta, Erethistes, Erethistoides, Gaga- ta, [glyptostemoids = ?], Glyptothorax, Mara, La- guvia, Nangra, [Pseudecheneis = ?], Pseudolagu- via, [Sisor = ?], [Aspredinidae = ?]. The first dorsal fin spine in catfishes is proxi- mally divided into two broad arms not particularly elongate ventroposteriorly. The ventral tips of each of these arms are sometimes extended posteroven- trally in the form of a long process continuous with the arms themselves, as in amblycipitids, most akysids, and many doradoids. In the derived con- dition for the taxa above, the process of each arm is inserted not at its tip but rather at a point dorsal to it (Fig. 36). This insertion site makes the whole arm of the spine appear bifurcated, with one side of the bifurcation continuous with the rest of the spine and the other diverging posteroventrally from it. The taxa with this character coded as missing are those in which the whole first spine is absent or extremely reduced (aspredinids, glyptoster- noids, Sisor) or in which the spine is present but lacks ventral processes {Pseudecheneis). 89. Element corresponding to second dorsal fin spine not spinous (CI = 100, RI = 100). Glyptostemoids, Pseudecheneis. The second dorsal fin spine in catfishes (often referred to as the dorsal spine) is primitively mod- ified into a pungent structure, with a morphology unique for the group (Reed, 1 924a). A typical sec- ond spine is present in nearly all members of the group studied, with the exception o{ Pseudecheneis and the glyptostemoids. Those two exceptions are clearly derived within siluriforms. Even in other members of the ingroup with a nonpungent spine, such as Sisor, the morphology of the ray retains the condition primitive for catfishes, with a series of imbricated V-shaped segments formed by suc- cessive branchings of one half-ray. Only in the two taxa above is the ray completely nonspinous, com- posed of simple rectangular lepidotrichia (except for the processes for the pinnate expansions of the dorsal fin in some glyptostemoids) similar to those seen in more posterior dorsal fin rays. 90. Posterior nuchal plate with anterior process forming facet for articulation with anterior nuchal plate (CI = 100, RI = 100). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Bagarius, Conta, Erethistes, Erethistoides, Gagata, glyptostemoids, Glyptothorax, Hara, La- guvia, Nangra, Pseudecheneis, Pseudolaguvia, Si- sor, Aspredinidae. Primitively in siluriforms, the anterior and pos- terior nuchal plates, when they contact each other, lack facet-like bony structures at the area of con- tact. In amblycipitids, the distal margin of the pos- terior plate is twisted dorsally to meet its coun- terpart anteriorly and may be somewhat thickened 40 FIELDIANA: ZOOLOGY at this site. However, it lacks a well-differentiated process. In the apomorphic state, the posterior nu- chal plate has a well-defined strut of bone forming a facet that articulates anteriorly with the anterior nuchal plate element. This process is well differ- entiated and is not simply an enlargement of the distal margin of the plate. It is located proximally relative to the lateral margin of the plate. The anterior and posterior nuchal plates are very re- duced in glyptostemoids and do not contact each other. Often, the pterygiophore portion corre- sponding to the first nuchal plate is altogether ab- sent in that taxon. However, when both plates are present, they are linked by a ligament that occupies the same position as a bony bridge in other taxa. There is a bony protuberance on the anterior mar- gin of the posterior nuchal plate at the site where the ligament attaches. This protuberance is pos- tulated as homologous to the process seen in the remaining terminal taxa with the derived condi- tion, therefore serving as evidence that glyptoster- noids display the apomorphic state for this char- acter. The extensive surface of contact between the first and second nuchal plates in aspredinids and all taxa referred to below in the family Ereth- istiidae modifies the process such that it is wider than long. The facet-like structure of the contact is still evident, however, on the mesial portion of the posterior plate. The anterior and posterior nu- chal plates are distally fused to each other in adult specimens of Parakysis verrucosa examined in the area of interest for this character, which makes it uncertain whether or not the process is primitively present in this taxon. Nevertheless, this fusion has apparently not yet taken place in the much smaller examples of P. anomalopteryx available; these clearly show an anterior process on the posterior nuchal plate (Fig. 37). The process at this stage is linked to the anterior nuchal plate by a ligament in an arrangement similar to that described above for some glyptostemoids. On the basis of this data, Parakysis is also assigned the derived condition for this character. Examination of a more com- plete ontogenetic series will be necessary to con- firm the coding of this character for both Parakysis and glyptostemoids. 9 1 . Ventral tip of first dorsal fin pterygiophore and corresponding neural spines with con- tacting facets (CI = 100, RI = 100). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Amblyceps, Liobagrus, Bagarius, Conta, Erethistes, Erethistoides, Gagata, glyptostemoids, PNP Fig. 37. Parakysis anomalopteryx. CAS 49421, pu- tatively juvenile condition of dorsal fin spines and nuchal plates, dorsal view. Arrows indicate anterior processes on posterior nuchal plates. Wavy lines represent liga- ments. Scale bar = 1 mm. Glyptothorax, Hara, Laguvia, Nangra, Pseudeche- neis, Pseudolaguvia, Sisor, Aspredinidae. The first dorsal fin pterygiophore in siluriforms usually extends ventrally to fit between the arms of the bifurcated neural spine of the complex cen- tmm (Royero, 1 987). Primitively within the order, both the ventral portion of the pterygiophore and its corresponding neural spine are formed mostly of smooth laminar bone, frequently with long lam- inar flanges for muscle insertion. In all terminal taxa treated in this study, the ventral portion of the first pterygiophore and the distal tip of the neural spine of the complex centrum are composed of thick, rugose nonlaminar bone. Additionally, the surfaces of both structures are expanded at their region of contact, forming a facet-like thick- ening on the lateral face of the pterygiophore. The mgose aspect is more pronounced where the sur- faces are in contact, i.e., the lateral surface of the pterygiophore and the mesial surface of the distal tip of the neural spine. All these characteristics seem to represent an incipient stage of the mod- ifications that lead to the formation of a stridu- latory apparatus, described below as a character of a more restrictive clade (cf character 109). DE PINNA: ASIAN CATFISH FAMILIES 41 Fig. 38. Erethistes pusillus, ummz 208697, right lac- rimal, dorsal view, anterior on top. Scale bar = 1 mm. 92. Posterior nuchal plate straight and thin, its width approximately even from base to tip (CI = 100, RI = 100). [Bagarius = ?], Gagata. Nangra. The plesiomorphic condition for the posterior nuchal plate in siluriforms is that of a wide plate of bone, approximately twice as long as broad and expanding gently from base to tip. Its margins are usually curved, forming a convex profile. In Ga- gata and Nangra, the posterior nuchal plate is dif- ferent from that primitive condition in several re- spects. It is roughly five times as long as broad, its width either narrows or remains constant from base to tip, and its margins are straight in dorsal view. The character is variable in Bagarius, with the plesiomorphic condition seen in small B. ba- garius (ummz 1 86793) and the apomorphic state clearly present in the largest B. yarrelli available for study (iacm 38294-55). Medium-sized B. yar- relli show a condition intermediate between the derived and primitive states, which seems to in- dicate that the apomorphic state may only become evident in large specimens of the genus. However, in the absence of any ontogenetic information on species other than B. yarrelli, this character is cod- ed as missing for Bagarius. Laterosensory System 93. First lateral-line ossicle enlarged, adjoined to posterior portion of supracleithrum (CI = 50, RI = 75). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Aspredinidae. The lateral-line sensory canal in siluriforms is usually enveloped by a series of tubular ossicles, believed to be the only remnants of scales in the fishes of this order (Fink & Fink, 1981). In the plesiomorphic condition, the first (anteriormost) ossicle is similar to those following and is loosely connected with the portion of the supracleithrum where the lateral-line canal emerges from the skull. The lateral-line canal, therefore, runs exclusively through soft tissue for a distance between the back of the skull and the first ossicle. In the taxa above, the first ossicle is more heavily ossified than the subsequent ones and is also larger and more com- plex in shape. Additionally, it is closely adjoined to the back of the skull (Fig. 1 0). The lateral-line canal emerges from the supracleithrum and goes directly into the first ossicle, being thus almost continuously enclosed in bone until the posterior end of the first ossicle. Within Akysis species, the character is fully developed only in Akysis sp. 1 , which includes the largest specimens of the genus available for study. In remaining Akysis species, the derived condition is not as clearly present, perhaps as a result of the juvenile condition of these small specimens. The first lateral-line ossicle, however, is clearly larger than those remaining and is more closely associated with the posterior por- tion of the neurocranium than it is in the plesio- morphic condition. 94. Lacrimal modified into a bowl-shaped struc- ture located ventrolaterally to nasal capsule (CI = 50, RI = 86). Conta, Erethistes, Erethistoides, glyptosternoids. Hara, Laguvia, Pseudecheneis, Pseudolaguvia. The lacrimal (= first infraorbital) in catfishes displays a large amount of morphological varia- 42 FIELDIANA: ZOOLOGY tion not yet interpreted in a phylogenetic context. Usually it is composed of a tubular portion that encloses the anterior part of the infraorbital later- osensory canal, a plate-like portion developed to different degrees among various taxa, and one or more processes in various positions. Although it is futile to attempt to describe what would be the generalized primitive condition of the whole lac- rimal for siluriforms, some unique conditions can confidently by hypothesized as apomorphic. The condition in the taxa above is one of them. In those taxa, the lacrimal is modified into a large and simple bowl-like structure, adpressed to the ventrolateral surface of the nasal capsule and ap- parently providing partial shielding for it (Fig. 38). The canal portion normally associated with the lacrimal is still present and extends along the dor- sal portion of the bowl without interrupting its smooth profile. 95. Nasal with expansions beyond canal-bearing portion (CI = 100). Akysis. The nasal in catfishes is primitively a simple tubule lining a corresponding portion of the later- osensory canal system. It normally lacks any sig- nificant expansions of bone beyond the canal- bearing tube. In all Akysis species examined, the nasal has shelf-like expansions on its dorsal sur- face, extending considerably beyond the canal- bearing tube along its whole length (Fig. 39). The nasal in Acrochordonichthys is considerably thick- ened compared with the plesiomorphic condition in catfishes. Despite the heavy ossification, how- ever, the extent of ossification beyond the canal- bearing portion is minimal in that genus and very different from the condition observed in Akysis. 96. Nasal elongated anteriorly to the level of an- terior margin of mesethmoid comua (CI = 33, RI = 0). Bagarius, Nangra, Aspredinidae. The nasal bone in catfishes primitively does not extend anteriorly to the anterior margin of the mesethmoid comua. In aspredinids, Bagarius, and Nangra, the nasal is elongated, and its anterior tip reaches or passes the anterior margin of the un- derlying mesethmoid comua. Fig. 39. A kysis leucorhynch us. usnm 1 09 6 3 6 , left na- sal, dorsal view, anterior to right. Scale bar = 1 mm. 97. Anterior portion of lateral line mnning closely in parallel to lateral margin of Weberian lam- ina (CI = 100, RI = 100). Conta, Erethistes, Erethistoides, Hara, Laguvia, Pseudolaguvia, Aspredinidae. The lateral-line sensory canal and associated os- sicles in catfishes primitively follow a path along the body of the fish without any close associations with internal skeletal structures. In the derived condition, the anterior portion of the canal follows closely the ventral rim of the lateral margin of the Weberian lamina (Figs. 1 1 , 25). In cases where the lamina is curved, the canal follows the curve of the lamina. 98. Path of infraorbital laterosensory canal straight (CI = 50, RI = 50). Acrochordonichthys, Akysis, Breitensteinia. The infraorbital branch of the laterosensory sys- tem of the head in catfishes primitively follows a curved path roughly circling the ventral profile of the orbit (Arratia & Huaquin, 1995). In the taxa above, the infraorbital canal is straight or nearly so from the sphenotic to the lacrimal (Fig. 10). 99. Lacrimal with dorsal mesially directed shelf- like portion (CI = 50, RI = 75). Acrochordonichthys, Akysis, Breitensteinia, Par- akysis, Aspredinidae. The lacrimal (first infraorbital) in siluriforms displays a large amount of morphological varia- DE PINNA: ASIAN CATFISH FAMILIES 43 Flo. 40. Left lacrimals in dorsal view, anterior on top. A, Dysichthys sp., amnh uncat. B, Acrochordonich- thys rugosus, zrc 4415. Scale bar = 1 mm. tion, which makes it difficult to hypothesize its general plesiomorphic morphology (see discussion of character 94). Rather, it seems that the only way to adequately describe the phylogenetically informative variation in that bone is to subdivide it into various individual characters representing subsets of its general morphology. The character described here relates to the dorsal margin of the lacrimal, an area that in most catfishes is flat and devoid of marked modifications of shape. In the taxa above, the dorsal surface of the lacrimal has a large curved expansion extending dorsomesially (Fig. 40). The expansion is thick, not laminar, and usually has ornamentations on its surface. 100. Nasal adpressed to margin of neurocranium for posterior two thirds or more of its length, Fig. 41. Nangra nangra, cmk 6369, dorsal portion of preopercular branch of laterosensory canal system and related structures. Larger stippling represents cartilage. Scale bar = 1 mm. anterior portion abruptly diverging laterally (CI = 100, RI = 100). Erethistes, Erethistoides, Hara. The nasal in catfishes is primitively located along the lateral margin of the mesethmoid but sepa- rated from it throughout its length by a space great- er than the width of the nasal. In the derived con- dition, the posterior half or more of the nasal's length is adpressed to the lateral margin of the mesethmoid, and its anterior portion abruptly di- verges laterally. 101. Dorsal half of suprapreopercle strongly curved posteriorly, making nearly 90° angle with ventral half of bone (CI = 100, RI = 100). Gagata, [Glyptothorax = ?], Nangra. The suprapreopercle, a common ossification in the ingroup under study and many other catfishes, is a tubular ossicle lining the branch of the sensory canal system between the dorsal tip of the pre- opercle and the pterotic. Primitively in the order, the suprapreopercle is a roughly straight tube, fre- quently inclined posteriorly relative to the dorsal tip of the preopercle. In the derived state, the dor- sal half of the suprapreopercle is strongly curved posteriorly, forming an almost 90° angle with its ventral half (Fig. 41). Within Glyptothorax, there is widespread vari- ation in the states of this character, not only among 44 FIELDIANA: ZOOLOGY the various species but also on occasion between different specimens of the same species, and some- times even between the right and left sides of the same individual. This variation cannot be ade- quately treated in this study, and this character is coded as missing for Glyptothorax. 102. Hyomandibula with bony tunnel enclosing branch of the laterosensory canal between preopercle and pterotic (CI = 100). Aspredinidae. As described for character 101, the branch of the laterosensory canal system between the pre- opercle and the pterotic is typically enclosed in an independent ossification called the suprapreoper- cle. In aspredinids, that canal branch is instead enclosed in a bony tunnel in the hyomandibula. This tunnel probably represents a suprapreopercle indistinguishably fused to the hyomandibula, be- cause the hyomandibula is primitively not a canal- bearing bone. Nonetheless, there is no sign of an independent suprapreopercle in even the youngest specimen of the family examined (a representative oTAspredo aspredo, usnm 226072) or in the pos- sibly paedomorphic hoplomyzontins. Pelvic Fins and Girdle 103. Basipterygium with ventral longitudinal keel, extending anteriorly alongside internal arm (CI = 50, RI = 85). Bagarius, Conta, Erethistes, Erethistoides, Gaga- ta, [glyptostemoids = ?], Glyptothorax, Hara, La- guvia, Nangra, Pseudecheneis, Pseudolaguvia, [Si- sor = ?]. The ventral surface of the basipterygium in cat- fishes is primitively flat, without prominent pro- cesses or flanges. In the taxa above, there is a con- spicuous laminar flange on the ventral surface of the basipterygium. It usually originates on the main portion of the bone and extends anteriorly along- side the ventral surface of its internal arm (Fig. 42). For most of its length, the structure is a low flange, but in the majority of taxa there is a well- defined, small portion where it is abruptly more prominent. The process is reduced but still present in Bagarius, Pseudecheneis, Erethistes, Erethis- toides, and Hara. Among glyptostemoids exam- EA VCB Fig. 42. Laguvia ribeiroi, ummz 208955, right bas- ipterygium, ventral view. Larger stippling represents car- tilage. Scale bar = 1 mm. ined, only Exostoma labiatum has the derived state and in a rather reduced condition (Fig. 43). There- fore, this character was coded as missing for glyp- tostemoids. DE PINNA: ASIAN CATHSH FAMILIES 45 Fig. 43. Exostoma labiatum, nrm 25105, right bas- ipterygium, ventral view. Scale bar = 1 mm. The flange is an attachment point for the pro- tractor ischii muscle, which pulls the pelvic girdle anteriorly when contracted (Saxena & Chandy, 1 966). Elsewhere within siluriforms, a very similar structure is present in Amphiliidae (Skelton, 1 985, pp. 275, 276, 287). However, here the flange is on the dorsal surface of the bone instead of the ven- tral, therefore indicating nonhomology. A some- what similar keel is present in callichthyids (Reis, 1993) and hoplomyzonlin aspredinids, but in these taxa, as in amphiliids, it is located on the dorsal surface of the basipterygium. 104. External arm of basipterygium abruptly ex- panded distally (CI = 100). Glyptostemoids. In the primitive condition, the anterior arms of the basipterygium in siluriforms expand only slightly or instead narrow from base to tip. In cases where they expand, the expansion proceeds along their whole length. In all glyptostemoids, the nar- row external arm of the bone is abruptly expanded near its anterior tip (Fig. 43). The internal arm retains the primitive morphology. 105. Posterior cartilage of basipterygium ramified (a = 100). Bagarins. The ppsterior cartilage of the basipterygium in siluriforms is primitively a relatively short struc- ture bordering a more or less extensive part of the posterior process of the bone (Shelden, 1937). In Bagarius, this cartilage is markedly elongated pos- teriorly and distally ramified into a short dendritic arrangement. A somewhat similar condition is seen in the subfamily Cetopsinae of the Cetopsidae (de Pinna & Vari, 1995). In the Cetopsinae, nonethe- less, the dendritic portion of the cartilage is far longer than that observed in Bagarius. This dif- ference, plus the fact that cetopsines and Bagarius are very distantly related in catfish phylogeny (de Pinna, 1 993), indicates that the two conditions are not homologous. 106. Cartilage on posterior process of basiptery- gium reduced or absent (CI = 100). Aspredinidae. The posterior process of the basipterygium in siluriforms is usually lined mesially and posteri- orly with a narrow but relatively extensive carti- lage, often continuous mesially with the interba- sipterygial cartilage. In the Aspredinidae, the car- tilage either is reduced to a small distal plug iso- lated from the interbasipterygial cartilage (in aspredinines and bunocephalins) or is totally ab- sent (in hoplomyzontins). 46 FIELDIANA: ZOOLOGY 107, Anterior arms of basipterygium undifferen- tiated from main part of bone (CI = 100). Aspredinidae. In nearly all siluriforms the basipterygium has two anterior arms with cartilages at their tips (Shelden, 1937). Uniquely in aspredinids, there is no differentiation between the body of the basip- terygium and its arms, and the whole pelvic skel- eton is represented by a single plate. Anterolat- erally along the edge of the bone, a single cartilage extends from the region of fin-ray attachment to the anterior end of the basipterygium. Anterior basipterygial arms are also undifferentiated in scoloplacids (Schaefer, 1990, pp. 196-197). How- ever, these arms are present in remaining lorica- rioids, indicating that the condition in that family is autapomorphic. Miscellaneous 108. Epural curved (CI = 100). Parakysis. The epural in catfishes, when present, is prim- itively a straight or nearly straight rod of bone in lateral view (cf Lundberg & Baskin, 1 969). In Par- akysis, the bone is markedly curved posteriorly, leaving an elliptical space between it and the uro- neural. 109. Stridulatory apparatus (CI = 66, RI = 80). State 1: Conta. State 2: Gagata, Nangra, Sisor. The presence of a stridulatory apparatus is among the most interesting morphological modifications seen in the ingroup under study. Although sound- producing adaptations are common and have long been known in siluriforms (see Geoffroy St. Hi- laire, 1829; Tavolga, 1962, 1971, 1977), the spe- cific mechanism seen in Conta, Sisor, Nangra, and Gagata is unique within the order. In these genera, the apparatus is formed by a modified contact be- tween the neural spine of the complex centrum and the first dorsal fin pterygiophore (actually formed as a ventral extension of the supraneural that forms the anterior nuchal plate). The close connection between the two structures in itself is plesiomorphic for catfishes. In the derived state, however, the surfaces of contact are markedly en- FiG. 44. Gagata gagata, amnh 8358, SEM photo- graph, hemal spines of complex centrum, anterolateral view. Scale bar = 1 mm. larged and mobile. Additionally, their surfaces of contact are crenulated, i.e., provided with a series of short parallel ridges on the lateral surface of the pterygiophore and on the mesial surface of the parapophysis (Figs. 44-47). These and other mod- ifications form a well-differentiated stridulatory apparatus. The sound is produced when the ridged surface on the ventral tip of the supraneural and its counterpart on the neural arch rub against one another, and it can be simulated by manipulating the head of preserved specimens (as reported by Haddon, 1881). Two major versions of the ap- paratus exist. In one of them, coded as state 1 and seen only in Conta, the ventral ridged processes of the pterygiophore are separate and each has its own ridged area that contacts the neural spine arm on its side (Fig. 46). Sound-producing contact can be both on the right and left sides at the same time. In the other version, coded as state 2 and seen in the other three genera, the ridged surfaces of the pterygiophore are fused into a single median process, forming a round bony swelling with ver- tical ridges on its surface (Fig. 47). In this second version, the sound-producing contact can take place on the right or left sides of the modified neural DE PINNA: ASIAN CATFISH FAMILIES 47 Fio. 45. Gagata gagata, amnh 8358, SEM photograph, distal portion of hemal spines of complex centrum, showing internal ridges of stridulatory apparatus. Scale bar = 1 mm. Spine but not on both simultaneously. States 1 and 2 are ordered in this way because the ventral arms of the supraneural are primitively separate, and their fusion is therefore a subsequent step. The stridulatory apparatus in Gagata was described by Haddon (1881), and that o^ Sisor was described by Mahajan (1963). The apparatus is well devel- oped and in accordance with state 2 in Nangra viridescens, but in A^. nangra it seems to be some- what atrophied and corresponds to state 1 . Nan- gra, therefore, shows some ambiguity with regard to this character's coding. Despite that, the char- acter is assigned state 2 instead of a missing code for this genus. The reason for this decision is that a missing code would imply, for parsimony cal- culations, that this character in Nangra could be assigned any of states 0, 1, or 2. However, the genus never lacks the stridulatory apparatus and consequently the character state can only be 1 or 2. Adequate treatment of this case will depend on methodological developments regarding treat- ment of missing codes for multistate characters in parsimony analysis. 1 10. Integument of thorax or thorax and abdo- men modified into adhesive apparatus formed by patches of unculi underlain by ridges of integument (CI = 40, RI = 25). State 1 (ridges oblique or parallel to body): Conta, Glyptothorax, Laguvia, Pseudola- guvia. State 2 (ridges transverse to body): Pseude- cheneis. There is an extensive literature dealing with the morphology and evolutionary significance of the hill-stream adaptations in sisorids (e.g., Hora, 1 922, 1923, 1930; Bhatia, 1950; Lai et al., 1966; Saxena & Chandy, 1966; Tilak, 1976). Among the most conspicuous of these specializations is the pres- ence of an adhesive apparatus on the ventral sur- face of the body. The structure is formed by in- tegumentary folds forming a pattern of ridges and grooves usually disposed in a roughly bilateral ar- rangement. The area covered by the ridges varies from a small arc immediately posterior to the branchiostegal membranes (as in some Glypto- thorax; cf. Jayaram, 1981, pp. 255-259) to nearly the whole thoracic and abdominal region (as in Conta). Covering the integument ridges are patch- es of numerous unculi (cf. Fig. 48), which are pro- jections of single epidermal cells of ostariophysans (Roberts, 1982; Arratia & Huaquin, 1995) and seem to play a major role in adhesion. Although densely distributed over the ridges, unculi are ab- sent in the grooves of the adhesive apparatus (first 48 FIELDIANA: ZOOLOGY soc WL ANP+MNP PNP Fig. 46. Conta conta, ummz 208632, posterior portion of neurocranium and Weberian complex (above) and dorsal fin support (below, spines removed), dorsal view. Arrows indicate ridges of stridulatory apparatus; those indicated by lower arrow are seen by transparency. Upper and lower poriions shown separated for clarity; their natural relative portion corresj>onds to an overlapping of the two arrows. Scale bar = 1 mm. DE PINNA: ASIAN CATHSH FAMILIES 49 ANP+MNP Fio. 47. Gagata gagata, amnh 8358, dorsal fin support and associated vertebral elements, lateral view. Arrow indicates ridged sphere that forms part of stridulatory apparatus. This structure is located between the ridged surfaces of the neural spines shown in Figures 44 and 45. S«ile bar = 1 mm. SO FIELDIANA: ZOOLOGY Fig. 48. Pseudecheneis sulcatus, bmnh 1985.9.16:50, SEM photograph of adhesive apparatus. Notice profile of individual epithelial cells. observed by Hora, 1 922) (Fig. 49). There is a large degree of morphological variation in the adhesive apparatus among various taxa. The most deviant condition is found in Pseudecheneis, where the skin ridges are thick, flat, and oriented transversely relative to the longitudinal axis of the body (Fig. 49). However, the condition of the adhesive ap- paratus in juveniles of that genus is similar to that in other taxa. The integumentary folds in young Pseudecheneis sulcatus are oblique and gradually assume a transverse orientation with growth (Ti- lak, 1 976). This observation not only corroborates that the adhesive apparatus in Pseudecheneis is homologous to that in other taxa but also indicates that transverse folds are derived relative to their oblique orientation. Therefore, this character is separated into two states, ordered accordingly. In state 1 , the ridges of the adhesive apparatus are oblique or parallel to the longitudinal axis of the body, and in state 2, the ridges have a transverse orientation. The presence of unculi, taken by itself, is ple- siomorphic, because they are probably primitively present in most ostariophysans (Roberts, 1982) (however, they are absent in diplomystids; see Ar- ratia & Huaquin, 1995). Also, the association of unculi with adhesive functions is not exclusive to sisoroids; it also occurs in the pectoral fin of am- philiids (Roberts, 1982), seemingly with a similar function. What defines the derived condition dis- cussed here is specifically the presence of well- defined unculi patches associated with prominent skin ridges on the thorax and/or abdomen. The whole structure is evident on the ventral aspect of the fish and normally visible with the naked eye. The fine structure of the adhesive apparatus in sisorids has been reported in a number of papers, summarized and added to by Roberts ( 1 982). More recently, additional SEM observations on Pseu- decheneis were reported by Singh and Agarwal (1991). For this study, SEM examinations corrob- orated observations previously made for that ge- nus (Figs. 48, 49). Additionally, the SEM structure of the apparatus in Conta and Laguvia was ex- amined for the first time (Figs. 50, 51) and was found to conform to the general morphology so far known for the thoracic adhesive apparatus in other taxa. The first in-depth account of the adhesive ap- paratus in sisorids was by Hora (1922). According to that author, each of the unculi-bearing epider- mal cells (then not referred to by this name) has a cell nucleus right at the base of the unculus; sometimes half of the nucleus is actually inside the unculus. This positioning probably indicates intense nuclear synthesis directed toward the un- DE PINNA: ASIAN CATFISH FAMILIES 51 Fig. 49. Pseudecheneis sulcatus, bmnh 1 985.9. 1 6:50, SEM photograph of adhesive apparatus, showing transverse folds and limit between ridge (with unculi) and groove (naked) areas. At right is an enlargement of area framed by rectangle at left. cuius region. The cell nuclei are located deeper in the body of those epidermal cells that lack unculi, i.e., those lining the grooves of the adhesive ap- paratus (Hora, 1922). The taxonomic generality of these observations about the position of the nucleus remains to be determined. 111. Bony plates on siuface of body (CI = 100). Sisor. The primitive catfish condition is the lack of bony plates on the surface of the body, as seen in diplomystids, cetopsids, pimelodis, bagrids, and all other basal or generalized siluriforms. An ar- mored condition is certainly derived within the order, but the definition of a bony plate is rather loose, and their structural basis can be highly vari- able. Sometimes in the same taxon, plate-like structures are formed from widely different ana- tomical bases. In Sisor, the body is covered by three series of bony plates, each having a different structural basis. The series along the lateral sides of the body are formed by expanded lateral-line ossicles, which posteriorly become closely asso- ciated with the distal margin of the vertebral trans- verse processes (see character 72). Meanwhile, the series along the mid-dorsal line is formed by au- togenous (Mahajan, 1967b, p. 313; contra Regan, 1911) plates that are wider than long, each of which articulates with the expanded dorsal tip of its cor- responding neural spine. The posteriormost mid- dorsal plate is twice as long as the others and bears a well-defined spine alongside the anterior margin of the adipose fin. Another series of plates extends on each side along the base of the dorsal fin. These plates are expansions of the distal portions of prox- imal dorsal fin pterygiophores and bear strong re- semblance to the posterior nuchal plate. There- fore, the three series of bony plates in Sisor are formed in markedly different ways. It might be argued that the structural differences among the various series indicate three separate characters. This question is irrelevant for the present case because no member of the ingroup has only one or two of the three series, and a subdivision of the character would therefore be uninformative for hypothesized relationships. Bony plates seem to have originated indepen- dently a number of times within siluriforms, namely in higher loricarioids (Callichthyidae, Scoloplacidae, Astroblepidae, and Loricariidae), Ooumeinae (Amphiliidae), and Doradidae. In each S2 FIELDIANA: ZOOLCXJY Fig. 50. Conta conta, ummz 208632, SEM photograph of adhesive apparatus, showing oblique folds (A) and unculi morphology (B). of these groups, the hypotheses presently available 1 1 2. First (uppermost) principal caudal fin ray en- (cf. de Pinna, 1993, and references therein) indi- cate that the armored clades are more closely re- lated to naked taxa than to each other. The ar- mored condition in Sisor is therefore hypothesized as uniquely derived. larged and elongated as a filament (CI = 100). Sisor. The various principal caudal fin rays in siluri- forms usually have a rather uniform morphology. DE PINNA: ASIAN CATFISH FAMILIES S3 ••^H Fig. 51. Laguvia ribeiroi, ummz 208955, SEM photograph of adhesive apparatus, showing obUque folds (A) and unculi morphology (B). Although their relative number and branching pat- tern may vary markedly among various taxa (Lundberg & Baskin, 1 969), their size and struc- ture change little from the uppermost to the low- ermost ray in a single iin of an adult individual. In Sisor, the uppermost principal caudal fin ray is markedly enlarged (Mahajan, 1967b). The diam- eter of its proximal portion is about five times that of remaining rays, and its base covers over half the total area of the upper hypural plate. Addi- tionally, the uppermost ray is elongated into a fil- ament considerably longer than the standard length S4 FIELDIANA: ZOOLOGY / / . :# / / / / / / ///////V /////////// / V / Fio. 52. Most parsimonious hypothesis of sisoroid relationships (CI = 70, RI = 79) based on 1 1 2 characters studied, with branches numbered. Characters transforming in each branch or terminal taxon are as follows. BRANCH 1: 1, 32, 61, 64, 77, 87, 91; BRANCH 2: 14, 20, 43, 51, 64, 76; BRANCH 3: 24, 26, 90; BRANCH 4: 13, 45, 64, 93, 98, 99; BRANCH 5: 2, 38, 41, 55, 72; BRANCH 6: 25, 33, 53; BRANCH 7: 27, 57, 62, 64, 66, 77 (reversal), 79, 87 (reversal), 88, 103; BRANCH 8: 30, 41, 60, 67; BRANCH 9: 4, 29, 63; BRANCH 10: 7, 8, 25, 44, 56, 109; BRANCH 11: 12, 34, 35, 41 (reversal), 92, 101; BRANCH 12: 49, 52, 110; BRANCH 13: 11, 24 (reversal), 27 (reversal), 57 (reversal), 59, 80, 81, 82, 89, 94; BRANCH 14: 3, 38, 84, 85, 97; BRANCH 15: 37, 54, 73, 86, 94, 110; BRANCH 16: 27, 58, 73; BRANCH 17: 19, 22, 41, 85 (reversal); BRANCH 18: 100; BRANCH 19: 65, 68; Akysis: 83, 95; Parakysis: 9, 15, 17, 18, 39, 40, 73, 76, 81, 98, 108; Breitensteinia: 71; Bagarius: 10, 21, 46, 50, 96, 105; Sisor. 3, 72, 84, 1 1 1, 1 12; Gagata: 69, 75; Glyptothorax: 10, 31; Pseudecheneis: 16, 36, 42, 70, 1 10; glyptoster- noids: 3, 5, 6, 41 (reversal), 64 (reversal), 78, 79 (reversal), 104, 110 (reversal); Aspredinidae: 23, 28, 47, 72, 93, 96, 99, 102, 103 (reversal), 106, 107; Conta: 74, 109; Pseudolaguvia: 48. of the fish (Mahajan, 1967b). A certain degree of elongation of the uppermost principal caudal fin ray is not unusual in catfishes and other teleosts, being usually associated with deeply forked caudal fins. In cases where the upper lobe of the fin is elongated, the uppermost principal ray may be filamentous to a certain degree, such as in Hara filamentosa Blyth and in the genus Conta (cf Hora, 1950). However, the condition of the uppermost ray in Sisor is unique in its extreme elongation and overall enlargement relative to the remaining caudal fin rays. Phylogenetic Analysis The 1 1 2 characters described above were coded for each terminal taxon, forming the data matrix shown in Table 1 . This matrix was analyzed with the Hennig86 computer program (Farris, 1988). The exact algorithm ie found a single most par- simonious cladogram (Fig. 52), with a length of 167 steps, CI = 70, and RI = 79. An identical result was obtained using the much faster com- bination of heuristic algorithms mh*; bb*;. The same hypothesis was also encountered with PAUP DE PINNA: ASIAN CATFISH FAMILIES 55 Table 1. the text. Data matrix for superfamily Sisoroidea. Order of characters follows that presented and numbered in Hypothetical ancestor 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 0000000000 00 Acrochordonichthvs 1100000000 0010000000 0001110000 0110000100 1000100000 0010100000 1002000000 0100001000 0000001001 1010000110 0000000000 00 Akvsis iobooooo?o 0010000000 0001010000 0100000000 0000100000 0000000000 1002000000 0000001000 0010001001 1010100110 0000000000 00 Breitensteinia 0100000000 0010000000 0001110000 0110000100 1000100000 0010100000 1002000000 1100000000 0000000001 1010000110 0000000000 00 Parakysis iioodoooio 0000101100 0001010000 0100000111 1000000000 0000100000 100?000000 0120011000 1000001001 1010000010 0000000100 00 Amblvceps looobooooo 0001100001 0000000000 0100000000 0010000000 1000000000 1001000000 0000011000 0000001000 1000000000 0000000000 00 Liobagrus 1000000000 0001000001 0000000000 0100000000 0010000000 1000000000 1001000000 0000011000 0000001000 1000000000 0000000000 00 Bagarius 1001000001 0000000000 1001011011 0100000000 1000010001 0000001001 1113011000 0000000010 0000000101 1?00010000 0010100000 00 Conta 1?10000000 0000000000 0001011000 0100001100 0000000000 0001001000 1103010000 0011000010 0001110101 1001001000 0010000011 00 Erethistes 1000000000 0000000000 0001012000 0100001100 0000000000 0001001100 1103110100 0020000010 0001110101 1001001001 0010000000 00 Erethistoides 1000000000 0000000000 0001012000 0100001100 0000000000 0001001100 1103010000 0020000010 0001110101 1001001001 0010000000 00 Gagata 1001001100 0100000000 000111101? 0101100000 0001000000 0000011001 1113011010 0000100000 0000000101 1100000000 1010000010 00 glyptostemoids 1?1011000? 1000000000 0000010001 010000000? 0000000010 0100000011 1100011000 0000000101 1100000711 1001000000 00? 1000000 00 Glyptothorax 1000000001 0000000000 0001011001 1100000000 1000000010 0100001001 1103011000 0000000010 0000000101 1000000000 ?010000001 00 Hara 1000000000 0000000000 0001012000 0100001100 0000000000 0001001100 1103110100 0020000010 0001110101 1001001001 0010000000 00 Laguvia 1000000000 0000000010 0101012000 0100001100 10000000?0 0001001100 1103010000 0020000010 0001010101 1001001000 0010000001 00 Nangra 1001000100 0100000000 000111101? 0101100000 0000000000 0000011001 1113011000 00000000?0 0000000101 1100010000 1010000010 00 Pseudecheneis 1000000000 1000010000 0000010001 0100010000 1100000010 0100000011 1103011001 0000000011 1100000?11 1001000000 0010000002 00 Pseudolaguvia 1000000000 0000000010 0100012000 0100001100 1000000110 0001001100 1103010000 0020000010 0001010101 1001001000 0010000001 00 Sisor 1?10001100 0000000000 000111100? 0100000000 1001000000 0000011001 1113011000 0100000010 0001000?01 1000000000 00?0000010 11 Aspredinidae 1?1000?000 00000000 ?0 0011011100 0100000100 0000001000 000000 1?00 1103010000 01?0000010 000?100?01 1010011010 0100011000 00 56 HELDIANA: ZOOLOGY I 3.0 (SwofFord, 1989). Statistics calculated with PAUP Gength = 164, CI = 0.713, RI = 0.800) were slightly different from those calculated with Hennig86, supposedly as a result of the slightly different treatment of missing entries and non- additive (unordered) multistate characters. The tree was rooted with an all-zero hypothetical outgroup, representing a summary of the polarity assump- tions discussed for each character above. The same most parsimonious unrooted network is obtained with or without the all-zero outgroup. The robustness of the final result was tested first by an application of successive approximations weighting (Farris, 1969), implemented by the se- quence of commands ie; xs w; cc; in Hennig86 and repeated iteratively until the assigned character weights stabilized (Fitzhugh, 1989). The applica- tion of successive weighting may result in a dif- ferent tree even with data sets with a single most parsimonious solutions under an equal-weight analysis. In such cases, the procedure serves to reveal a subtle form of internal inconsistency in the data set. In the present case, however, the same result holds after successive approximations weighting. Another test for the hypothesized relationships was the investigation of trees slightly longer than the most parsimonious tree(s) (so-called decay in- dex or Bremer support; see review in Bremer [1994]). That procedure was implemented with PAUP. Eight trees were found with one extra step, and their semistrict consensus revealed a collapse of several components. Those that resisted col- lapse were composed of the taxa named below as Amblycipitidae, Acrochordonichthyini, clade B, Sisoridae, Sisorini, Nangrina, Glyptosteminae, Glyptostemini, Erethistidae, Laguviini, Erethis- tini, and Hara plus Erethistes. With two extra steps, 42 trees were found, and their semistrict consensus was similar to the previous one, except that within Sisoridae, only Nangrina and Glyptostemini re- sisted. With three extra steps, there were 1 82 trees, and only Amblycipitidae, clade B, Nangrina, Glyptostemini, and Laguviini were preserved. Nangrina collapsed at four extra steps (6 1 5 trees), and clade B disappeared at five extra steps (1,857 trees). The results of an analysis of component decay are difficult to interpret. Decay analysis can result in quick collapse of nodes even in data sets with no homoplasy, and a measure of clade ro- bustness based on these procedures is dubious at best. The results obtained here are therefore of- fered for comparison with other studies than as an actual indication of clade support. Under the admittedly nebulous premise that the sequence of clade decay actually reflects some measure of clade robustness, one can make a number of inferences about relative support for groups. For instance, clade A, clade C, and Akysidae are the weakest among the major components, collapsing at a sin- gle extra step. This is particularly important in the case of clade C, which unites the Neotropical As- predinidae with the Asiatic Erethistidae. Although the South American-Asian component is weak at that point, another component also supporting the same biogeographical relationship, clade B, is among the most resilient in the analysis, resisting up to four extra steps. Classificatory Scheme The single most parsimonious cladogram ob- tained in this analysis served as the basis for a new classification of the taxa studied. The main con- cem in translating the cladogram into a classifi- catory scheme was to minimize nomenclatural al- terations of the currently adopted system. For that purpose, a combination of subordinating and se- quencing conventions were adopted (see review by Wiley, 1 98 1 ) so as to maintain names with their presently recognized rank and composition when- ever possible. Nonetheless, a few changes of no- menclature were unavoidable because part of the favored phylogenetic hypothesis is directly incom- patible with previous notions about the compo- sition of some named groups. The single most im- portant change is the placement of Aspredinidae as the sister group to a subset of current Sisoridae. Two viable options were considered in attempting to modify the previous classification so as to in- corporate that information: (1) the current Sisor- idae could be expanded to include the Aspredi- nidae, which could then be lowered to subfamilial rank, and (2) Aspredinidae could be maintained as a family, with Sisoridae split into two different families. I chose the second option, for a variety of reasons. First, there are very few siluriform fam- ilies that are distributed on more than one con- tinent, and among freshwater taxa none are found in both Asia and South America. Thus, maintain- ing the Neotropical Aspredinidae separate from the Asiatic Sisoridae would conform better to the current classification of the order. Second, an ex- panded Sisoridae (i.e., including current Sisoridae plus Aspredinidae) would be a rather difficult group to diagnose for practical purposes because of the large degree of morphological variation observed DE PINNA: ASIAN CATHSH FAMILIES 57 Fio. 53. Familial cladogram based on hypothesized relationships. among its numerous members. A split into two families readily diagnosable by external mor- phology facilitates the making of keys for faunal surveys, manuals, and other identification tools. Accordingly, the hypothesis in Figure 52 is translated into the familial cladogram represented in Figure 53, with Amblycipitidae, Akysidae, Si- soridae (sensu stricto), Erethistidae, and Aspre- dinidae arranged in this order as a sequenced clas- sification. The scheme below also offers a classi- fication down to tribal level for taxa so far included in Sisoridae and Akysidae, not only because these formed the main subject of this study but also because the hypothesis of relationships is resolved enough to allow that. The hierarchies below fa- milial rank are all strictly subordinated. Figure 54 shows the relation between the most parsimonious cladogram in Figure 52 and the proposed familial and subfamilial classification. The family Parakysidae, proposed by Roberts (1989) for Parakysis alone, is not maintained in .^ / .* ^ # / / ■/ / »< .# / i / / /i / / / / / / / / / / // / // // /{ / Erethistinae Fio. 54. Familial and subfamilial classification superimposed on most parsimonious cladogram. 58 FIELDIANA: ZOOLOGY the classification below. Instead, it is lowered to the rank of subfamily and modified to include also Acrochordonichthys and Breitensteinia, the closest relatives of Parakysis. The assemblage composed of the five families is referred to as superfamily Sisoroidea. The first family names based on Sisor and Aspredo are available in the same publication (Bleeker, 1858). Sisoroidea was chosen over Aspredinidae simply because the former is listed first in Bleeker's clas- sification. This is a rather arbitrary decision be- cause the ICZN does not have a provision for line precedence of family-group names. Superfamily Sisoroidea Family Amblycipitidae Family Akysidae Subfamily Akysinae Subfamily Parakysinae Tribe Parakysini Tribe Acrochordonichthyini, new Family Sisoridae Subfamily Sisorinae Tribe Sisorini Subtribe Sisorina Subtribe Nangrina, new Tribe Bagariini Subfamily Glyptosteminae Tribe Glyptostemini Subtribe Glyptostemina Subtribe Pseudecheneidina, new Tribe Glyptothoracini, new Family Erethistidae Subfamily Continae, new Subfamily Erethistinae Tribe Erethistini Tribe Laguviini, new Family Aspredinidae Diagnoses for Clades and Synonymies for Named Taxa This section summarizes the character infor- mation discussed in the previous sections, orga- nized according to the most parsimonious hy- pothesis of relationships. The characters presented as diagnostic for each taxon are the apomorphic states of transitions optimized to occur at the node corresponding to that clade. Therefore, the list of diagnostic characters is technically a synapomor- phy list. Accordingly, the diagnostic characters are not necessarily present in all the terminal taxa in- cluded in their respective clade. Any reversals and convergences of a given character are noted in parentheses. The numbering for diagnostic char- acters follows that in the "Character Description and Discussion" section above. Tracing of char- acter transitions in the most parsimonious hy- pothesis was done with Clados 1.0 (Nixon, 1992). In cases of multiple equally parsimonious optim- izations (i.e., ambiguous characters), the option that maximized reversals relative to convergences was chosen (de Pinna, 1991). The fact that a given character is ambiguous is, nonetheless, noted for each case. Autapomorphies for terminal taxa are not in- cluded below (for a summary of autapomorphies, see Fig. 52). Diagnostic characters of a more prac- tical nature are presented in the identification key in the next section. Synonymies include the first published reference located for a given family-group name that meets current criteria of availability, followed by other selected publications relevant for the taxonomic history of the taxon. A list of genera included is provided for the lowest-level family-group taxon in each clade. These lists include only genera cur- rently considered valid. The diagnoses for Amblycipitidae and Aspre- dinidae should be regarded as provisional because these two families are currently the object of more in-depth studies by other authors. Superfamily Sisoroidea Bleeker, new usage Synonymy— See Sisoridae below. Families Included— Amblycipitidae, Akysi- dae, Sisoridae, Erethistidae, and Aspredinidae. Dl\gnosis 1. Posterior center of ossification of palatine compressed and expanded vertically (reversal in Breitensteinia). 32. Articular region of lateral ethmoid elongated as a process, with articular facet for palatine at tip. 6 1 . Parapophysis of fifth vertebra strong and at- tached to ventral side of centrum, directed directly transversely to centrum. 64. Humeral process or soft tissue around it con- nected to anterior portion of vertebral column by well-defined ligament (humerovertebral ligament). 77. Segments of pectoral fin spine very oblique, almost parallel to axis of spine, not evident DE PINNA: ASIAN CATFISH FAMILIES 59 (reversed in Breitensteinia and clade B; alter- native optimization possible). 87. Dorsal spine with medial ridges along its an- terior surface, forming bilateral longitudinal pouches (reversed in clade B and Breitenstei- nia; alternative optimization possible). 9 1 . Ventral tip of first dorsal fin pterygiophore and corresponding neural spines with con- tacting facets. Family Amblycipitidae Day Amblycepinae Day, 1873 (as subfamily of Siluridae). Amblycepidae; Regan, 1911 (in part). Amblycipitidae; Berg, 1940; Greenwood et al., 1966; Chardon, 1968; Jayaram, 1981; Burgess, 1989; Eschmeyer, 1990; Mo, 1991. Genera VscujUED—Amblyceps Blyth, Lioba- grus Hilgendorf, Xiurenbagrus Chen and Lund- berg. Type GEt^s—Amblyceps Blyth, 1858. Diagnosis 14. Metapterygoid with well-defined thickening for ligamentous connection with ectoptery- goid. 20. Hyomandibula with convex posterior expan- sion. 43. Second hypobranchial with longest axis nearly perpendicular to first. 5 1 . Branchiostegal rays forming a closed, round arrangement. 64. Humeral process or region around it con- nected to anterior portion of vertebral column by well-defined ligament (humerovertebral ligament)— state 1. Unordered character; sta- tus as evidence for monophyly uncertain. 76. First proximal radial elongated as a cartilag- inous rod (convergent with Parakysini). Unnamed Clade A (Akysidae + Sisoridae + Erethistidae + Aspredinidae) Diagnosis 24. Supratemporal fossae present (reversed in Glyptostemini and Pseudolaguvia). 26. Supracleithrum strongly t ;tached to skull. 90. Posterior nuchal plate with anterior process forming facet for articulation with anterior nuchal plate. Family Akysidae Gill Type GENUS— y^Acy^w Bleeker, 1858. Akyses Gill, 1861 (as Group of the subfamily Pime- lodinae). Akyses; Bleeker, 1862 (as phalanx 9 of the Slirps Bag- rini). Akysidae; Weber and De Beaufort, 1913; Berg, 1940; Greenwood et al., 1966; Chardon, 1968; Jayaram, 1981; Burgess, 1989; Eschmeyer, 1990; Mo, 1991. Parakysidae Roberts, 1989; Eschmeyer, 1990. Amblycepidae (in part); Regan, 1911. Subfamilies Included- nae. Diagnosis ■Akysinae, Parakysi- 1 3. Metapterygoid with long anterior process (re- versed in Parakysini; alternative optimization possible). 45. Anterior cartilage of anteriormost basibran- chial enlarged (reversed in Parakysini; alter- native optimization possible). 64. Humeral process or region around it con- nected to anterior portion of vertebral column by well-defined ligament (humerovertebral ligament)— state 2 (reversed to 1 in Parakys- ini). Unordered character; status as evidence for monophyly uncertain. 93. First lateral-line ossicle enlarged, overlapping with posterior portion of supracleithrum (convergent in Aspredinidae). 98. Path of infraorbital laterosensory canal straight (reversal in Parakysini; alternative optimi- zation possible). 99. Lacrimal with dorsal shelf-like portion di- rected mesially (convergent in Aspredinidae). Subfamily Akysinae Gill Type Genus and Synonymy— See Akysidae. Genus Included— /I A:>'5/5 Bleeker. Dl^gnosis 83. Anterior margin of pectoral fin spine with notch visible in dorsal view. 95. Nasal with expansions beyond canal-bearing portion. Subfamily Parakysinae Roberts Type GEtnjs— Parakysis Herre, 1940. Parakysidae; Roberts, 1989; Eschmeyer, 1990. 60 FIELDL\NA: ZOOLOGY Tribes Included— Parakysini, Acrochordoni- chthyini. Diagnosis 2. Mandibular canal entering bone through its midsection (coded as missing in Aspredini- dae, Sisor, glyptostemoids, and Contd). 38. Second hypobranchial unossified (convergent in clade C). 41. Extra cartilaginous element anterior to last basibranchial (convergent in Sisoridae and Laguviini). 55. Posterior ceratohyal with dorsal process near its connection with suspensorium. 72. Vertebrae with transverse processes (conver- gent in Aspredinidae and Sisorina). Tribe Parakysini Roberts Type Genus and Synonymy— See Parakysinae. Genus IfiCUJTyED—Parakysis Herre. Diagnosis 9. Posterior half of palatine twisted laterally (coded as missing in Akysinae). 15. Quadrate with anterior spur-like process (convergent in Amblyceps). 17. Metapterygoid reduced to its center of os- sification. 18. Very large subpreopercle. 39. First hypobranchial unossified. 40. All basibranchials unossified (coded as miss- ing in Glyptostemina). 73. Reduced number of pleural ribs, state 2 (con- vergent with Erethistidae and Erethistinae; coded as missing in Aspredinidae). 76. First proximal radial elongated as a cartilag- inous rod (convergent with Amblycipitidae). 81. Humeral process absent (convergent with Glyptostemini). 98. Reversal to state 0 (alternative optimization possible). 108. Epural curved. Tribe Acrochordonichthyini, new Type GEt^us—Acrochordonichthys Sleeker, 1858. Genera IhiCLUDED—Acrochordonichthys Bleek- er, Breitensteinia Steindachner. Diagnosis 25. Paired fontanels between supraoccipital, fron- tals, and sphenotics (convergent in Sisorini). 33. Mesethmoid comua with dorsal subdivision. 53. Proximal end of anterior ceratohyal with nar- row anterior extension directed anterolater- ally. Unnamed Clade B (Sisoridae + Erethistidae + Aspredinidae) Diagnosis 27. Posterior portion of supracleithrum anky- losed to margin of Weberian lamina (state 1; reversed to 0 in Glyptostemini; further modified as state 2 in Erethistinae). 57. Parapophysis of fifth vertebra strongly flat- tened and expanded (reversed in Glyptoster- nini). 62. Parapophysis of fifth vertebra long, almost or quite reaching lateral surface of body wall. 64. Humeral process or region around it con- nected to anterior portion of vertebral col- umn by well-defined ligament (humerover- tebral ligament)— state 3 (reversed to 0 in Glyptostemina). Unordered character; sta- tus as evidence for monophyly uncertain. 66. Posterior part of Weberian lamina exten- sively contacting parapophysis of fifth ver- tebra. 77. [Reversal of] anterior half of segments of pectoral-fin spine elongate, almost parallel to axis of spine (alternative optimization possible). 79. Coracoid with ventral anterior process (re- versed in Glyptostemina and Gagata; coded as missing in Nangra). 87. [Reversal of] second dorsal-fin spine with medial ridge along its anterior surface, form- ing bilateral longitudinal pouches (alterna- tive optimization possible). 88. Ventral arms of first dorsal-fin spine with posterior subprocesses attached dorsal to their tip (coded as missing in Sisorina, all Glyptostemini, and Aspredinidae). 103. Basipterygium with ventral longitudinal keel, anteriorly extending alongside intemal arm (reversed in Aspredinidae; coded as missing in Sisorina and Glyptostemina; ambiguous optimization possible). Family Sisoridae Bleeker TYPEGENUS—5/5or Hamilton, 1822. Sisorichlhyoidei Bleeker, 1858 (as Subfamilia I of Familia Siluroidei). DE PINNA: ASIAN CATFISH FAMILIES 61 Sisoriformes; Bleeker, 1862, 1863 (as Subfamilia I of Familia Siluroidei). Sisoridae; Gill, 1 872; Regan, 1911; Berg, 1 940; Green- wood et al.,. 1966; Chardon, 1968; Misra, 1976; Jayaram, 1981; Burgess, 1989; Eschmeyer, 1990. Bagariidae; Weber and De Beaufort, 1913; Berg, 1 940; Greenwood et al., 1 966. Sisolidae (misspelling); Kimura, 1934. Incertae sedis— genus Sundagagata Boese- man, 1966. Subfamilies Included— Sisorinae, Glyptoster- ninae. Dl\gnosis 30. Lateral ethmoid with narrow lateral exten- sions directed posteriorly alongside lateral margin of frontals (coded as missing in all Sisorini). 41. Extra cartilaginous element anterior to last basibranchial (reversed in Nangrina and Glyptostemina; convergent in Parakysinae and Laguviini). 60. Distal tip of os suspensorum closely linked and/or adpressed to ventral process of com- plex centrum. 67. Posterior portion of Weberian lamina with lateral extension alongside anterior margin of parapophysis of fifth vertebra. Subfamily Sisorinae Bleeker Type Genus and Synonymy— See Sisoridae. Tribes included— Sisorini, Bagariini. Dl\gnosis 4. Distal tip of maxilla pointed (reversed in Si- sor; (alternative optimization possible). 29. Epioccipital with slender posterior process (reversed in Sisor; alternative optimization possible). 63. Posterodorsal comer of Weberian lamina with pointed process directed posteriorly. Tribe Sisorini Bleeker Type Genus and Synonymy— See Sisoridae. SuBTRiBES included— Sisorina, Nangrina. Dl\gnosis 7, Premaxillary dentition reduced or absent in adults (reversed in Nangra; coded as missing in Aspredinidae; alternative optimization possible). 8. Premaxillae with well-defined facet for ar- ticulation with mesethmoid. 25. Paired fontanels between supraoccipital, frontals, and sphenotics (convergent in Ac- rochordonichthyini). 44. Last basibranchial narrow (reversed in Nan- gra; alternative optimization possible). 56. Weberian capsule complete. 109. Stridulatory apparatus (state 2; state 1 con- vergent with Continae). Subtribe Sisorina Bleeker Genera Included— Swor Hamilton. Dl\gnosis 3 . Mandibular laterosensory canal absent (con- vergent in Glyptostemina and clade C). 72. Vertebrae with transverse processes (con- vergent in Aspredinidae and Parakysinae). 84. Anterior margin of pectoral spine with ser- rations (convergent in Erethistidae and As- predinidae). 111. Bony plates on surface of body. 112. First (uppermost) principal caudal fin ray en- larged and elongated as a filament. Subtribe Nangrina, new Type Genus— A^an^ra Day, 1876. Genera Included— (7a^a/a Bleeker, Nangra Day. Dl\gnosis 1 2. Direction of main axis of maxilla, when part- ly abducted, with strong ventral component. 34. Lateral ethmoid with well-defined dorsal facet contacting ventral surface of frontal. 35. Lateral ethmoid cartilage for palatine with extension for lacrimal. 4 1 . [Reversal of] extra cartilaginous element an- terior to last basibranchial. 92. Posterior nuchal plate straight and thin, its width approximately even from base to tip (coded as missing in Bagariini). 101. Dorsal half of suprapreopercle strongly curved posteriorly, making nearly 90° angle with ventral half (coded as missing in Glyp- tothoracini). Tribe Bagariini Giinther Type GENvs—Bagariiis Bleeker, 1853. Bagarina Giinther, 1864 (as the ninth group of sub- family Siluridae Proteropterae). Bagarinae; Misra, 1976; Burgess, 1989. Bagariidae (in part); Weber and De Beaufort, 1913; Berg, 1940; Greenwood et al., 1966. 62 FIELDLMSTA: ZOOLOGY Genus Included— Bagarius Bleeker. Diagnosis 1 0. Distal portion of premaxilla detached (con- vergent in Glyptothoracini; coded as missing in Glyptostemina). 21. Endopterygoid with spur-like process an- terolaterally. 46. Fourth ceratobranchial with tooth plates. 50. Proximal portions of epibranchials 1 and 2 adpressed to one another. 96. Nasal elongated anteriorly to level of ante- rior margin of mesethmoid comua (conver- gent in Nangra and Aspredinidae). 105. Posterior cartilage of basipterygium rami- fied. Subfamily Glyptosterninae Gill Type Genus— Glyptosternon McClelland, 1 842. Glyptostemi Gill, 1 86 1 (as Group of the sub- family Pimelodinae). Glyptosterninae; Misra, 1976; Burgess, 1989. Tribes Included— Glyptostemini, Glyptothor- acini. Diagnosis 49. Proximal portion of first and/or second epi- branchial curved posteriorly (convergent in Pseudolaguvia; coded as missing in Lagu- via). 52. Lateral wings of urohyal extremely reduced or absent. 1 10. Integument of thorax or thorax and abdo- men modified into adhesive apparatus formed by patches of unculi underlain by ridges of integument (reversed to state 0 in Glyptostemina; further modified as state 2 in Pseudecheneidina; convergent in Ereth- istidae). Comments— A suprageneric name based on Glyptosternum was first proposed by Gill (1861) as Group Glyptostemi of the subfamily Pimelo- dinae. Although the family-group name has been formally available for over 1 30 years, it has almost never been used in Latin (exceptions: Misra, 1 976; Burgess, 1989). Curiously, the collective name glyptostemoid sisorids, or simply glyptostemoids, has been widely cited in vernacular English to refer to the suprageneric group composed of Glypto- sternum and its closest relatives (e.g., Hora &. Menon, 1948; Hora & Silas, 1952a,b;Gauba, 1969; Steinitz, 1961; TUak, 1963, 1976; Chu, 1979, 1981). Most uses of the vernacular name glypto- stemoids are equivalent in composition to the tax- on ranked below as sublribe Glyptostemina. Tribe Glyptostemini Gill Type Genus and Synonymy— See Glyptoster- ninae. Subtribes Included— Glyptostemina, Pseu- decheneidina. Dl\gnosis 1 1. Presence of incisiform teeth on jaws. 24. [Reversal of] supratemporal fossae. 27. [Reversal of] posterior portion of supracleith- mm ankylosed to margin of Weberian lamina. 57. [Reversed of] parapophysis of fifth vertebra strongly flattened and expanded. 59. Complex centmm as short as or shorter than free centra. 80. Eleven or more pectoral fin rays. 81. Humeral process absent. 82. Upper limb of cleithmm with laminar expan- sion directed mesially. 89. First dorsal fin ray flexible, not spinous. 94. Lacrimal modified into spoon-shaped stmc- ture located ventrolateral to nasal capsule (convergent in Erethistidae). Subtribe Glyptostemina Gill Type Genus and Synonymy— See Glyptoster- ninae. Genera Included— Glyptosternon McClelland, Glaridoglanis Norman, Oreoglanis Smith, Exos- toma Blyth, Myersglanis Hora and Silas, Corag- lanis Hora and Silas, Euchiloglanis Regan, Pseu- dexostoma Chu. Dl\gnosis 3. Mandibular laterosensory canal absent (con- vergent in Sisorina and clade C). 5. Maxilla dorsoventrally flat and strongly ex- panded distally. 6. Interopercle narrow and elongate, its lami- nar portion reduced. 4 1 . [Reversal of] extra cartilaginous element an- terior to last basibranchial (also homoplas- tically reversed in Nangrina). 64. [Reversal of] humeral region connected with anterior portion of vertebral column by well- defined ligament (reversal to state 0). 78. Proximal pectoral radials 2 and 3 flared at distal tip. 79. [Reversal of] coracoid with ventral process. DE PINNA: ASL\N CATFISH FAMILIES 63 104. External arm of basipterygium abruptly ex- panded distally. 1 10. [Reversal of] integument of thorax or thorax and abdomen modified into adhesive ap- paratus formed by patches of unculi under- lain by ridges of integument (reversed to state 0). Subtribe Pseudecheneidina, new Type Genus— Pseudecheneis Blyth, 1860. Genus Included— Only the type genus. Diagnosis 16. Endopterygoid surrounding metapterygoid dorsally. 36. Sphenotic not contacting supraoccipital. 42. First epibranchial expanded anteriorly. 70. Dorsal expansion on first pleural rib. 110. Integument of thorax or thorax and abdo- men modified into adhesive apparatus formed by patches of unculi underlain by ridges of integument (state 2). Tribe Glyptothoracini, new Type Genus— G/>'p/o//2or(2x Blyth, 1860. Genus Included— Only the type genus (with three subgenera: Glyptothorax, Paraglyptothorax Li, and Superglyptothorax Li; Li, 1986). Dl^gnosis 10. Distal portion of premaxilla detached (con- vergent in Bagariini; coded as missing in Glyptostemina). 3 1 . Lateral arms of vomer long and thin, extend- ing underneath whole length of articular pro- cess of lateral ethmoid. Unnamed Clade C (Erethistidae + Aspredinidae) Dl\gnosis 3. Mandibular laterosensory canal absent (re- versed in Erethistinae; convergent in Sisorina and Glyptostemina; alternative optimization possible). 38. Second hypobranchial unossified (convergent in Parakysinae). 84. Anterior margin of pectoral spine with ser- rations (convergent in Sisor; coded as missing in Aspredinidae). 85. Internal support for pectoral fin rays small in size (reversed in Laguviini). 97. Anterior portion of lateral line running closely in parallel to lateral margin of Weberian lam- ina. Family Erethistidae Bleeker Type Genvs— Erethistes Miiller and Troschel, 1849. Erethistides Bleeker, 1862, 1863 (as phalanx 10 of Stirps Bagrini). Subfamilies Included— Continae, Erethistin- ae. Dl\gnosis 37. Last hypobranchial with anterolateral exten- sion. 54. Posterior margin of anterior ceratohyal with laminar expansion forming process directed laterally. 73. Five or fewer pleural ribs (convergent with Parakysini; coded as missing in Aspredini- dae). 86. Extensive contact between expansions of first and second pterygiophores of dorsal-fin spines. 94. Lacrimal modified into spoon-shaped struc- ture located ventrolaterally to nasal capsule (convergent in Glyptostemini). 1 1 0. Integument of thorax or thorax and abdo- men modified into adhesive apparatus formed by patches of unculi underlain by ridges of integument (reversed in Erethistini; convergent in Glyptosteminae; alternative optimization possible). Subfamily Continae, new Type Genus— Con/a Hora, 1950. Genus Included— Only the type genus. Dl\gnosis 74. First two pleural ribs flattened and expand- ed. 109. Stridulatory apparatus (state 1 convergent with Sisorini). Subfamily Erethistinae Bleeker Type Genus and Synonymy— See Erethisti- idae. 64 FIELDIANA: ZOOLOGY Tribes Included— Erethistini, Laguviini. Diagnosis 27. Posterior portion of supracleithrum anky- losed to margin of Weberian lamina (as state 2). 58. Parapophysis of fifth vertebra branched or notched distally (coded as missing in Aspre- dinidae). 73. Four or fewer pleural ribs— state 2 (conver- gent with Parakysini; coded as missing in As- predinidae). Tribe Erethistini Bleeker Type Genus and Synonymy— See Erethisti- idae. Genera li^cujDED—Erethistes Miiller and Troschel, Erethistoides Hora, Hara Blyth. Diagnosis 100. Nasal adpressed to margin of neurocranium for two thirds or more of its posterior length, then abruptly diverging laterally at anterior portion. Tribe Laguviini, new Type Genus— La^v/a Hora, 1921. Genera Included— Lagwv/a Hora, Psendola- guvia Misra. Dl\gnosis 1 9. Opercle very narrow (coded as missing in As- predinidae). Endopterygoid bifurcated anteriorly. Extra cartilaginous element anterior to last basibranchial (convergent in Sisoridae and Parakysinae). [Reversal of] internal support for pectoral-fin rays small in size. 22. 41. 85 Family Aspredinidae Bleeker Type GENVS—Aspredo ScopoM, Mil. Aspredinoidei Bleeker, 1858 (as Familia II), 1862, 1863 (as Familia IV). Aspredinina; Giinther, 1864 (as 14thgroup, sixth sub- family—Siluridae Proteropodes). Aspredinidae; Gill, 1872; Boulenger, 1 904; Goodrich, 1909; Chardon, 1968; Nelson, 1994. Bunocephalidae; Regan, 1911. Genera Included— ^mara//a Fowler, Xyli- phius Eigenmann, Dysichthys Cope, Bunocephalus Kner, Agmus Eigenmann, Hoplomyzon Myers, Dupouyichthys Schultz, Ernstichthys Femandez- Yepez. Diagnosis 23. Condyle for opercle on hyomandibula di- rected ventrally. 28. Pterotic with lateral shelf-like expansion. 47. Lateral cartilage of first hypobranchial en- larged. 72. Vertebrae with transverse process (conver- gent in Sisorina and Parakysinae). 93. First lateral-line ossicle enlarged, overlap- ping with posterior portion of supracleith- rum (convergent in Akysidae). 96. Nasal elongated anteriorly to level of ante- rior margin of mesethmoid comua (conver- gent in Nangra and Bagariini). 99. Lacrimal with dorsal shelf-like portion di- rected mesially (convergent in Akysidae). 102. Hyomandibula with bony tunnel enclosing branch of the laterosensory canal between preopercle and pterotic. 103. [Reversal of] basipterygium with ventral longitudinal keel, extending anteriorly alongside internal arm (coded as missing in Sisorina and Glyptostemina; alternative op- timization possible). 106. Cartilage on posterior process of basiptery- gium reduced or absent. 107. Anterior arms of basipterygium undifferen- tiated from main part of bone. Identification Key to Families Through Subtribes This key has been elaborated as an aid to the implementation of the proposed changes in clas- sification. The key is eminently practical and not intended to follow the branching sequence of the phylogenetic hypothesis advanced. Likewise, the characters used are not necessarily synapomor- phies for the groups but rather are easily observed traits, of external anatomy whenever possible, that allow identification of taxa without the need for dissections or other more sophisticated prepara- tions. Because the key is artificial, the user must keep in mind that certain taxa may have several different entries, and the key must be followed from the beginning each time. DE PINNA: ASIAN CATFISH FAMILIES 65 la. Adhesive apparatus, formed by integumentary ridges and grooves, present on thorax (between pectoral fins or immediately posterior to branchiostegal membranes) or abdomen 2 lb. Adhesive apparatus absent, skin of thorax and abdomen smooth 5 2a. Adhesive apparatus long and narrow, approximately 6 times or more as long as broad Erethistidae, Continae 2b. Adhesive apparatus shorter and broader than in 2a 3 3a. Grooves of adhesive apparatus transverse to longitudinal axis of body Sisoridae, Glyptosteminae, Glyptostemini, Pseudecheneidina 3b. Grooves of adhesive apparatus parallel or oblique to longitudinal axis of body 4 4a. Presence of a narrow bony area immediately posterior to and approximately as long as the humeral process, appearing to be a continuation of that process (not always visible on the surface of skin but evident to the touch Erethistidae, Erethistinae, Laguviini 4b. Area immediately posterior to tip of humeral process soft or, if bone is present, its length is far shorter than that of humeral process Sisoridae, Glyptosteminae, Glyptothoracini 5a. Humeral process poorly developed or absent, not visible through skin 6 5b. Humeral process well developed, visible through skin 7 6a. Pectoral fin rays 1 1 or more Sisoridae, Glyptosteminae, Glyptostemini, Glyptostemina 6b. Pectoral fin rays 9 or fewer Amblycipitidae 7a. Serrations absent on anterior margin of pectoral-fin spine 8 7b. Serrations present on both anterior and posterior margins of pectoral fin spine 9 8a. Prominent, semispherical, round skin tubercles over body 11 8b. Skin tubercles present either absent or, when present, flattened, ovoid in shape, and not prominent 13 9a. Broad bony plates present along mid-dorsal line posterior to dorsal fin Sisoridae, Sisorinae, Sisorini, Sisorina 9b. No bony plates as in 9a 10 1 Oa. Adipose fin absent Aspredinidae 10b. Adipose fin present Erethistidae, Erethistinae, Erethistini I la. Tip of pectoral fin, when adpressed to body, almost or quite reaching vertical through origin of pelvic fins Akysidae, Akysinae I I b. Tip of pectoral fin, when adpressed to body, not reaching vertical through origin of pelvic fins . 12 1 2a. Skin tubercles arranged in longitudinal rows along body Akysidae, Parakysinae, Acrochordonichthyini 1 2b. Skin tubercles not arranged in longitudinal rows Akysidae, Parakysinae, Parakysini 1 3a. Flattened skin tubercles present Sisoridae, Sisorinae, Bagariini 1 3b. Skin tubercles absent Sisoridae, Sisorinae, Sisorini, Nangrina Discussion and Comparison with Previous Qassifications The results of the phylogenetic hypothesis ad- vanced herein help to settle a number of contro- versial issues about the systematics and taxonomy of the Sisoridae and their relatives. The genus La- guvia was created by Hora (1921) to include three species, L. aspera (the type species), L. sham, and L. ribeiroi, the two latter described as new in the same paper. According to Hora (1921), Laguvia was in most resp)ects intermediate between Ereth- istes and Glyptothorax. Later, the genus was placed as a synonym o{ Glyptothorax by Menon (1954), a move subsequently accepted without comment by Li (1986). Menon's reason for his action was that the characters given by Hora to distinguish Laguvia from Glyptothorax were minor and were actually present in some species of the latter. Men- on particularly noted the similarities between La- guvia and Glyptothorax tuberculatus, a species lat- er made the type of a separate genus, Pseudola- guvia, by Misra (1976). The genus Laguvia was resurrected by Gauba (1970b), who studied the internal anatomy of the genus and recorded a num- ber of similarities between Laguvia and Glypto- thorax and between Laguvia and Erethistes. Com- paring the evidence for the alternatives, Gauba decided that there was more support for grouping 66 FIELDIANA: ZCX)LOGY Laguvia and Erethistes and further suggested that the synonymy oi Laguvia under Glyptothorax pro- posed by Menon (1954) was incorrect. Gauba (1967, 1970b) informally recognized a natural group similar to the family Erethistidae as defined in the present study and on occasion even used the vernacular "erethistid" (e.g., Gauba, 1970b, p. 63). Apparently unaware of Gauba's previous work, Jayaram (1973) also proposed to revalidate Laguvia and further suggested that the genus was most closely related to Hara. Again without ref- erence to Gauba, Tilak ( 1 976) also informally rec- ognized a relationship between Laguvia and Er- ethistes-WVt taxa when he said that "the fishes of the genus Laguvia tend to show more affinities to the Erethistes group of sisorid fishes" (Tilak, 1 976, p. 259). The suggestions of Gauba, Jayaram, and Tilak, although not formulated within a strictly comparative phylogenetic framework, are corrob- orated by the results of this study. Numerous apo- morphic characters support the position of La- guvia (and the closely related Pseudolaguvia) in an intemested subset of Erethistidae. Menon's (1954) previous suggestion of synonymy evidently had been based on a sense of overall similarity between Laguvia and Glyptothorax, which turns out to con- sist only of symplesiomorphies. Gunther (1864) and Bleeker (1873) included Pimelodus asperus (later the type species of Laguvia) in Hara, a place- ment that is close to the actual phylogenetic affin- ities of the genus. A controversy about the distinctiveness of Ga- gata and Nangra (cf Hora & Law, 1941; Tilak, 1963) is also affected by the hypothesis presented in this paper. These two genera are hypothesized as sister groups on the basis of a large set of syn- apomorphies. Although only a few species of each genus were examined, preliminary evidence sug- gests that each forms a monophyletic group. In this situation, the question of their generic sepa- ration becomes rather irrelevant because both al- ternatives are compatible with the proposed phy- logenetic relationships. I agree with Tilak (1963) that the morphological distinctions between Nan- gra and Gagata are at least as pronounced as those among most other genera in Sisoroidea. Hora and Law's suggestion of synonymy was not based on a detailed knowledge of the internal anatomy of these fishes and relied mostly on their smaller ex- ternal morphology. The phylogenetic affinities of the genus Pseu- decheneis have also been controversial. Early stud- ies pointed out similarities and differences be- tween that genus and various "glyptostemoid si- sorids" but were not conclusive about their pos- sible phylogenetic affinities (Hora & Chabanaud, 1930; Hora, 1952; Hora & Silas, 1952a,b). Various speculations about convergent morphology of the adhesive apparatus adaptations led to the recog- nition of Parapseudecheneis (Hora & Chabanaud, 1930) and Propseudecheneis (Hora, 1937). Hora and Chabanaud (1930) proposed that Pseudeche- neis and Parapseudecheneis were derived from dif- ferent lineages within glyptostemoids. Essentially the same view was reiterated by Hora (1952) but with the addition of one more allegedly convergent lineage represented by Propseudecheneis. In any case, the origin of Pseudecheneis, Parapseudeche- neis, and Propseudecheneis was considered to be within the glyptostemoid lineage, therefore im- plicitly indicating that the whole assemblage (glyp- tostemoids plus the three genera) formed a mono- phyletic group. Both Parapseudecheneis and Prop- seudecheneis were later considered junior syn- onyms of Pseudecheneis by Chu (1982), who proposed that the three genera formed a mono- phyletic group, thus refuting Hora's convergence hypothesis. Of the three putative synapomorphies proposed by Chu (1982, p. 435) for the expanded Pseudecheneis, two can be considered valid evi- dence for monophyly: the stmcture of the adhesive organ and the lack of contact between the sphen- otic and supraoccipital (his characters nos. 1 1 and 13, respectively, the latter originally proposed by Gauba [1968, p. 233]; a third character, the ab- sence of a dorsal fin spine [Chu's no. 1 2], is actually a synapomorphy for glyptostemoids plus Pseu- decheneis). Although these two characters are re- liable evidence for monophyly, the internal anat- omy of the most distinctive species, P. paviei, re- mains unknown, therefore leaving the stmcture of the adhesive organ as the only derived character observed in all species of Pseudecheneis. Regard- less of these limitations, the monophyly of Pseu- decheneis is a reasonable working hypothesis, al- though one that needs further testing. Chu (1982) focused exclusively on the problem of the mono- phyly of Pseudecheneis and did not address the question of that genus's closest relatives. Tilak (1976, p. 260) proposed a close relationship be- tween Pseudecheneis and glyptostemoids, mainly on the basis of an overall similarity in their pec- toral spine morphology. In support of Tilak's view, a large set of synapomorphies advanced in the present study support the alignment of Pseude- cheneis with the glyptostemoid group of sisorids. These characters plus the numerous synapomor- phies exclusive to glyptostemoids leave little doubt DE PINNA: ASIAN CATFISH FAMILIES 67 that Pseudecheneis is the sister group of glypto- stemoids. In certain ways, Pseudecheneis is less specialized for torrential life than are glyptoster- noids, as has been repeatedly observed (e.g., Hora and Silas, 1952a). From the phylogeny advanced herein, it is clear that the fewer specializations of Pseudecheneis are a result of the basal position of that genus rather than a reversal from a more ex- treme condition. Thus, the adaptations to torren- tial environments observed in Pseudecheneis are structurally and phylogenetically intermediate be- tween those of Glyptothorax and glyptostemoids. A familial separation between the Akysidae and the Amblycipitidae has been recognized since We- ber and de Beaufort (1913). Prior to that, fishes now included in Akysidae were part of a broader Amblycipitidae (e.g., Regan, 1911). In delimiting his Amblycepidae [sic], Regan (1911) considered similarities in the edentulous palate and anterior vertebrae between Akysis and Acrochordonichthys and between Amblyceps and Liobagrus. None of these resemblances, however, can be considered putative synapomorphies by today's standards. Hora ( 1 936) discussed this problem and concluded that the morphological differences justified rec- ognition of a separate Akysidae. Although Hora's reasoning relied on degree of difference rather than genealogical kinship, his conclusions are essen- tially correct in a phylogenetic context. The most parsimonious hypothesis favored in this work in- dicates that members of the Akysidae and Am- blycipitidae are not each other's closest relatives. Recognition of an expanded Amblycipitidae in Regan's sense would result in a paraphyletic as- semblage, thus requiring a separate Akysidae, as adopted in most classifications of this century. Some of the results above will likely be shown to be incomplete under a more fine-grained anal- ysis because of incomplete representation of spe- cies in the studied taxa. In particular, the genera Akysis and Glyptothorax may be deemed para- or polyphyletic in more encompassing studies. The genus Akysis comprises approximately 10 de- scribed species (He & Chen, 1981), but only four were available for examination here. Two char- acters (8 1 and 9 1 ) were found to corroborate that these four species compose a monophyletic group. Some species currently in Akysis may be more closely related to species in other akysid genera. For example, character 9 is shared between Par- akysis and only some Akysis species. Such syna- pomorphies should serve as indicating that the monophyly of Akysis should not be taken for granted in more complete studies on the genus. Species in Akysis probably compose an assemblage far more complex and diverse than so far realized, a complexity that has not yet been recognized be- cause of the rarity of representatives of the genus in collections. The genus Glyptothorax also is probably com- plex, and the hypothesis proposed herein for the position of the taxon may be quite incomplete. One uniquely derived character (no. 31) for the genus was observed in all representatives available for examination or for which published data were available. However, the condition of this character in most of the approximately 40 species of Glyp- tothorax remains to be determined. A second char- acter transition is also hypothesized to be apo- morphic for Glyptothorax (character 10), but this character is homoplastic with other clades. If the genus is actually shown to be nonmonophyletic, it is unlikely that its species will be spread through- out widely diverging levels of sisorid phylogeny. Most probably they will be restricted to the Glyp- tosteminae, some of them perhaps as sister group to the whole subfamily or to the tribe Glyptoster- nini. One interesting feature too poorly known to be included in the character analysis but which deserves discussion is a peculiar foramen found in the pterosphenoid. The lateral portion of the braincase in Pseudecheneis is remarkably com- plex, with a number of foramina and subdivisions not observed in other siluriforms. Of particular relevance is the foramen formed by a roughly Y-shaped postero ventral process of the pterosphe- noid. The arms of the process are widely divergent at the base and then curve strongly toward each other, forming a well-defined, rounded space in the center. The foramen thus formed is actually located in the area corresponding the prootic but is enclosed by ossifications derived from the pter- osphenoid. Such a configuration is certainly de- rived for siluriforms and is seen also in remaining members of the Glyptostemini, although in a somewhat modified form. A very similar situation is found in some of the species examined of Glyp- tothorax, such as G. major, but not in others, such as G. sinense. This may indicate that some species of Glyptothorax aTe more closely related to species of other genera in the Glyptostemini than to other species currently in Glyptothorax. Evidently there is some character incongruence involved in the situation because the species of Glyptothorax in question share a derived character not seen else- where in the Glyptostemini. Solution of this prob- lem will have to await a more complex study of Glyptothorax, a genus in urgent need of a species- 68 FIELDIANA: ZOOLOGY level revision and smaller-scale phylogenetic treat- ment. There is also some indication that the genera Gagata and Nangra may be nonmonophyletic as presently constituted. A striking synapomorphy was found for two species, one from each genus. The posterior portion of the palatine in Nangra nangra and Gagata cenia is expanded dorsoven- trally to an extreme degree (Fig. 55). This expan- sion goes far beyond the normal synapomorphic condition of that bone in sisoroids (cf character 1 ) and is evidently a character exclusive to those two species among the taxa examined. This ex- treme condition of the palatine expansion is not present in the remaining species examined of both Gagata and Nangra and thus may indicate that neither genus is monophyletic. Again, both genera include a number of species not available for ex- amination in the present study, but present evi- dence hints at a rather complex systematic situa- tion involving Gagata and Nangra. Two patterns of homoplasy are noteworthy in this study. The first is the set of synapomorphies for aspredinids and akysids. The most conspicu- ous of these is the presence of large vertebral trans- verse processes, which sometimes extend all the way to the body surface (character 72). This char- acter is also homoplastically present in Sisor. In that taxon, however, the processes are associated with the bony plates derived from lateral-line os- sicles and clearly provide partial mechanical sup- port for them. This situation does not occur in akysids and aspredinids, where there are no bony plates on the surface of the body. This structural- functional difference, when interpreted in light of the phylogenetic hypothesis, provides a posteriori evidence that the transverse processes in Sisor are not homologous to those in other sisoroid taxa. The structure of the transverse processes in as- predinids and akysids, however, is very similar and is only revealed as homoplastic by the topol- ogy of the most parsimonious cladogram. Addi- tional synapomorphies for aspredinids and akys- ids are the enlarged first lateral-line ossicle, which is also adjoined to the posterior portion of the supracleithrum (character 93), and the morphol- ogy of the lacrimal, which has a dorsal shelf-like portion directed mesially (character 99). Again, no structural detail seems to indicate, even a poster- iori, that the conditions of these traits are non- homologous in akysids and aspredinids. Their ho- moplastic condition is a conclusion stemming only from parsimony considerations. Another noteworthy pattern of homoplasy is ob- B Fig. 55. Nangra nangra, cmk 6369, right palatine. A, lateral view. B, ventral view. Larger stippling repre- sents cartilage. Scale bar = 1 mm. served between Parakysis and amblycipitids. The morphology of the first proximal pectoral radial in Parakysis and amblycipitids is very similar (character 76), having the shape of an elongate cartilaginous rod unique among sisoroids and very distinctive from the plesiomorphic condition in most other siluriforms. Fitting the same pattern, the presence of a spur-like process on the quadrate (character 1 5) occurs exclusively in Parakysis and Amblyceps (but not in Liobagrus). The state of the humerovertebral ligament (character 64) should also be mentioned in this regard. Parakysis and amblycipitids are the only taxa in which the lig- ament is not attached to the humeral process. Al- though the ordering of the three states of the hu- merovertebral ligament is as yet undetermined, as is the correct coding for the character in Parakysis, the distribution of the various states in the most parsimonious scheme leaves little doubt that any derived similarities found in that complex for Par- akysis and amblycipitids are homoplastic because Parakysis is adjacent to two clades showing state 2 for character 64 {Akysis and Breitensteinia + Acrochordonichthys). An additional similarity between Parakysis and amblycipitids that was not included in the data analysis is the shape of the endopterygoid. In both taxa, the bone has the shape of an inverted L, with its anterior portion strongly deflected laterally, forming roughly a 90° angle with the posterior por- tion of the bone. The endopterygoid is much smaller in Parakysis than in the amblycipitids, but DE PINNA: ASIAN CATFISH FAMILIES 69 the similarity in general shape is undeniable and is certainly derived within siluriforms. The prob- lem with this similarity is the state assignment to other terminal taxa in the Sisoroidea. The endop- terygoid is present in all of them, and in several cases it has a shape similar to that in Parakysis and amblycipitids, although with a broader ossi- fied area. In some cases, the L shape of the en- dopterygoid is simply obscured by additional os- sification. A comparison of various endoptery- goids shows that there are conditions in which the L shape is clearly present, others in which it is identifiable but poorly defined, and a few in which it is not present by any stretch of the imagination. Examination of early ontogenetic stages in mem- bers of the ingroup may reveal that the L-shaped endopterygoid is actually present in a large subset and that it is subsequently obscured by additional bone expansions in adult individuals of most in- group taxa. Such ontogenetic data are presently unavailable, but without them we cannot ade- quately understand and code this character. With present information, the addition of this character to the data matrix would result in an unacceptably large number of taxa with the character coded as missing, which in turn would obscure more in- formation than it would reveal. For these reasons, the similarity in endopterygoid shape between Parakysis and amblycipitids is considered a char- acter impossible to adequately delimit with cur- rent knowledge and accordingly was left out of the parsimony analysis. Comments on Miniaturization and Gigantism in Sisoroidea The evolutionary and ecological significance of miniaturization in vertebrates has been discussed in a number of recent publications. For freshwater fishes, Weitzman and Vari (1988) and Kottelat and Vidthayanon (1993) provided lists of minia- turized species known from South America and South or Southeast Asia, respectively. Both pub- lications adopted an arbitrary maximum of 26 mm standard length (SL) for a species to be considered as a miniature. Weitzman and Vari (1988), how- ever, recognized that this cutoff" is simply a first approximation and that numerous other traits characterize miniaturized species, some of which may exceed that length. Sisoroidea includes the only truly miniaturized catfish reported in South Asia (the erethistid Erethistes maesotensis Kot- telat, according to Kottelat & Vidthayanon, 1993). Other Asiatic taxa that are obviously miniatures, even though they may on occasion slightly exceed 26 mm SL, are Harajerdoni (Day) and all species included in Laguvia and Pseudolaguvia (of Ereth- istidae). Other recognized sisoroid miniatures oc- cur in the tribe Hoplomyzontini (Aspredinidae: Dupouyichthys sapito Schultz and Hoplomyzon papillatus Stewart; WeiUman & Vari, 1988). At the other extreme, sisoroids also include one gigantic species, the sisorid Bagarius yarrelli. This species is reported to reach close to 2 m in length (Roberts, 1983, 1989) and is one of the largest catfishes in Asia. It has been repeatedly observed that paedo- morphic traits commonly present in miniatures tend to occur in identical form in phylogenetically unrelated lineages (Weitzman & Vari, 1988; Schaefer et al., 1989;Begle, 1991; Grande, 1994). Such traits include various reductions in the la- terosensory canal system, number of fin rays, and degree of ossification. In some instances, reduc- tions are carried on to total absence of certain structures. Anatomical changes associated with gi- gantism seem to be far less profound than those related to miniaturization. This is not surprising, because reduction in size is nearly always associ- ated with a truncation of ontogeny, an association not occurring in cases of phylogenetic size increas- es. Most of the morphological alterations typically undergone by gigantic species seem to involve an increased degree of ossification, often resulting in thicker bones and tighter sutures. Although these changes may in some cases obscure some osteo- logical characters, their underlying condition can usually still be observed in half-grown individuals. From the viewpoint of phylogenetic reconstruc- tion, problems in character coding related to gi- gantism are far less critical than those related to miniaturization. All miniature sisoroids are in the families Er- ethistidae and Aspredinidae. The usual way to as- sess the number of miniaturization events in a group is to treat miniaturization as a character and optimize it on the most parsimonious tree topol- ogy. The number of independent origins plotted on the tree is the number of miniaturization events hypothesized to have occurred in the clade. On the cladogram advanced in this study, miniatur- izations within erethistids probably occurred at least twice: once in the Laguvia + Pseudolaguvia clade (tribe Laguviini) and once within the Hara + Erethistes clade. Miniaturization in this clade, involving Hara jerdoni and Erethistes maesoten- sis, is here conservatively considered as a single 70 FIELDIANA: ZOOLOGY event. Although the hypothesis that Erethistes and Hara form a monophyletic group is well corrob- orated, there is no evidence that either genus is monophyletic as presently constituted. Both gen- era are in need of revision and a fine-grained phy- logenetic treatment. At present, there is no clear reason to discard the possibility that E. maeso- tensis and H.jerdoni are sister groups, despite their present allocation to different genera, and there- fore accountable for by a single miniaturization event. This reasoning is in accordance with the purpose of assessing only the minimum number of miniaturization events. The two miniaturized species within hoplo- myzontin aspredinids have been hypothesized to represent independent miniaturization events (Weitzman & Vari, 1988, p. 456). Although those species are the smallest in the family, the whole tribe Hoplomyzontini can be considered minia- turized under a less restrictive definition of the term. All hoplomyzontines are very small in body size and have reductions in the laterosensory canal system and in the number of fin rays. The taxa phylogenetically close to miniaturized sisoroid clades are all of small size. All erethistids have small overall body size, and their closest rel- atives, the aspredinids, show a similar pattern. The only medium-sized aspredinids are the few species in the subfamily Aspredininae. However, all as- predinines have a markedly elongated caudal re- gion, which accounts for a large part of their SL and thus obscures the actual size condition of other body regions. This effect is similar to that referred to by Weitzman and Vari ( 1 988) as elongated min- iatures, when truly miniaturized species exceed a certain SL because of an elongated body shape. In contrast, the lineage that includes the gigantic Bagarius yarrelli comprises several species of me- dium size. Among its closest relatives, Sisor and some species ofGagata can reach beyond 200 mm SL, a size unknown in erethistids. The same is true of Glyptosteminae, where some Glyptothorax, Pseudecheneis, and Glyptostemina may reach 200 mm SL. Similarly, other species of Bagarius are also medium-size to large fishes, with B. bagarius reaching 192 mm SL and B. suchus reaching 515 mm SL (Roberts, 1983). These observations indicate that both minia- turization and gigantism are not abrupt events in the phylogenetic context for these groups. Extreme modifications of body size, toward either increase or decrease, seem to occur gradually across lin- eages. The definitions of miniaturization and gi- gantism apply to the most extreme cases of what seems to be a gradual phenomenon. Although de- tailed comparative data are not available, it seems that this rule applies to other catfish groups as well. All miniaturized members of the Neotropical fam- ily Trichomycteridae (Weitzman & Vari, 1988; de Pinna, 1989) belong to clades in which most or all species have small body size. The only miniatur- ized loricariid {Microlepidogaster lophophanes{E\- genmann and Eigenmann)) is a member of a sub- family (Hypoptopomatinae) where most species are small relative to the remainder of the family. Similarly, the various miniaturized callichthyids (Weitzman «fe Vari, 1988) are all related to small members of the family (Reis, 1993). How well other miniature siluriforms (and other fish groups) fit into this pattern still needs to be determined. Data on gigantism are far less extensive than those on miniaturization because fewer species fit in the former category. Preliminary observations, however, seem to indicate that a pattern similar to that in Sisoridae occurs in the Asiatic Panga- siidae and the Neotropical Pimelodidae. There are at least two pangasiid species that unquestionably qualify as %\%dinxic— Pangasius gigas and P. sani- twongsei, both reported to reach 3 m SL and 300 kg (Roberts & Vidthayanon, 1991). Little is known about the phylogenetic position of P. sanitwongsei, but the closest relative of P. gigas is probably P. hypophthalmus (Durand, 1949; Roberts & Vid- thayanon, 1 99 1), a species reaching at least 80 cm SL and 7 kg, and perhaps up to 1.3 m and 15.5 kg (Roberts & Vidthayanon, 1991). Pangasiidae includes exclusively medium-sized to large fishes, with most of them reaching beyond 50 cm SL as adults, several reaching 80 cm SL, and the smallest to ca. 20 cm SL (Roberts &. Vidthayanon, 1991). Within pimelodids, one member of the subfamily Sorubiminae is reported to reach beyond 2 m SL and 100 kg (e.g., Brachyplatystomafilamentosum; see Goulding, 1980). All other members of the Sorubiminae, probably a monophyletic group, are all medium or large in size, many of which reach- ing beyond 1 m SL as adults. Evolution of the Sound-producing Apparatus in Sisoroids The remarkable sound-producing apparatus present in Sisor, Gagata, Nangra, and Conta has been described above as character 109. The first published account of this complex was by Haddon (1881), who described its structure in Gagata (= DE PINNA: ASL\N CATFISH FAMILIES 71 Callomystax). The same observations were re- peated by Bridge and Haddon (1893), and later a very similar sound-producing apparatus was re- ported in Sisor by Mahajan (1963). Tilak (1963) briefly recorded the presence of a stridulatory ap- paratus in Gagata gagata and Nangra viridescens and reported its absence in the closely related Nan- gra punctata. The sound-producing ability of the stridulatory apparatus in the living fish has been reported only in Sisor (Mahajan, 1963), but the anatomical sim- ilarities of this apparatus in other taxa makes it likely that the mechanism plays a similar role in all of them. The sound emission is at least in part related to defense; Majahan (1963; p. 723) re- ported that '"''{Sisor] produces a very distinct and audible sound when caught in the aquarium and the intensity of sound increases considerably when the fish is taken out of water." Mahajan also noted that the sound produced by the fish was similar to that produced by manipulating the stridulatory apparatus of preserved fish. Whether the same stress reaction is displayed by Gagata, Nangra, and Conta remains to be investigated. The degree of development of the stridulatory apparatus varies among species. In Gagata and Sisor, it is always well developed and heavily os- sified. The same is true for Nangra viridescens. In A^. nangra, however, all elements of the apparatus are weak and thinly ossified, and the ridges are few and poorly differentiated. Nonetheless, the whole skeleton in that species is slight and thinly ossified, indicating that the reduced condition of the stridulatory apparatus is part of a trend toward reduction throughout the whole skeletal system in that species and not a reduction of the mechanism per se. Also in A^. nangra, the ridged surfaces of the supraneural (= first nuchal plate) are not tightly united into a solid structure as in A^. viridescens, Gagata, and Sisor. The same situation observed in A^. nangra probably occurs in N. punctata, judg- ing from illustrations by Tilak (1963, p. 305, Figs. 46, 47). The neural spine of the complex centrum in A^. punctata is broad at the tip, a condition typical of species with a stridulatory apparatus. As in A^. nangra, the ridged surfaces in A^. punctata probably are rather attenuated and visible with clarity only at certain angles and considerable magnification. The morphological modifications of the dorsal fin elements involved in the basic functioning of the stridulatory mechanism are very similar in all taxa and leave little room for an a priori hypothesis of nonhomology. In the phylogenetic hypothesis advanced here, Sisor, Nangra, and Gagata form a monophyletic group and the presence of the sound-producing mechanism is a synapomorphy (i.e., homologous) for that clade. However, a strid- ulatory apparatus based on a similar set of ana- tomical modifications and resulting in a very sim- ilar functional complex occurs in the genus Conta. Here, the apparatus is likewise formed by modi- fications of the skeletal elements associated with the dorsal fin (first nuchal plate) and complex cen- trum, and the sound-producing mechanism relies on the same overall physical basis, that is, a series of fine bony ridges on opposing surfaces. I assume a priori, based on a number of anatomical simi- larities (see description and discussion of character 109), that the conditions observed in Sisor, Nan- gra, and Gagata and those observed in Conta are homologous. Although that assumption may be the most reasonable course of action in view of the data available, assumptions about morpho- logical homology (the primary homologies of de Pinna, 1991) involve an element of subjectivity that may be critical in certain situations. There are borderline situations where two or more struc- tures are neither similar enough to unquestionably be considered homologous nor distinct enough to be considered nonhomologous. In a certain way, the stridulatory apparatus is in this category. In the phylogenetic hypothesis presented here, the Sisor/ Gagata/ Nangra clade (tribe Sisorini) is a number of nodes apart from Conta in the clado- gram, and this topology leaves no room for con- sidering the derived similarities in sound-produc- ing abilities as homologous in the two lineages. This result holds despite the initial working as- sumption, encased in the character coding used to build the data matrix, that the stridulatory mech- anism is a primary homologue in Conta and Si- sorini. Thus, some of the differences between the stridulatory apparatus in Conta and Sisorini can be interpreted a posteriori as evidence of nonhom- ology. The most obvious distinction is the plane of the ridged surfaces, nearly horizontal in Conta and vertical in Sisorini. Also, in Sisorini there is a trend toward approaching the ridged surface of either side of the supraneural, often forming a sin- gle median vertically ridged sphere. In Conta, there is an opposite tendency, with the crenulated sur- faces of the supraneural positioned wide apart and rather independent of one another except for their attachment on the same supraneural. Although these differences in themselves are not grounds for discarding the primary homology of the stridula- tory apparatus in Sisorini and Conta, they may 72 FIELDIANA: ZOOLOGY gain relevance under the available phylogeny, once the ontogeny of the different varieties of the mech- anism become known. A hypothesis of nonhom- ology would be supported if it is found that neither version of the apparatus is a stage in the devel- opment of the other, i.e., that they derive inde- pendently from a condition similar to that in gen- eralized sisoroids (see below). Based only on adult morphology, it is impossible to objectively estab- lish transformation series for the two states of the stridulatory apparatus. Both seem to be highly di- vergent from the plesiomorphic sisoroid condi- tion, but in opposite directions. In Conta, the trend has been for the crenulated surfaces of each side to diverge laterally so that the apparatus is broad as seen in dorsal view. In Sisorini, however, the crenulated surfaces approach the midline, and the apparatus is narrow. The primitive condition for the homologous structures in other members of Sisoridea is somewhere between the two derived conditions and cannot be unambiguously said to be more similar to either of them. I The parallel occurrence of such similar stridu- latory mechanisms within Sisoridea is intriguing, especially considering that the complex is un- known elsewhere in siluriforms. Nonetheless, there are synapomorphic modifications of sisoroids that can be interpreted as precursors to the stridulatory mechanism seen in some members of the group. As described for character 9 1 , the ventral portion of the first nuchal plate and the distal tip of the neural spine of the complex centrum contact one another by a modified surface in all members of Sisoridea. These modified surfaces are positionally homologous to the sound-producing ridged sur- faces of the stridulatory apparatus. Moreover, in several cases the contacting surfaces have a rough texture, indicating that whatever their function may be in the plesiomorphic sisoroid state, friction is important, just as with the ridged surfaces of the stridulatory apparatus. The transformation from the typical sisoroid condition to that of a fully formed stridulatory apparatus is therefore not dif- ficult to visualize. The stridulatory complex is cer- tainly more similar to the situation in generalized sisoroids than to the primitive siluriform condi- tion. The functional significance of the sisoroid mod- ifications of the supraneural and neural spine of the complex centrum (character 91) are unknown. These structures may also be involved in a sound- producing mechanism, although obviously a less efficient one than that of the Sisorini and Nangra. It is equally possible that these modifications serve as a reinforced mechanical tie between the dorsal fin complex and the vertebral column. Whatever the case may be, it is plausible that the repeated occurrence of such an unusual structure as the stridulatory apparatus in some sisoroids is in part accounted for by the preadaptation represented by specializations in the supraneural and correspond- ing neural spine typical of all the Sisoroidea. The case of the nonhomology of the stridulatory apparatus is an example of the importance of com- pleteness of taxonomic representation in phylo- genetic studies. The hypothesis of nonhomology is dependent on the presence of taxa without the mechanism phylogenetically located between Conta and Sisorini in the cladogram. It is only through the inclusion of those taxa in the analysis that optimization procedures bearing on a hy- pothesis of nonhomology become conclusive. In the absence of taxa such as the aspredinids, Ba- garius, and Glyptosteminae in this analysis, the nonhomology of the sound-producing mechanism could not possibly be discovered, no matter how accurate the hypothesis might otherwise be. Rea- sons for not including certain taxa can range from unavailability for study to extinction prior to dis- covery. Whichever the reason, it is clear that the completeness of taxonomic representation is crit- ical for hypotheses of character evolution, a con- clusion that was corroborated in a previous study dealing with a different group of fishes (Stiassny & de Pinna, 1 994) and in others dealing with fossil taxa(Gauthieretal., 1988;Donoghueetal., 1989). Evolution of the Thoracic Adhesive Apparatus The hypothesis of phylogenetic relationships among sisoroids presented here permits a more objective insight into the evolution of adhesive modifications in the group. This topic, in connec- tion with the broader question of adaptations to torrential environments, has been the object of copious discussion in the past, and much differ- ence of opinion has been the result of exiguous knowledge about relationships among the various taxa involved. As described in the discussion of character 1 10, the thoracic adhesive apparatus is formed by a series of roughly parallel integumen- tary ridges and grooves, the ridges having a dense covering of unculi. This structure, in a thoracic and/or abdominal location, is unique within Sil- uriformes. DE PINNA: ASIAN CATFISH FAMILIES 73 The optimization of the character in the most- parsimonious hypothesis makes it clear that the thoracic adhesive apparatus, as defined here, orig- inated twice within sisoroids: once in Glyptoster- ninae and again in Erethistidae. In both cases, the character is subsequently reversed in a subgroup. Within Glyptosteminae, the adhesive apparatus is hypothesized as secondarily lost in the subtribe Glyptostemina ("glyptostemoids"), and within er- ethistids a similar reversal happens in the tribe Erethistini. In both cases, an alternative optimi- zation of the character is possible, in which the adhesive apparatus is independently developed in Continae and Laguviini (of Erethistidae) and in Glyptothoracini and Pseudecheneidina (of Glyp- tosteminae). However, for reasons presented else- where (de Pinna, 1991), optimizations maximiz- ing reversals (ACCTRAN) are theoretically pref- erable and therefore adopted as the favored alter- native of character evolution. In accordance with this reasoning, the adhesive apparatus is consid- ered homologous in Continae and Laguviini and in Glyptothoracini and Pseudecheneidina. A hypothesis of nonhomology of the apparatus in Glyptosteminae and Erethistidae is not subject to considerations regarding treatment of ambig- uous optimizations. The transitions are unambig- uously convergent, and it is impossible to consider these two occurrences of the thoracic apparatus as homologous without violating parsimony. As with the stridulatory apparatus, it is puzzling that such a distinctive and unique structure within siluri- forms has arisen twice in a relatively small group distributed in the same general area. One expla- nation that is always possible when tracing char- acter evolution is that there are unknown taxa (extinct or not yet discovered) with the apparatus that would fit into key positions in the cladogram and overturn the optimization. For example, two successive sister groups to Sisorinae that happened to have the apparatus would suffice to reconfigure the optimization in such a way that the structure would then be hypothesized as homologous in all taxa. Thus, the inferred convergence may be an artifact of incomplete taxonomic representation. This scenario, although always a possibility, can- not affect the current hypothesis unless the rele- vant taxa are actually found. If we accept the hypothesis that the thoracic apparatus actually originated twice within siso- roids, we can look for some character in sisoroids in general that may have served as a preadaptation for the development of the stmcture. Roberts (1982) reported that Bagarius, a fish lacking the adhesive apparatus as defined here and in other papers, has well-defined patches of unculi over some areas of the skin. Such well-circumscribed patches differ from the more diffuse condition common in other ostariophysans and resemble those present, for example, at the edges of the apparatus in Conta (Figs. 50A,B). If the presence of these patches is confirmed in additional sisoroid taxa, it is possible that what is here called the adhesive apparatus is in fact an extreme condition of a widespread modification found in other si- soroids. If so, then the explanation for the repeated occurrence of the adhesive apparatus in closely related taxa is the same offered above for a similar situation observed for the stridulatory mecha- nism, i.e., the presence of a synapomorphic con- dition for a more inclusive group that served as a precursor. Another intriguing aspect of the evolution of the adhesive apparatus as inferred from the proposed phylogeny is its reversal in Glyptostemina and Erethistini. There is little doubt that the adhesive apparatus is important in providing the fish with a means of maintaining its position against strong currents, i.e., as an organ of attachment to the substrate. One would expect, therefore, that the taxa that lost the adhesive apparatus would be those that secondarily occupied lentic environ- ments. This is not always the case, however. The Glyptostemina comprises some of the fishes most highly adapted for torrential life, and all of them lack the thoracic apparatus (although they have other morphological specializations for that en- vironment). Among members of Erethistini, Hara and Erethistes are reported to inhabit muddy slug- gish waters overgrown with vegetation (Hora, 1 950, pp. 189-190; Jayaram, 1973, p. 388), but there is some discrepant information about Erethistoides (the sister group to the remainder of the tribe, lacking the thoracic apparatus). According to Hora (1950, pp. 190, 192) the genus is known to occupy fast-running waters, but Jayaram (1973, p. 388) stated that "'Erethistes montana [= Erethistoides montana] and Hara hara were collected from shal- low sluggish rivers of the Teesta drainage system." Clearly, more detailed information on the biology of erethistids and sisorids is needed for a more conclusive discussion of the adaptive/phylogenet- ic significance of the thoracic adhesive apparatus. A number of parallel adaptations to torrential environments have been previously observed be- tween Sisoridae and the Neotropical Loricariidae and Astroblepidae (cf. Hora, 1930). However, the most striking example of adaptation to life in tor- 74 FIELDIANA: ZOOLOGY rents among Neotropical catfishes is found in the family Trichomycteridae. The organs of attach- ment in that family are represented by modified opercular and interopercular bones armed with enlarged odontodes at their margins, a set of mod- ifications that is synapomorphic for the family. Contrary to the situation in other catfishes with organs of attachment, some trichomycterids have modified these adaptations for parasitic habits rather than for resisting water currents. In such cases, the attachment organs are used to penetrate and fix the parasitic fish inside its feeding site on the host (Gudger, 1930; Kelley & Atz, 1964; Ma- chado & Sazima, 1983). Nevertheless, most tri- chomycterids, including the basal subfamilies Co- pionodontinae and Trichogeninae (successive sis- ter groups to the remainder of the family; de Pinna, 1992), indeed use the attachment organs for life in torrents, indicating that this is the primitive functional role of the mechanism in the family. In trichomycterids that are neither parasitic nor in- habitants of torrential waters, there tends to be reduction or loss of the opercular and interoper- cular armature for attachment. Such is the case with several sand-dwelling members of the sub- families Glanapteryginae and Sarcoglanidinae (Myers, 1944; Myers & Weitzman, 1966; de Pin- na, 1989). Most trichomycterids inhabit fast-run- ning waters, often small creeks, and occur in nearly all highlands and mountainous regions in South America on both sides of the Andes. The torrent- inhabiting trichomycterids use their "adhesive" mechanism not only to resist currents but also to actively move against them and climb waterfalls. Their "elbowing" head movement to ascend cur- rents has been described on some occasions (e.g., Eigenmann, 1918, p. 262). Upriver migratory ac- tivities have been described for Trichomycterns migrans (cf. Dahl, 1960). Because of their comparable levels of diversi- fication and their adaptations to similar environ- ments, a comparison between the patterns of di- versity displayed by trichomycterids and sisoroids reveals some common elements about the evo- lutionary significance of adhesive adaptations. Within sisoroids, the most speciose clade is Glyp- tosteminae, with close to 50% of all species in the superfamily. This clade is apomorphically diag- nosed by the thoracic adhesive apparatus. Their sister group, subfamily Sisorinae, lacks such ad- aptations, and includes only about 1 5 species. In trichomycterids, the situation is more complex be- cause very few members of the family lack the attaching mechanism. But all clades that are apo- morphically diagnosed by the lack of the opercular and interopercular armature are small in absolute number of species. Nematogenyiidae, the sister group to Trichomycteridae (de Pinna, 1 992), lacks the opercular and interopercular armature and in- cludes a single species {Nematogenys inermis) en- demic to a small area in central Chile. Thus, in both trichomycterids and sisorids, the presence of attaching mechanisms seems to have been a de- termining factor for their distribution over wide areas and their concomitant diversification. One can say that the Andes are for trichomyc- terids what the Himalayas are for sisorids and er- ethistids. Both mountain chains represent a vast environment where adhesive and anchoring ad- aptations are obviously an advantage for disf>ersal over wide areas. Such wide distribution is prone to high levels of vicariant events. As expected, sisorids and trichomycterids are the most diverse and widely distributed siluriforms groups around each of those mountain chains. Relationships of the Neotropical Aspredinidae The family Aspredinidae is the most aberrant component of the Neotropical catfish fauna. The family's relationships have always been obscure, and there have been few explicit proposals about its possible closest relatives. Even the internal anatomy of aspredinids has rarely been studied, a situation that by and large continues to the present. Chardon (1967) proposed that aspredinids were related to loricarioids, composed the Neotropical families Nematogenyidae, Trichomycteridae, Cal- lichthyidae, Scoloplacidae, Loricariidae, and As- troblepidae. There is long-standing evidence that these six families form a natural group (e.g., Peyer, 1 922), and more recent phylogenetic treatment has strongly supported the monophyly of loricarioids (Baskin, 1973; Schaefer& Lauder, 1986;Schaefer, 1990). Chardon's (1967) proposal was that aspre- dinids should be added to that assemblage. How- ever, subsequent studies showed that Aspredini- dae did not actually share any putative synapo- morphies exclusively with loricarioids (Baskin, 1973). That family has remained incertae sedis within Siluriformes until recently, probably more as a result of the paucity of studies than because of real controversy. The anatomical distinctiveness of aspredinids puts a large gap between it and other Neotropical DE PINNA: ASIAN CATFISH FAMILIES 75 catfish families. Comparisons restricted in their coverage to a single continent will meet with dif- ficulty finding any special similarities joining as- predinids with any other Neotropical family or group of families. The first suggestion that aspre- dinids might be related to non-Neotropical taxa was by Ferraris (1989), who proposed that they were the sister group to the Akysidae. Although Ferraris's results were not published in full, they stimulated renewed interest in the problem and for the first time raised the possibility that the relationships of aspredinids might be outside the Neotropics. The hypothesis presented in this study for the placement of Aspredinidae relies on evidence at several levels. The evidence for individual clades is not particularly overwhelming, but the total ev- idence for all intemested clades bearing on aspre- dinid relationships is rather strong. Aspredinids share with the other members of Sisoroidea seven synapomorphies (nos. 1, 32, 61, 64, 77, 87, 91). Of these, four (nos. 32, 61, 64, 91) are unreversed and nearly unique within siluriforms. The pres- ence of the humerovertebral ligament (character 64), in particular, has not so far been observed in any other catfishes. The group called clade A above (Akysidae + Sisoridae + Erethistidae -I- Aspre- dinidae) relies on three synapomorphies; one of these (no. 24) shows reversals, and the other two (nos. 26, 90) have a number of putatively ho- moplastic occurrences elsewhere in siluriforms. Support for this clade is therefore weak. Another group including aspredinids, clade B (Aspredini- dae + Sisoridae + Erethistidae), is supported by 10 characters (nos. 27, 57, 62, 64, 66, 77, 79, 87, 88, 103), two of which (nos. 77, 87) are reversals. Although several of the other characters are them- selves reversed within the clade, most of the re- versals are concentrated in the highly modified tribe Glyptostemini, in particular in the subtribe Glyptostemina. The sister-group relationship be- tween aspredinids and erethistids (clade C) relies on five characters (nos. 3, 38, 84, 85, 97), of which only one (no. 97) is neither reversed nor conver- gent elsewhere in sisoroids. The placement of aspredinids within an inter- nested set of sisoroids, therefore, is supported by a total of 25 character transitions at several levels. Of these clades, the strongest seems to be the one joining sisorids, aspredinids, and erethistids (clade B). This clade is supported by the largest set of synapomorphies, and five extra steps are needed to collapse it. This is more steps than it takes to collapse any other component involving aspredi- nids, except for the whole superfamily. In com- parison, clade C and clade A both collapse at a single extra step. Therefore, although the relation- ships of aspredinids with erethistids alone are rel- atively fragile, their relationships with erethistids and sisorids are robust. Thus, in terms of the bio- geographical significance of hypothesized relation- ships, the alignment of Neotropical aspredinids with Asiatic taxa can be considered well corrob- orated. A remarkable process of skin molting so far ob- served only in aspredinids and akysids may even- tually help define some sisoroid subgroups. Fer- raris (1991, pp. 86, 87, 164) reported that aspre- dinids and the akysid Breitensteinia have been ob- served in aquarium conditions to shed their skin periodically. According to Ferraris, in Breiten- steinia the newly exposed skin is brighter than the old layer and shows the pigmentation pattern very vividly. Also, the skin is shed in large straps, oc- casionally in a nearly entire piece that looks like a peel of the fish. Both aspredinids and akysids have a very similar skin surface, with a rough tex- ture and numerous large tubercles. The nature of their probably heavily keratinized skin may ne- cessitate periodic molting. If this is so, skin shed- ding should also be searched for in sisorids and erethistids, most of which have a similar skin tex- ture, though normally less extreme. Unfortunate- ly, skin molting has to date been observed only accidentally and has never been studied in any detail; Ferraris's (1991) report seems to be the first and only published reference to the phenomenon. Sisoroids are only rarely kept in artificial condi- tions, and in the few cases observed the shed skin seems to be quickly consumed by other fish (Fer- raris, 1991, p. 86), making it even less likely to be detected. Skin molting has so far not been ob- served in any other siluriforms, and it is certainly a process of considerable biological and systematic significance. More observations are critically needed to understand the phenomenon in a phy- logenetic context. The inclusion of aspredinids within sisoroids, a group otherwise distributed exclusively in Asia, seems to clearly represent a trans-Pacific relation- ship. This kind of relationship has been observed in numerous zoological and botanical taxa, and its biogeographical implications are important (cf. Croizat, 1958; Melville, 1981; Nelson & Platnick, 1981). A number of models have been proposed to explain trans-Pacific relationships, the most fa- mous of which is the Pacifica hypothesis. That model jxjstulates a former continent located around 76 FIELDIANA: ZOOLOGY the upcoming Pacific basin that split up into frag- ments, which then colhded and fused with the continents presently around the Pacific rim (Hal- lier, 1912; Melville, 1966, 1967;Nur&Ben-Avra- ham, 1981). The Pacific Ocean, thus, originated between the rifting margins of Pacifica, and the expansion of the basin coincided with the frag- mentation of the continent. The Pacifica hypoth- esis is controversial, but trans- Pacific relationships unquestionably provide support for it. Most cases of trans- Pacific relationships, however, relate North America and East/Southeast Asia and involve mainly plant taxa. Trans-Pacific relationships linking animal groups in South America and Asia are rare. The nature of the trans-Pacific aspredinid/si- soroid relationship, however, is unclear. For it to be fully conclusive, sisoroids would have to con- tain some elements in Africa, and aspredinids would then have to be demonstrably closer to the Asiatic than to the African components. Only in that case would all elements relevant to the prob- lem be present. Based on our current understand- ing of sisoroid distribution, the trans-Pacific hy- pothesis relies partly on the absence of known rep- resentatives of the group in Africa rather than on the known presence of taxa in all the relevant ar- eas. Despite this, the relationships of aspredinids with Asiatic taxa is an intriguing phenomenon that deserves further study, perhaps using a data set of a different nature, and should stimulate the search for yet undiscovered fossil or recent sisoroids in Africa. Acknowledgments This study was begun during the course of a postdoctoral fellowship in the Division of Fishes, National Museum of Natural History, funded by the Smithsonian Institution. The work was com- pleted during a second postdoctoral appointment in the Department of Geology, Field Museum of Natural History, funded by the National Science Foundation. I am especially grateful for the sup- port and advice received from Richard Vari and Stanley Weitzman (at usnm), and Lance Grande (at fmnh). Naercio Menezes and the remaining staff of the Departamento de 2k)ologia da Univ- ersidade de Sao Paulo were most helpful in making the arrangements which permitted me to have the research time to complete this project. I also thank Carl Ferraris, Flavio Bockmann, John Lundberg, John Friel, Olivier Rieppel, and Heok Hee Ng for informative conversations on various topics re- lated to this paper. Carl was also the first to call my attention to the possible Asian affinities of aspredinids. Betty Strack kindly assisted me with SEM facilities at fmnh, and John Weinstein pre- pared all non-SEM photographs. For loans and exchanges of specimens and/or courtesies offered me during visits to their insti- tutions, I thank Barbara Brown, Norma Feinberg, Gareth Nelson, and Melanie Stiassny (amnh); Wil- liam Saul and Scott Schaefer (ansp); Darrell Sie- bert and Gordon Howes (bmnh); David Catania and William Eschmeyer (cas); John Lundberg (formerly duvc); Barry Chemoff, Mary Anne Rog- ers, and Kevin Swagel (fmnh); Richard Feeney and Robert Lavenberg (lacm); Heraldo Britski, Jose Figueiredo, Naercio Menezes, and Osvaldo Oyakawa (mzusp); Sven Kullander and Eric Ah- lander (nrm); William Fink and Douglas Nelson (ummz); Susan Jewett, Richard Vari, and Stanley Weitzman (usnm); Maurice Kottelat (cmk, for- merly zsm); and Yang Chang Man (zrc). The manuscript benefited greatly from com- ments and suggestions by Gloria Arratia, Carl Fer- raris, Joaquim Maia, Scott Schaefer, and Richard Vari. A particularly thorough review was done by Flavio Bockmann, for which I am especially grate- ful. Finally, I must add a word of appreciation for Raj Gauba, C. L. Mahajan, and Raj Tilak for their contributions to the knowledge of sisorid anato- my, which provided an initial impetus for the pres- ent study. Research funding is currently provided by CNPq and FAPESP. Literature Cited Alexander, R. M. 1 965. Structure and function in the catfish. Journal of Zoology, 148: 88-152. Arratia, G. 1987. Description of the primitive family Diplomystidae (Siluriformes, Teleostei, Pisces): Mor- phology, taxonomy and phylogenetic implications. Bonner Zoologische Monographien, 24: 1-123. . 1992. Development and variation of the sus- pensorium of primitive catfishes (Teleostei: Ostario- physi) and their phylogenetic relationships. Bonner Zoologische Monographien, 32: 1-149. Arratia, G., and L. Huaquin. 1995. Morphology of the lateral line system and of the skin of diplomystid and certain primitive loricarioid catfishes and system- atic and ecological considerations. Bonner Zoolo- gische Monographien, 36: 1-110. Arratia, G., AND H. -P. ScHULTZE. 1990. Theurohyal: Development and homology within osteichthyans. Journal of Morphology, 203: 247-282. DE PINNA: ASL\N CATFISH FAMILIES 77 Bailey, R. M., and J. N. Baskin. 1976. Scoloplax dicra, a new armored catfish from the Bolivian Am- azon. Occasional Papers of the Museum of Zoology, University of Michigan, 674: 1-14. Baskin, J. N. 1973. Structure and relationships of the Trichomycteridae. Ph.D. diss.. City University of New York, New York. Begle, D. p. 1991. Relationships of the osmeroid fishes and the use of reductive characters in phylogenetic analysis. Systematic Zoology, 40: 33-53. Berg, L. S. 1 940. Classification of fishes, both Recent and fossil. Travaux de Tlnstitut Zoologique de I'A- cademie des Sciences de I'URSS, 5(2): 87-517. Bhatia, B. 1950. Adaptive modifications in a hill- stream catfish, Glyptothorax telchittaiyidimWxon). Pro- ceedings of the National Institute of Sciences of India, 16(4): 271-285. Bleeker, M. p. 1853. Nalezingen op de ichthyolo- gische fauna van Bengalen en Hindostan. Verhande- lingen van het bataviaasch Genootschap van Kunsten en Wetenschappen, 25: 1-164. . 1858. De visschen van den Indischen Archipel. Siluri. Acta SocietatisScientiarum Indo-Neerlandicae, 4: 1-370 + xii. . 1862. Atlas Ichthyologique des Indes Orien- tales Neerlandaises, public sous les auspices du Gou- vemement colonial neerlandais. Tome II. Siluroides, Characoides et Heterobranchoides. Government pub- lication, Amsterdam. . 1863. Sistema silurorum revisimi. Neder- landsch Tijdschrift voor de Dierkimde, 1: 77-122. 1873. Memoire sur la faune ichthyologique de Chine. Nederlandsch Tijdschrift voor de Dierkunde, 4: 113-154. Blyth, E. 1858. Report of the Curator, Zoological De- partment, for May, 1858. Journal of the Asiatic So- ciety of Bengal, 27: 267-290. 1 860. Report on some fishes received chiefly from the Sitang River and its tributary streams, Ten- asserim Provinces. Journal of the Asiatic Society of Bengal, 29: 138-174. Boeseman, M. 1966. A new sisorid catfish from Java, Sundagagata robusta gen. et spec. nov. Proceedings of the Koninklijke Nederlandse Akademie van We- tenschappen Series C Zoology, 69: 242-247. Boltlenger, G. a. 1904. Teleostei (systematic part), pp. 541-727. In Harmer, S. F., and A. E. Shipley, eds.. The Cambridge Natural History, vol. 7. Macmillan, London. Bremer, K. 1994. Branch support and tree stability. Cladistics, 10: 295-304. Bridge, T. W., and A. C. Haddon. 1893. Contribu- tions to the anatomy of fishes. II. The air-bladder and Weberian ossicles in the siluroid fishes. Philosophical Transactions of the Royal Society of London Series B, 84: 65-333. Brooks, D. R., R. T. O'Grady, and E. O. Wiley. 1 986. A measure of the information content of phylogenetic trees, and its use as an optimality criterion. Systematic Zoology, 35:571-581. Brown,B. A., AND C.J. Ferraris. 1988. Comparative osteology of the Asian catfish family Chacidae, with the description of a new species from Burma. Amer- ican Museum Novitates, 2907: 1-16. Burgess, W. E. 1989. An Atlas of Freshwater and Marine Catfishes: A Preliminary Survey of the Silur- iformes. TFH PubUcations, Neptune City, N.J. Carpenter, J. M. 1988. Choosing among multiple equally parsimonious cladograms. Cladistics, 4: 291- 296. Chardon, M. 1967. Reconnaissance d'un groupe na- turel de six families de Siluriformes Sud-Americains grSce k I'etude anatomique de Tappareil de Weber au sens large. Annales de la Societe Royale Zoologique de Belgique, 97(1): 35-38. . 1968. Anatomic comparee de I'appareil de We- ber et des structures connexes chez les Siluriformes. Annales du Musee Royale de I'Afrique Centrale Serie 8, 169: 1-277. Chaudhuri, B. L. 1919. Report on a small collection of fish from Putao (Hkamti Long) on the northern frontier of Burma. Records of the Indian Museum, 16: 271-287, pi. 22. Chen, X. 1994. Phylogenetic studies of the amblycipi- tid catfishes (Teleostei, Siluriformes) with species ac- counts. Ph.D. diss., Duke University, Durham. [Not seen.] Chen, X., and J. G. Liwdberg. 1995. Xiurenbagrus. a new genus of amblycipitid catfishes (Teleostei: Siluri- formes), and phylogenetic relationships among the genera of Amblycipitidae. Copeia, 1995(4): 780-800. Chu, X.-L. 1979. Systematics and evolutionary pedi- gree of the glyptostemoid fishes (family Sisoridae). Acta Zootaxonomica Sinica, 4(1): 72-82. [In Chinese.] . 1981. Taxonomic revision of the genera Pa- reuchiloglanis and Euchiloglanis. Zoological Re- search, 2(1): 25-31. [In Chinese.] 1982. Phylogeny of the genus Pseudecheneis (Siluriformes, Sisoridae), with descriptions of two new species. Acta Zootaxonomica Sinica, 7(4): 428-437. [In Chinese.] CoAD, B. W. 1981. Glyptothorax silviae, a new species of sisorid catfish from southwestern Iran. Japanese Journal of Ichthyology, 27(4): 291-295. CoAD, B. W., AND G. B. Delmastro. 1985. Notes on a sisorid catfish from the Black Sea drainage of Turkey. Cybium, 9(3): 221-224. Croizat, L. 1958. Panbiogeography. Published by the author, Caracas, Venezuela. Dahl, G. 1 960. Nematognathous fishes collected dur- ing the Macarena Expedition, 1959. Novedades Co- lombianas, 1(5): 302-317. Day, F. 1871. On the freshwater siluroids of India and Burmah. Proceedings of the Zoological Society, Lon- don, 1871: 703. . 1873. Report on the Fresh Water Fish and Fisheries of India and Burma. Office of the Superin- tendent of Government Printing, Calcutta, India. Ding, R.-H., T.-Y. Fu, AND M.-R. Ye. 1991. Two new species of the genus Pareuchiloglanis from China. Acta Zootaxonomica Sinica, 6(3): 369-374. [In Chinese.] DiNGERKUS, G., AND C. Uhler. 1 977. Enzyme clearing of alcian-blue stained whole small vertebrates for dem- onstration of cartilage. Stain Technology, 52: 229-232. 78 FIELDIANA: ZOOLOGY DoNOGHUE, M., J. Doyle, J. Gauthier, A. Kluge, and T. RowE. 1989. The importance of fossils in phy- logeny reconstniction. Annual Review of Ecology and Systematics, 20: 431^60. DuRAND, J. 1949. Les foraies jeunes du Pa/i^oy/ano- h don gigas Chevey. Le genre Pangasianodon. Buletin W de la Societe Zoologique de France, 74: 112-116. I EiGENMANN, C. H. 1918. The Pygidiidae, a family of South American catfishes. Memoirs of the Carnegie Museum, 7: 259-398. Eldredge, N., and J. Cracraft. 1980. Phylogenetic Patterns and the Evolutionary Process. Columbia Uni- versity Press, New York. Eschmeyer, W. N. 1990. Catalog of the Genera of Recent Fishes. California Academy of Sciences, San i Francisco. Farris, J. S. 1969. A successive approximations ap- proach to character weighting. Systematic Zoology, 18: 374-385. . 1983. The logical basis of phylogenetic anal- Erethistoides montana Hora. Zoologische Beitrage, 15: 185-201. 1970b. On the cranial osteology of two Indian ysis, pp. 7-36. In Platnick, N. I., and V. A. Funk, eds., Advances in Cladistics, vol 2. Columbia University Press, New York. . 1988. Hennig86, version 1.5. Distributed by the author, Port Jefferson Station, N.Y. Ferraris, C. 1988. The Auchenipteridae: Putative monophyly and systematics, with a classification of the Neotropical doradoid catfishes (Ostariophysi, Sil- uriformes). Ph.D. diss.. City University of New York, New York. . 1989. On the interrelationships between the Aspredinidae and the Akysidae (Ostariophysi, Silur- iformes), p. 86. In Abstracts of the 69th Annual Meet- ing, American Society of Ichthyologists and Herpe- tologists. ASIH, Arizona State University, Tempe. . 1991. Catfish in the Aquarium. Tetra Press, Morris Plains, N.J. Fink, S. V., and W. Fink. 1981. Interrelationships of the ostariophysan fishes (Pisces, Teleostei). Zoological Journal of the Linnean Society, 72: 297-353. Fitzhugh, K. 1 989. Cladistics in the fast lane. Journal of the New York Entomological Society, 97: 234-241. Friel, J. P. 1994. A phylogenetic study of the Neo- tropical banjo catfishes (Teleostei: Siluriformes: As- predinidae). Ph.D. diss., Duke University, Durham. Galtba, R. K. 1962. The endoskeleton of Bagarius bagarius (Ham.), part I— The skull. Agra University Journal of Research, 11(1): 75-90. . 1966. Studies on the osteology of Indian sisorid catfishes— II. The skull oi Glyptothorax cavia. Copeia, 1966:802-810. . 1967. Studies on the skull of the Indian Sisorid catfish, Erethistes pussilus. Journal of Zoology Lon- don, 151: 379-388. . 1968. On the morphology of the skull of catfish Pseudecheneis sulcatus. Zoologische Anzeiger, 181: 226-236. . 1969. The head skeleton of G/yp/o^/er/iwm re- ticulatum McClelland & Grifith. Monitore Zoologico Italian©, 3: 1-17. . 1970a. Contributions on the skull of a catfish. catfishes of the genus Laguvia. Zoologische Anzeiger, 185: 55-67. Gauthier, J., A. G. Kluge, and T. Rowe. 1988. Am- niote phylogeny and the importance of fossils. Cla- distics, 4: 105-209. Geoffroy St. Hillaire, I. 1829. Histoire Naturelle des Poissons du Nil, pp. 141-338. In Description de I'Egypte (ed. Commission des Monuments de I'E- gypte), 2nd ed., vol. 24. C. L. F. Panckoucke, Paris. Gill, T. 1861. Synopsis ofthe genera of the sub-family of Pimelodinae. Proceedings of the Boston Society of Natural History, 8: 46-55. . 1872. Arrangement ofthe families of fishes, or classes Pisces, Marsipobranchii, and Leptocardii. Smithsonian Miscellaneous Collection, 247: 1-49 + xlvi. Gcx)DRiCH, E. S. 1909. Vertebrata craniata (first fas- cicle: cyclostomes and fishes), pp. xvi + 1-518. In Lankester, R., ed., A Treatise on Zoology, part 9. Adam and Charles Black, Lx>ndon. GotJLDiNG, M. 1980. The Fishes and the Forest: Ex- plorations in Amazonian Natural History. University of California Press, Berkeley. Grande, L. 1987. Redeschption of "fHypsidoris far- sonensis (Teleostei, Siluriformes), with a reassessment of its phylogenetic position. Journal of Vertebrate Pa- leontology, 7: 24-54. Grande, T. 1994. Phylogeny and paedomorphosis in an African family of freshwater fishes (Gonorynchi- formes: Kneriidae). Fieldiana: Zoology, 78: 1-20. Greenwood, P. H., D. E. Rosen, S. H. Weitzman, and G. S. Myers. 1966. Phyletic studies of teleostean fishes, with a provisional classification of living forms. Bulletin ofthe American Museum of Natural History, 131: 341^55. Gudger, E. W. 1930. On the alleged penetration of the human urethra by an Amazonian catfish called candiru, with a review of the allied habits of other members of the family Pygidiidae, parts I and II. American Journal of Surgery, 8: 170-188, 443-457, Gunther, A. 1864. Catalogue ofthe Fishes in the Brit- ish Museum, vol. 5. British Museum of Natural His- tory, London. Haddon, a. C. 1881. On the stridulating apparatus of Callomystax gagata. Journal of Anatomy, 15: 323- 326. Hallier, H. 1912. Uber friihere Landbriicken, Pflan- zen- imd Volkerwanderungen zwischen Australaien und Amerika. Mededeelingen van's Rijks Herbarium, 13(13): 1-32. Hamilton, F. 1822. An Account of the Fishes Found in the River Ganges and Its Branches. Edinburgh. He, M., AND Y. Chen. 1981. Two new catfishes of the genus Akysis from China. Zoological Research, 2(3): 209-214. [In Chinese.] Hennig, W. 1950. Grundziige einer Theorie der phy- logenetischen Systematik. Deutscher Zentralverlag, Berlin. . 1966. Phylogenetic Systematics. University of Illinois Press, Urbana. DE PINNA: ASIAN CATFISH FAMILIES 79 . 1968. Elementos de una Sistematica Fil(^- netica. Editorial Univcrsitaria de Buenos Aires, Ri- vadavia. Herre, a. W. 1940. New species of fishes from the Malay Peninsula and Borneo. Bulletin of the Raffles Museum, 16: 5-26. HoRA, S. L. 1921. On some new or rare species of fish from the eastern Himalayas. Records of the Indian Museum, 22:731-744. . 1922. Structural modifications in the fish of mountain torrents. Records of the Indian Museum, 24:31-61. . 1923. Notes on fishes in the Indian Museum. V. On the composite genus Glyptosternon McClelland. Records of the Indian Museum, 25: 1-44. 1 930. Ecology, bionomics and evolution of the torrential faima, with special reference to the organs of attachment. Philosophical Transactions of the Roy- al Society of London Series B, 218: 171-282. 1936. Siluroid fishes oflndia, Burma and Cey- lon. II. Fishes of the genus Akysis Bleeker. Records of the Indian Museum, 38: 199-202. . 1937. Notes on fishes in the Indian Museum— XXXVI. On a new genus of Chinese catfishes allied to Pseudecheneis. Records of the Indian Museum, 39: 348-350. . 1 950. Siluroid fishes oflndia. Burma and Cey- lon. XIII. Fishes of the genera Erethistes Miiller & Troschel, Hara Blyth and of two new allied genera. Records of the Indian Museum, 47: 183-202. 1952. Parallel evolution of Pseudecheneis Blyth and similar fishes of South-East Asia. Journal of the Asiatic Society, Science, 18: 123-128. HoRA, S. L., AND P. Chabanaud. 1930. The siluroid fish Pseudecheneis and an allied new genus. Records of the Indian Museimi, 32: 2 1 5-222. HoRA, S. L., AND N. C. Law. 1941. Siluroid fishes of India, Burma and Ceylon. Records of the Indian Mu- seum, 43: 9-42. HoRA, S. L., AND A. S. Menon. 1948. Systematic po- sition of three glyptostemoid fishes of the family Si- soridae (order Siluroidae). Proceedings of the National Institute of Sciences oflndia, 46: 55-62. H<»A, S. L., AND E. G. SnAS. 1952a. Evolution and distribution of glyptostemoid fishes of the family Si- soridae (order Siluroidea). Proceedings of the National Institute of Sciences of India, 18: 309-322. . 1952b. Notes on fishes in the Indian Museum. XL VII. Revision of the glyptostemoid fishes of the family Sisoridae, with description of new genera and species. Records of the Indian Museum, 49: 5-30. Howes, G. J. 1983. The cranial muscles of loricarioid catfishes, their homologies and value as taxonomic characters (Teleostei, Siluriformes). Bulletin of the British Museum of Natural History, Zoology, 45: 309- 345. ICZN. 1985. International Code of Zoological No- menclatiu^, 3rd ed. University of Califomia Press, Berkeley. Jay ARAM, K. C. 1973. Siluroid fishes of India, Burma and Ceylon. 18. Systematic position of the genus La- guvia Hora and its relationships. Records of the Zoo- logical Survey of India, 67: 385-389. . 1981. The Freshwater Fishes of India, Paki- stan, Bangladesh, Burma and Sri Lanka— A Hand- book. Zoological Survey of India, Calcutta. Kelley,W. A.,andJ. W. Atz. 1964. A pygidiid catfish that can suck blood from goldfish. Copeia, 1964: 702- 704. KiMURA, S. 1934. Descriptions of the fishes collected from the Yangtze-kiang, China, by late Dr. K. Kish- inouye and his party in 1927-1929. Journal of the Shanghai Science Institute, Section 3, 1: 1 1-247. KoTTEiAT, M. 1983. A new species of fre/Aisrw MQller & Troschel from Thailand and Burma (Osteichthyes: Siluriformes: Sisoridae). Hydrobiologia, 107: 71-74. KoTTEU^T, M., AND C. ViDTHAYANON. 1993. Boraros micros a new genus and species of minute freshwater fish from Thailand (Teleostei: Cyprinidae). Ichthyo- logical Exploration of Freshwaters, 4: 161-176. Lm-, M. B., a. N. Bhatnagar, and J. P. Uniyal. 1 966. Adhesive modifications of a hillstream fish Glypto- thorax pectinopterus (McClelland). Proceedings of the National Academy of Sciences India, 36: 109-1 16. Li, S. 1986. Systematics, distribution and evolution of Glyptothorax (Siluriformes: Sisoridae), pp. 521-528. In Uyeno, T., R. Arai, T. Taniuchi, and K. Matsuura, eds., Indo- Pacific Fish Biology: Proceedings of the Sec- ond International Conference on Indo- Pacific Fishes. Ichthyological Society of Jaiian, Tokyo. Lundberg, J. G. 1970. The evolutionary history of North American catfishes, family Ictaluridae. Ph.D. diss.. University of Michigan, Ann Arbor. . 1975. Homologies of the upper shoulder girdle and temporal region bones in catfishes (order Siluri- formes), with comments on the skull of the Helogen- idae. Copeia, 1975: 66-74. 1 982. The comparative anatomy of the tooth- less blindcat, Trogloglanis pattersoni Eigenmann, with a phylogenetic analysis of the ictalurid catfishes. Mis- cellaneous Publications of the Museum of Zoology, University of Michigan, 163: 1-85. Lundberg, J. G., and J. N. Baskin. 1 969. The caudal skeleton of catfishes, order Siluriformes. American Museum Novitates, 2399: 1-49. Machado, F. a., and I. Sazima. 1983. Comporta- mento alimentar do peixe hematofago Branchioica bertonii (Siluriformes, Trichomycteridae). Ciencia e Cultura, 35: 344-348. Maddison, W. p., M. J. DoNOGHUE, and D. R. Mai> DisoN. 1 984. Outgroup analysis and parsimony. Sys- tematic Zoology, 33: 83-103. Mago-Lecoa, F. 1978. Los peces de la familia Ster- nopygidae de Venezuela. Acta Cientifica Venezolana, 29: 1-89. Mahajan, C. L. 1963. Soimd producing apparatus in an Indian catfish Sisor rhabdophorus Hamilton. Jour- nal of the Linnean Society, Zoology, 44: 72 1-724. . 1966a. Sensory canals of the head in Sisor rhabdophorus Hamilton. Transactions of the Ameri- can Microscopical Society, 85: 548-555. 1966b. Sisor rhabdophorus— \ study in ad- aptation and natural relationship. I. The head skele- ton. Journal of Zoology London, 149: 365-393. 1967a. Sisor rhabdophorus— \ study in ad- aptation and natural relationship. II. The interrela- 80 FIELDIANA: ZOOLOGY tionship of the gas bladder, Weberian apparatus, and membranous labyrinth. Journal of Zoology Lx)ndon, 151:417-432. . 1967b. Sisor rhabdophorus—A study in ad- aptation and natural relationship. III. The vertebral column, median fins and their musculature. Journal of Zoology Lx)ndon, 152: 297-318. . 1967c. The integument, exoskeleton and cu- taneous sense organs of Sisor rhabdophoms Hamilton. Proceedings of the National Institute of Sciences of India B, 33: 27-36. McCleixand, J. 1842. On the freshwater fishes col- lected by William Grifith, Esq., F. L. S. Madras Med- ical Service, during his travels under the orders of the Supreme Government of India, from 1835 to 1842. Journal of Natural History, Calcutta, 2: 560-589. Melville, R. 1966. Continental drift, mesozoic con- tinents and the migrations of the angiosperms. Nature, 211: 116-120. . 1967. The distribution of land around the Te- thys Sea and its bearing on modem plant distribution, pp. 291-310. In Adams, C. G., and D. V. Ager, eds., Aspects of Tethyan Biogeography, no. 7. The System- atics Association, London. 1981. Vicarious plant distributions and paleo- geography of the Pacific region, pp. 238-274. In Nel- son, G., and D. Rosen, eds., Vicariance Biogeography: A Critique. Columbia University Press, New York. Menon, M. a. S. 1 954. Notes on fishes of genus Glyp- tothorax. Records of the Indian Museum, 52: 27-54. Misra, K. S. 1976. The Fauna of India and the Ad- jacent Countries— Pisces, vol. III. Zoological Survey of India, Calcutta. Mo, T. 1991. Anatomy, Relationships and Systematics of the Bagridae (Teleostei: Siluroidei)— With a Hy- pothesis of Siluroid Phylogeny. Theses Zoologicae 17. Koeltz Scientific Books, Konigstein. Mo, T., AND X. Chu. 1986. A revision of the sisorid catfish genus Glyptothorax from China. Zoological Re- search, 7: 339-350. [In Chinese.] MiJLLER, J., AND F. H. Troschel. 1849. Beschreibung und Abbildung neuer Fische; die Familie Characinen. Horae Zoologicae, 3: 1-27. Myers, G. S. 1 944. Two extraordinary new blind ne- matognath fishes from the rio Negro, representing a new subfamily of Pygidiidae, with a rearrangement of the genera of the family and illustrations of some pre- viously described genera and species from Venezuela and Brazil. Proceedings of the California Academy of Sciences, 23: 591-602. Myers, G. S., and S. H. Weftzman. 1966. Two re- markable new trichomycterid catfishes from the Am- azon basin in Brazil and Colombia. Journal of Zool- ogy, 149: 277-287. Nelson, G. J., and N. I. Platnick. 1981. Systematics and Biogeography: Cladistics and Vicariance. Colum- bia University Press, New York. Nelson, J. S. 1994. Fishes of the World, 3rd ed. John Wiley & Sons, New York. Nixon, K. C. 1 992. CLAEKDS, version 1 .0. Distributed by the author, Ithaca, New York. NixoN,K. C.,andJ. I. Davis. 1991. Polymorphic taxa. missing values and cladistic analysis. Cladistics, 7: 233- 241. NuR, A., AND Z. Ben-Avraham. 1981. Lx>st Pacifica continent: a mobilistic speculation, pp. 341-358. In Nelson, G., and D. Rosen, eds., Vicariance Biogeog- raphy: A Critique. Columbia University Press, New York. Peyer, B. 1922. Uber die Flossenstacheln der Welse und Panzerwelse, sowie des Karpfens. Morphologisch- esjahrbuch, 51:493-554. de Pinna, M. C. C. 1 989. A new sarcoglanidine catfish, phylogeny of its subfamily, and an appraisal of the phyletic status of the Trichomycterinae. American Museum Novitates, 2950: 1-39. . 1991. Concepts and tests of homology in the cladistic paradigm. Cladistics, 7: 367-394. . 1992. A new subfamily of Trichomycteridae (Teleostei, Siluriformes), lower loricarioid relation- ships and a discussion on the impact of additional taxa for phylogenetic analysis. Journal of the Linnean Society, Zoology, 106: 175-229. 1 993. Higher-level phylogeny of Siluriformes, with a new classification of the order (Teleostei, Os- tariophysi). Ph.D. diss.. City University of New York/ American Museum of Natural History, New York. DE Pinna, M. C. C, AND R. P. Vari. 1995. Monophyly and phylogenetic diagnosis of the family Cetopsidae, with synonymization of the Helogenidae (Teleostei: Siluriformes). Smithsonian Contributions to Zoology, 571: 1-26. Platnick, N. I., C. E. Griswold, and J. A. Coddington. 1991. On missing entries in cladistic analysis. Cladis- tics, 7: 337-343. Reed, H. D. 1924a. The morphology and growth of the spines of siluroid fishes. Journal of Morphology, 38:431^51. . 1924b. The morphology of the dermal glands in nematognathous fishes. Zeitschrift fur Morphologie und Anthropologic, 24: 227-264. Regan, C. T. 1911. The classification of the teleostean fishes of the order Ostariophysi. 2. Siluroidea. Annals and Magazine of Natural History, 8: 553-577. Reis, R. E. 1993. Filogenia da familia Callichthyidae (Ostariophysi, Siluriformes), com uma revisao tax- onomica do genero Hoplosternum. Ph.D. diss., Tese de Doutoramento, Institute de Biocifincias, Univer- sidade de Sao Paulo, Sao Paulo. RiEPPEL, O. 1988. Fundamentals of Comparative Bi- ology. Birkhauser Verlag, Basel. Roberts, T.R. 1982. Unculi (homy projections arising from single cells), an adaptive feature of the epidermis of ostariophysan fishes. Zoologica Scripta, 11: 55-76. . 1983. Revision of the south and southeast Asian sisorid catfish genus Bagarius, with description of a new species from the Mekong. Copeia, 1983: 435-445. . 1989. The freshwater fishes of western Borneo (Kalimantan Barat, Indonesia). Memoirs of the Cali- fomia Academy of Sciences, 14: 1-210 + xii. Roberts, T. R., and C. Vidthayanon. 1 99 1 . System- atic revision of the Asian catfish family Pangasiidae, with biological observations and descriptions of three new species. Proceedings of the Academy of Natural Sciences of Philadelphia, 143: 97-144. DE PINNA: ASIAN CATHSH FAMILIES 81 ROYERO, R. 1987. Morfologia de la aleta dorsal en los bagres (Teleostei: Siluriformes), con especial referen- da a las familias americanas. Bachelor's thesis, Tra- bajo Especial de Grado, Universidad Central de Ven- ezuela, Facultad de Ciencias, Caracas. Saxena, S. C, AND M. Chandy. 1966. The pelvic gir- dle and fin in certain Indian hill stream fishes. Journal of Zoology London, 148: 167-190. ScHAEFER, S. A. 1987. Osteology of Hypostomus ple- costomus (Linnaeus), with a phylogenetic analysis of the loricariid subfamilies (Pisces: Siluroidei). Contri- butions in Science (Los Angeles), 394: 1-31 + iv. . 1990. Anatomy and relationships of scolopla- cid catfishes. Proceedings of the Academy of Natural Sciences of Philadelphia, 142: 167-210. ScHAEFER, S., AND G. V. Laijder. 1986. Historical transformation of fimctional design: Evolutionary morphology of feeding mechanisms in loricarioid cat- fishes. Systematic Zoology, 35: 489-508. ScHAEFER, S., S. H. Weftzman, AND H. A. Brttski. 1989. Review of the neotropical catfish genus Scoloplax (Pi- sces: Loricarioidea: Scoloplacidae) with comments on reductive characters in phylogenetic analysis. Pro- ceedings of the Academy of Natural Sciences of Phil- adelphia, 141: 181-211. ScHOCH, R. M. 1 986. Phylogeny Reconstruction in Pa- leontology. Van Nostrand Reinhold, New York. Scopou, G. A. 1777. Introductio ad historiam natur- alem, sistens genera lapidum, plantarum et animalium hactenus detecta, caracteribus essentialibus donata, in tribus divisa, subinde ad leges naturae. Prague. Shelden, F. F. 1937. Osteology, myology, and prob- able evolution of the nematognath pelvic girdle. An- nals of the New York Academy of Sciences, 37: 1-96. Siebert, D. J. 1 987. Interrelationships among families of the order Cypriniformes (Teleostei). Ph.D. diss.. City University of New York, New York. Singh, N., and N. K. Agarwal. 1 99 1 . The SEM sur- face structure of the adhesive organ oi Pseudoecheneis sulcatus McClelland (Teleostei: Sisoridae), from Ga- rhwal Himalayan hillstreams. Acta Ichthyologica et Piscatoria, 21(2): 29-35. Skelton, p. H. 1 985. Two new Amphilius (Pisces, Sil- uroidei, Amphiliidae) from the Zaire River system, Africa. Revue de Zoologie Africaine, 99: 263-291. Skelton, P. H., L. RiscH, AND L. De Vos. 1984. On the generic identity of the Gephyroglanis catfishes from southern Africa (Pisces, Siluroidei, Bagridae). Revue de 2^1ogie Africaine, 98: 337-372. Steinitz,H. 1961. Notes on some glyptostemoid fishes (Sisoridae, Siluroidea). Mitteilungen aus dem Ham- burgischen Zoologischen Museum und Institut, 59: 105-115. Steyskai-, G. C. 1980. The grammar of family-group names as exemplified by those of fishes. Proceedings of the Biological Society of Washington, 93: 168-177. Stiassny, M. L.J.,andM.C. C. dePinna. 1994. Basal taxa and the role of cladistic patterns in the evaluation of conservation priorities: A view from freshwater, pp. 235-249. In Forey, P. L., C. J. Humphries, and R. I. Vane- Wright, eds., Systematics and Conservation Evaluation. Systematics Association special volume 50. Clarendon Press, Oxford, England. Swofford, D. L. 1989. PAUP: Phylogenetic analysis using parsimony, version 3.0. Illinois Natural History Survey, Champaign. Swofford, D. L., and W. P. Maddison. 1987. Re- constructing ancestral character states under Wagner parsimony. Mathematical Biosciences, 87: 199-229. Swofford, D. L., and G. J. Olsen. 1990. Phylogeny reconstruction, pp. 41 1-501. In Hillis, D. M., and C. Moritz, eds.. Molecular Systematics. Sinauer, Sun- derland, Mass. Ta VOLGA, W. N. 1962. Mechanisms of sound produc- tion in the ariid catfishes Galeichthys and Bagre. Bul- letin of the American Museum of Natural History, 124: 1-30. 1971. Sound production and detection, pp. 1 35-205. In Hoar, W. S., and D. J. Randall, eds.. Fish Physiology, vol. 5. Academic Press, New York. 1977. Sound Production in Fishes. Dowder, Hutchinson & Ross, Stroudsberg, Germany. Taylor, W. R., and G. Van Dyke. 1985. Revised procedures for staining and clearing small fishes and other vertebrates for bone and cartilage study. Cy- bium, 9: 107-119. TiLAK, R. 1963. The osteocranium and the Weberian apparatus of the fishes of the family Sisoridae (Silu- roidea): A study in adaptation and taxonomy. Zeit- schrift fiir Wissenschaftliche Zoologie, 169: 281-320. . 1976. The adhesive thoracic apparatus in the evolution of glyptothoracoid fishes (Sisoridae: Siluri- formes). Zoologische Anzeiger, 196: 255-261. Vari, R. p. 1979. Anatomy, relationships and classi- fication of the families Citharinidae and Distichodon- tidae (Pisces, Characoidea). Bulletin of the British Mu- seum of Natural History, Zoology, 36: 26 1-344. Weber, M., and L. F. De Beaufort. 1913. The Fishes of the Indo-Australian Archipelago. II. Malacoptery- gii, Myctophoidea, Ostariophysi: I. Siluroidea. Brill, Leiden. Weitzman, S. H. 1954. The osteology and relation- ships of the South American characid fishes of the subfamily Gasteropelecinae. Stanford Ichthyological Bulletin, 4: 212-263. . 1962. The osteolc^ of Brycon meeki, a gen- eralized characid fish, with an osteological definition of the family. Stanford Ichthyological Bulletin, 8: 1- 77. . 1974. Osteology and evolutionary relation- ships of the Stemoptychidae, with a new classification of stomiatoid families. Bulletin of the American Mu- seum of Natural History, 153: 327-478. Weftzman, S. H., and R. P. Vari. 1988. Miniaturi- zation in South American freshwater fishes: An over- view and discussion. Proceedings of the Biological So- ciety of Washington, 101: 444-465. Wiley, E. O. 1981. Phylogenetics: The Theory and Practice of Phylogenetic Systematics. John Wiley & Sons, New York. Wiley, E. O., D. Siegel-Causey, D. R. Brooks, and V. A. Funk. 1991. The compleat cladist: A primer of phylogenetic procedures. University of Kansas Mu- seum of Natural History Special Publication, 19: 1- 158 + X. 82 HELDIANA: ZOOLOGY p Yeates, D. 1 992. Why remove autapomorphies? Cla- decheneis with comments on osteological differentia- distics, 8: 387-389. tions at species level. Acta Zootaxonomica Sinica, 17: Zhou, W., & X.-L. Chu. 1992. A new species of P^ew- 110-115. DE PINNA: ASIAN CATHSH FAMILIES 83