Novjitates PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

AMERICAN MUSEUM NovJitates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 2950, 39 pp., 25 figs., 1 table August 9, 1989

A New Sarcoglanidine Catfish, Phylogeny of Its Subfamily, and an Appraisal of the Phyletic Status of the Trichomycterinae (Teleostei, Trichomycteridae)

MARIO C. C. de PINNA1

ABSTRACT Stauroglanis gouldingi is described as new genus Scleronema and two species of Trichomycterus, T. and species of the Trichomycteridae. Characters boylei and T. santaeritae, are tentatively proposed traditionally used to delimit subgroups of tricho- to be more closely related to the Sarcoglanidinae mycterids would include Stauroglanis in the than to the Trichomycterinae. In addition, it is subfamily Trichomycterinae. Nevertheless, simi- suggested that T. hasemani and T. johnsoni are larities between Stauroglanis and trichomycter- more closely related to the Tridentinae than to any ines cannot be considered as derived characters Trichomycterinae. These cases are given as posi- for the two taxa. Instead, a number ofother newly tive evidence that the Trichomycterinae is non- proposed characters constitute unequivocal syn- monophyletic as presently constituted. apomorphies supporting monophyly ofStaurogla- By current criteria, Stauroglanis can be consid- nis plus the Sarcoglanidinae. Further analysis in- ered as one more miniature trichomycterid. In view dicates that Sarcoglanis and Malacoglanis (the two ofnew data, three miniaturization events are pro- previously known sarcoglanidines) form a mono- posed within the family: at the base of Glanap- phyletic group, with Stauroglanis as its sister group. teryginae/Sarcoglanidinae clade (including All characters used until the present to define Trichomycterus santaeritae), at the base of Tri- the Trichomycterinae are either synapomorphies dentinae clade (including Trichomycterus hase- for the whole family (i.e., symplesiomorphies for mani and T. johnsoni) and inside Vandelliinae the subfamily) or features of uncertain polarity. (Paravandellia).

' Doctoral Student, Department of Herpetology and Ichthyology, American Museum of Natural History and City University of New York.

INTRODUCTION The subfamily Sarcoglanidinae was estab- tential value as indicators ofmonophyly. This lished by Myers and Weitzman (1966) to in- analysis showed that there is no known char- clude two new genera and species, Sarco- acter that conflicts with the inclusion ofStau- glanis simplex and Malacoglanis gelatinosus. roglanis within the Sarcoglanidinae. None of The new subfamily was characterized, among the characters previously used to diagnose other things, by the toothless upper jaw, the the Trichomycterinae can be considered as presence of an adiposelike fin, and the pec- apomorphic at the level of this intended as- toral-fin rays projecting beyond the fin mem- semblage, thus, no evidence supports mono- brane. The description of S. simplex was phyly of the subfamily. On the other hand, based on a single specimen from the upper the derived characters shared by Stauroglanis Rio Negro below Sao Gabriel rapids (Ama- and sarcoglanidines are unique to them, and zon Basin, Brazil), whereas M. gelatinosus represent unambiguous synapomorphies in- was known from two examples collected in dicating that they form a monophyletic group. a small side channel of the Rio Orteguaza Although evidence supporting the pro- (tributary ofthe Rio Caqueta, Amazon Basin, posed relationship of Stauroglanis with the Colombia). The Sarcoglanidinae have to date remaining Sarcoglanidinae was not scarce, the been known only from the three specimens analysis was limited by the lack of access to originally studied by Myers and Weitzman material ofS. simplex and M. gelatinosus. As (1966), and by two additional examples of noted above, known material ofthe Sarcogla- Malacoglanis reported by Stewart et al. nidinae consists only of the holotype of S. (1987). Myers and Weitzman (1966) sug- simplex and the holotype and single paratype gested that, in spite of their scarcity in col- (previously cleared and stained) ofM. gelati- lections, it is probable that the Sarcoglanidi- nosus, all ofthem deposited in the California nae occur throughout the entire Amazon Academy of Sciences. This material was Basin, as well as in the Orinoco Basin. loaned several years ago, subsequent to its The present paper is based on material col- use by Myers and Weitzman, and never re- lected by Dr. Michael Goulding in the Rio turned. Since I was unable to examine these Negro (Amazon Basin), and deposited in the specimens for the present paper, compari- fish collection of the Museu de Zoologia da sons relative to them were limited to the data Universidade de Sao Paulo, Sao Paulo, Bra- in the original description (Myers and Weitz- zil. Two specimens included within this ma- man, 1966), to the osteological sketches pro- terial, from separate localities, represent a vided in Baskin (1973, unpubl. thesis) ofthe distinctive new taxon so far completely un- single cleared and stained paratype ofM. ge- known, here proposed as a new genus and latinosus, and to some osteological drawings species, Stauroglanis gouldingi. of the same specimen made for me by Dr. The relationships ofthe new taxon are not Scott Schaefer. immediately evident. Characters traditionally used to delimit subgroups of Tricho- mycteridae place it in the subfamily Tricho- AcKNowLEDGmENTs mycterinae. Nevertheless, a comparative A previous version of this work was a dis- anatomical survey ofStauroglanis and other sertation submitted in partial fulfillment ofa trichomycterids revealed additional charac- bachelor's degree in biology (major in zool- ters in disagreement with that initial subfa- ogy), at the Universidade Federal do Rio de milial placement. The apomorphies relevant Janeiro, Brazil. The final version of the for intrafamilial relationships indicate that manuscript was prepared at the Department the new taxon is mostly closely related to the of Ichthyology of the American Museum of Sarcoglanidinae, rather than to any Tricho- Natural History. Support by both institutions mycterinae, as this subfamily is presently to the author is gratefully acknowledged. I constituted. This situation led to an exami- further wish to thank Dr. Heraldo Britski, the nation of all characters so far used to define advisor ofthis study, for his continuous help the Trichomycterinae, in terms of their po- and encouragement, and Prof. Mario Brum 1989 de PINNA: SARCOGLANIDINE CATFISH 3 and Drs. Naercio Menezes and Jose Figuei- surements followed Tchernavin (1944: 251- redo for valuable advice. I owe much, in both 252). Dorsal- and anal-fin ray counts for academic and personal grounds, to Drs. Erica Stauroglanis included all rays and splints. Caramaschi, Ricardo Castro, and Wilson Dorsal- and anal-fin ray counts for taxa other Costa, and I am pleased for their friendship. than Stauroglanis included two anterior un- Dr. Carl Ferraris shared with me his knowl- branched rays plus all following rays. The two edge of catfish anatomy and systematics, in posteriormost closely positioned rays, when many enjoyable conversations. Dr. Gareth present, were counted separately. Principal Nelson has been most helpful in various as- caudal-fin ray counts included all branched pects ofmy permanence at the American Mu- rays plus one unbranched ray in each lobe. seum. I am also indebted to Dr. Gloria Ar- Counts are given for each lobe (upper first), ratia for her attention and interest in my work. separated by a plus sign. All values in paren- Dr. Silvina Menu-Marque kindly provided theses are those of the paratype. specimens of Trichomycterus boylei for study. Osteological preparations were cleared and Dr. Scott Schaefer sent useful anatomical in- counterstained for cartilage and bone, ac- formation on the cleared and stained para- cording to a modified version of the method type ofMalacoglanis gelatinosus. My special of Dingerkus and Uhler (1977). A consider- recognition is extended to Dr. Michael able part of the osteological material exam- Goulding, for depositing his material in the ined was prepared prior to this study, and collections ofthe MZUSP, where Staurogla- part of it is stained only for bone. nis ultimately became available for study. Drawings were sketched with the aid ofan Various versions of the manuscript benefit- Aus Jena stereomicroscope and camera lu- ted from reviews by Drs. Gloria Arratia, Her- cida. Data checking and addition of details aldo Britski, Erica Caramaschi, Ricardo Cas- were accomplished by means ofan Olympus tro, Wilson Costa, Carl Ferraris, John stereomicroscope at greater magnifications. Lundberg, Naercio Menezes, Gareth Nelson, In all osteological illustrations (except in dia- Scott Schaefer, Richard Vari, and Stanley grammatic drawings), bony tissue is repre- Weitzman. sented by stipple and cartilage is represented Financial support was provided by the by open circles. Conselho Nacional de Desenvolvimento Only two specimens ofStauroglanis gould- Cientifico e Tecnologico (CNPq), Brazilian ingi are known, with the one in better con- Federal Government. Travel support was also dition chosen as holotype. The paratype, in provided by Funda-aco de Amparo a Pesquisa a very poor state of preservation, was made no Estado do Rio de Janeiro (FAPERJ), Gov- into an osteological preparation. In spite of ernment ofthe State ofRio de Janeiro. Prep- its fragile condition, the paratype cleared well aration ofthe final version ofthe manuscript and stained satisfactorily for both bone and was also financially supported by a grant cartilage. Its bones are, however, more loose- (Graduate-Undergraduate Research Pro- ly connected and the cleared tissues less firm gram) by the American Museum of Natural than is usual for well-preserved fishes sub- History. sequently cleared and stained. Morphometric data for the paratype were taken prior to the MATERIALS AND METHODS clearing and staining of the specimen. All measurements were point-to-point, taken with calipers on the left side of speci- ANATOMICAL ABBREVIATIONS mens. The caudal peduncle depth was mea- USED IN FIGURES sured at the vertical through the insertion of principal caudal-fin rays. The caudal pedun- b basioccipital cle length was from the base of the last anal- ba basibranchials cl-5 ceratobranchials 1 to 5 fin ray to the middle of the caudal fin base. e epioccipital Lengths ofthe dorsal- and anal-fin bases were epibranchials 1 to 4 measured from the first visible ray to the point el lateral ethmoid of posterior attachment. Remaining mea- ex exoccipital 4 AMERICAN MUSEUM NOVITATES NO. 2950 f frontal mens given (ex.) refers to the number ofspec- h hyomandibula imens examined, not to the total number of hb1-3 hypobranchials 1 to 3 specimens available in the respective lot. Un- hyl, hypurals 1 to 5 less otherwise noted, specimens are undis- i interopercle I lacrimal sected, alcohol preserved. The symbol CS in- m maxilla dicates that the specimens were available as me mesethmoid cleared and stained preparations; CS (b & c) mt metapterygoid refers to material stained for both cartilage o orbitosphenoid and bone, CS (b) refers to material stained op opercle only for bone. CS (b & c) indicates that at p pterotic least one of the specimens is stained for car- p3-4 pharyngobranchials 3 and 4 tilage, not necessarily the whole sample. Gen- pa parasphenoid era and species are listed alphabetically with- pal palatine in each suprageneric taxon. ph parhypural pm premaxilla Abbreviations for institutions are: Amer- Po preopercle ican Museum ofNatural History, New York pr prootic (AMNH); Academy of Natural Sciences of ps pterosphenoid Philadelphia, Philadelphia (ANSP); Califor- pu,+u, preural centrum 1 plus ural centrum 1 nia Academy of Sciences, San Francisco q quadrate (CAS); Museu de Zoologia da Universidade s sphenotic de Sao Paulo, Sao Paulo (MZUSP); Museu sc supracleithrum (posttemporosupraclei- Nacional, Rio de Janeiro (MNRJ); National thrum?) Museum of Natural History, Washington, so supraorbital (antorbital?, frontolacrimal D.C. (USNM); Museu de Ciencias da Pon- tendon bone?) su supraoccipital tificia Universidade Catolica do Rio Grande tp toothplate of pharyngobranchial 4 do Sul, Porto Alegre (MCP); Museu Anchie- u uroneural ta, Porto Alegre (MAPA -collections now ub unnamed submaxillary bone under the care of researchers working at v vomer MCP). w Weberian complex and capsule Additional comparative data for taxa of the subfamily Trichomycterinae were ob- tained from Arratia et al. (1978), Arratia and MATERiAL EXAMINED Menu-Marque (1981), Arratia and Menu- Comparative material examined for this Marque (1984), Arratia (1983), and Arratia study is listed below. The number of speci- and Chang (1975).

Trichomycteridae Trichomycterinae Eremophilus mutisii MZUSP 35409 1 ex. CS (b & c) Hatcheria macraei MZUSP 35687 1 ex. CS (b) Scleronema minutum MCP 9315 2 ex. CS (b & c) Scleronema sp. MAPA 2409 1 ex. CS (b & c) Scleronema sp. MAPA 1864 1 ex. CS (b & c) Trichomycterus amazonicus MZUSP 37763 1 ex. CS (b & c) Trichomycterus brasiliensis MZUSP 37145 2 ex. CS (b & c) Trichomycterus boylei MZUSP uncat. 2 ex. CS (b) Trichomycterusflorensis MZUSP uncat. 2 ex. CS (b) Trichomycterus hasemani MZUSP uncat. 5 ex. CS (b) Trichomycterus johnsoni MZUSP uncat. 1 ex. Trichomycterus quechuorum AMNH 20351 1 ex. CS (b) Trichomycterus reinhardti MZUSP uncat. 2 ex. CS (b) Trichomycterus tiraque AMNH 39740 2 ex. CS (b) Trichomycterus vermiculatus MZUSP uncat. 2 ex. CS (b) Trichomycterus sp. MZUSP 36966 1 ex. CS (b) 1989 de PINNA: SARCOGLANIDINE CATFISH 5

Glanapteryginae Listrura nematopteryx MZUSP 37138 5 ex. CS (b & c) Glanapteryx anguilla MZUSP 36530 2 ex. CS (b & c) Pygidianops eigenmanni CAS 11121 1 ex. CS (b & c) Typhlobelus ternetzi CAS 11119 1 ex. CS (b) Vandelliinae Paracanthopoma sp. MZUSP 29524 1 ex. CS (b) Paravandellia bertonji MZUSP 27707 1 ex. CS (b) Paravandellia sp. MZUSP uncat. 2 ex. CS (b & c) Vandellia sp. MZUSP 29155 2 ex. CS (b) Tridentinae Tridensimilis sp. MZUSP 36302 2 ex. CS (c & b) Tridentopsis sp. MZUSP 36303 2 ex. CS (b & c) Stegophilinae Haemomaster venezuelae AMNH 10194 1 ex. CS (b) Henonemus sp. MZUSP uncat. 1 ex. CS (b) Homodiaetus anisitsi MCP uncat. 2 ex. CS (b) Ochmacanthus alternus MZUSP 30467 2 ex. CS (b) Ochmacanthus orinoco MZUSP 30473 1 ex. CS (b) Ochmacanthus reinhardti AMNH 27693 1 ex. CS (b) Parastegophilus maculatus MZUSP 35736 1 ex. CS (b) Pseudostegophilus nemurus MZUSP 30427 2 ex. CS (b) Trichogeninae Trichogenes longipinnis MZUSP uncat. 6 ex. CS (b & c) Nematogenyidae Nematogenys inermis MZUSP 36956 1 ex. CS (b & c) AMNH 55329 1 ex. CS (b & c) Callichthyidae Brochis coeruleus AMNH 43411 2 ex. CS (b & c) Callichthys callichthys AMNH uncat. 3 ex. CS (b) Corydoras aeneus AMNH 21772 1 ex. CS (b) Hoplosternum punctatum AMNH 11580 4 ex. CS (b) Scoloplacidae Scoloplax dicra AMNH 10204 1 ex. CS (b) Scoloplax sp. MZUSP 37489 2 ex. CS (b & c) Astroblepidae Astroblepus grixalvii USNM 167876 2 ex. CS (b & c) Astroblepus sp. AMNH 20873 1 ex. CS (b) Loricariidae Farlowella sp. AMNH 21813 1 ex. CS (b & c) Hypoptopoma carinatum AMNH 39828 2 ex. CS (b & c) Hypostomus sp. AMNH 43453 1 ex. CS (b & c) Loricaria cataphracta AMNH 40081 1 ex. CS (b & c) Pterygoplichthys sp. AMNH uncat. 3 ex. CS (b & c) Diplomystidae Diplomystes chilensis AMNH 55318 1 ex. CS (b & c) Diplomystes sp. MZUSP 36953 1 ex. CS (b & c) Pimelodidae Chasmocranus rosae ANSP 137968 2 ex. CS (b & c) Microglanis sp. AMNH 20878 1 ex. CS (b & c) Rhamdia hypselurus AMNH 24870 2 ex. CS (b) 6 AMERICAN MUSEUM NOVITATES NO. 2950

Pimelodus pictus AMNH uncat. 2 ex. CS (b & c) Sorubim lima AMNH 42128 1 ex. CS (b) Auchenipteridae Asterophysus batrachus ANSP 158294 1 ex. CS (b & c) Auchenipterus demerarae AMNH 55352 2 ex. CS (b & c) Trachycorystes fisheri AMNH 5332 1 ex. CS (b) Aspredinidae Aspredinichthys tibicen AMNH 7094 1 ex. CS (b) Bunocephalus coracoideus AMNH 21815 1 ex. CS (b & c) Hypophthalmidae Hypophthalmus edentatus AMNH 55396 2 ex. CS (b & c) Helogenidae Helogenes marmoratus AMNH 7113 1 ex. CS (b) AMNH 13332 1 ex. CS (b) Ariidae Genidens genidens AMNH 20725 2 ex. CS (b) Bagridae Bagrus bayad USNM 229884 1 ex. CS (b & c) Chrysichthys auratus USNM 229728 1 ex. CS (b & c) Pseudobagrusfluvidraco AMNH 10438 2 ex. CS (b & c) Ictaluridae Ameiurus nebulosus AMNH 37682 1 ex. CS (b) Amblycipitidae Amblyceps mangois ANSP 59316 1 ex. CS (b & c) Siluridae Kryptopterus bicirrhis AMNH 43171 I ex. CS (b & c) Sisoridae Glyptosternon sinense AMNH 10265 2 ex. CS (b & c) Exostoma kishinouyei AMNH 15261 1 ex. CS (b & c) Amphiliidae Amphilius pictus AMNH 12314 1 ex. CS (b) Phractura scaphirhynchura AMNH 6622 2 ex. CS (b)

SYSTEMATIC ACCOUNTS by supraoccipital, sphenotic, and frontal; (5) vomer roughly rhombus shaped, with deep Stauroglanis, new genus necklike constriction at point of articulation TYPE SPECIES: Stauroglanis gouldingi, n. sp. with palatine; (6) single ossified basibranchial DIAGNOSIS: A member of subfamily Sar- element, crucifix shaped, with two lateral ar- coglanidinae, of the Trichomycteridae. The ticular processes. None of these characters following features distinguish Stauroglanis has been checked in Sarcoglanis simplex from other taxa of the family: (1) premaxilla (known only from the holotype), and char- with narrow lateral process, extending par- acters 2, 3, and 6 have also not been checked allel to maxilla, ofabout same length as basal in Malacoglanis gelatinosus. All six charac- portion of premaxilla; (2) premaxilla with ters are otherwise unique to Stauroglanis ventral flat expansion near distal limit ofpre- within trichomycterids. Other characters use- maxillary teeth distribution; (3) small irreg- ful for the identification and systematic ular-shaped bone below posterior articular placement of Stauroglanis include: maxilla process of maxilla; (4) two paired dorsal greatly enlarged, longer than premaxilla; car- openings on neurocranium, each one framed tilaginous head ofpalatine with thick anterior 1989 de PINNA: SARCOGLANIDINE CATFISH 7

Fig. 1. Lateral view of Stauroglanis gouldingi, n. gen. et sp., holotype, MZUSP 31088. Scale = 5 mm. ossification; quadrate with very long and wide DIAGNOSIS: See generic diagnosis. anterodorsal process, distally directed back- DESCRIPTION: Morphometric data for the ward; metapterygoid very reduced in size, holotype and paratype given in table 1. partly surrounded by anterodorsal process of Body elongate, roughly rounded in cross quadrate; interopercle reduced, with very section in trunk region and gradually more narrow odontode attachment facet and only compressed toward caudal region. Dorsal about three odontodes; opercle with only two surface of body flat along its whole length. large odontodes; palatine anteroposteriorly Dorsal body profile nearly straight, gently elongate; lacrimal elongate; number of teeth convex along its middle third. Ventral body on premaxilla reduced (about six); size ofanal profile concave shortly after insertion ofpec- fin reduced (six rays); eyes large and con- torals, slightly convex along abdominal re- spicuous; nasal barbel present but very short; gion, and straight from pelvic-fin insertion to snout region of head elongate; mouth sub- base ofcaudal fin. Body deepest at about level terminal, not suckerlike; all teeth conical, ofpelvic-fin insertion, gradually less deep to- without evident specializations in shape; no ward caudal fin. Ventral profile ofcaudal pe- lateral saclike adipose organ dorsoposterior duncle straight, dorsal profile very slightly to pectoral fin insertion. convex along its anterior half. Dorsal and ETYMOLOGY: From the Greek stauros ventral profiles of caudal peduncle slightly (cross) and glanis (the catfish ofAristotle), in allusion to the crucifix shape of the single TABLE 1 ossified basibranchial element. Gender mas- Morphometric Data for Holotype (A) and Para- culine. type (B) of Stauroglanis gouldingi, n. gen et sp. INCLUDED SPECIES: Only one known species, (All measurements expressed in mm) Stauroglanis gouldingi, n. sp. A B Standard length 23.3 23.8 Stauroglanis gouldingi, new species Total length 27.3 26.8 (Figures 1-3) Head length 4.2 4.7 HOLOTYPE: MZUSP 31088, 23.3 mm SL, Body depth 2.7 3.0 Brazil, Estado do Amazonas, Rio Daraa (Rio Caudal peduncle length 6.4 6.2 Caudal peduncle depth 1.4 1.5 Negro drainage system), Cachoeira do Aracu, Predorsal length 13.6 13.9 near waterfall; coll. Michael Goulding, 10 Feb. Preanal length 16.1 17.0 1980. Head width 3.2 3.0 PARATYPE: MZUSP 30411, 23.8 mm SL, Head depth 1.8 1.6 Brazil, Estado do Amazonas, Rio Negro, Is- Interorbital 0.5 0.8 land of Cumuru, near Rio Arirara; lake on Snout length 2.1 2.3 island, muddy beach; coll. Michael Gould- Dorsal fin base length 2.4 2.2 ing, 1 Feb. 1980, 14:00 hr. Specimen stored Anal fin base length 1.3 1.2 in glycerol, partly dissected, cleared, and Prepelvic length 12.2 11.6 stained for cartilage and bone. Breadth of mouth 1.1 1.6 8 AMERICAN MUSEUM NOVITATES NO. 2950

Fig. 2. Dorsal view ofthe head ofStauroglanis Fig. 3. Ventral view of the head of Stauro- gouldingi, n. gen. et sp., holotype, MZUSP 31088. glanis gouldingi, n. gen. et sp., holotype, MZUSP Scale = 1 mm. 31088. Scale = 1 mm. and evenly converging toward caudal fin. No iously sized globular bodies, irregularly dis- dorsal or ventral expansion on caudal pe- tributed in apparently more than one layer. duncle due to presence of procurrent rays. Globular bodies with appearance ofoil drops Dorsal profile ofhead flat, nearly continuous merged in water, but quite rigid to contact. in straight line with dorsal profile of body. Band's outer surface rough and irregular, Ventral profile ofhead convex from lower lip matching contours of numerous individual to gular region (apparently due to expanded globular bodies. position of hyoid arch at time of fixation), Integument very thin and transparent, su- then horizontally straight along short dis- perficial muscles, and some other internal tance continuous with ventral trunk region. structures readily visible. Papillae not visible Myotomes conspicuous, readily visible on skin, even on lips and barbels of holotype, along whole body, more prominent along re- but visible in paratype as tiny alcian-blue gion from anal-fin base to middle length of stained dots over entire skin. caudal peduncle. Myotomes progressively Head elongate, distinctly narrowing ante- narrower and more angled toward caudal fin. riorly, triangle shaped in dorsal view and flat. Deep horizontal skeletogenous septum sep- Rostral part ofhead very long. Anterior mar- arating epaxial and hypaxial muscle masses, gin ofsnout rounded and convex. Mouth sub- deepest at about same region where myo- terminal and narrow, lips not notably fleshy. tomes are most prominent. Ventral region Branchial membranes very thin and trans- with easily visible muscle segments, more lucent, attached to isthmus only at anterior- conspicuous anteriorly. most mesial point, gill openings unconstrict- Long lateral band of apparently adipose ed. Five to six branchiostegal rays seen tissue running from region ofpectoral-fin at- without difficulty through branchial mem- tachment to posterior margin ofanal-fin base, branes. Eyes very large and conspicuous, dor- becoming narrower posteriorly. Band slightly sally located. Eyeball entirely visible through hyaline, composed of large number of var- skin, with laterosuperior orientation. Eyes lo- 1989 de PINNA: SARCOGLANIDINE CATFISH 9 cated almost completely on posterior half of Fins very damaged in preservation, origi- head length. Transparent areas below skin nal shape difficult to determine accurately, just in front ofeach eye, apparently filled with interradial membrane mostly destroyed. Re- gelatinous matter, merging gradually with re- constructed approximate outlines (from rel- mainder of head surface. Opercular patch of ative lengths of rays) shown in figure 1. Pec- odontodes small but visible in both lateral toral fin moderately elongate, approximately and dorsal views, partly covering pectoral fin of same length as head. First pectoral ray insertion. Interopercular patch of odontodes (unbranched) not prolonged as filament, with located below each eye, extremely small and same general appearance as following hyaline, inconspicuous on head. No fold of branched rays. Pectoral-fin rays 8 (7). Origin skin underlying interopercular patch ofodon- of dorsal fin on posterior half of standard todes in holotype, very small fold present in length, behind insertion of pelvic fins. Dor- paratype. Interopercular patch of odontodes sal-fin rays 8 (8). Anal-fin origin approxi- located only slightly below horizontal through mately below posterior halfofdorsal fin base. opercular patch. Anal-fin base short, much shorter than dor- Bony base of maxillary barbel very large, sal-fin base. Anal-fin rays 6 (6). Pelvic fins but more slender than in Sarcoglanis, Mala- extending beyond anal and urogenital open- coglanis, and Scleronema. Distance between ings, but not reaching origin of anal fin. Pel- tips of bony portions of barbels approxi- vic-fin rays 5 (5). Pelvic splint absent. All fins mately equal to head width, when maxillae with some branched rays (exact proportion completely protruded. Soft portion of max- ofbranched and unbranched rays impossible illary barbel very slender, about ofsame length to verify due to their poor condition). Prin- as bony section. Rictal barbel of approxi- cipal caudal rays 6 + 6 (6 + 6). Procurrent mately same length and diameter as soft por- caudal rays small, inconspicuous, and few in tion of maxillary barbel. Origin ofrictal bar- number; 4 dorsal and 4 ventral (data from bel under anterior part of bony base of paratype). maxillary barbel. Maxillary and rictal barbels Pigmentation in preservative: Body almost with slender central rods. Nasal barbel very totally white, dark pigment restricted to eyes short and thin, hardly visible from above, its and some isolated melanophores, forming origin at rim ofanterior nare. Nares tiny and very ill-defined pigmentation pattern. Indi- difficult to see, due to combination ofreduced vidual melanophores very dark and quite size and general transparency of tissues. An- large, of variable sizes. A group of melano- terior nare surrounded by fold of skin, con- phores scattered over neurocranium, over tinuous posterolaterally with nasal barbel. area corresponding to supraoccipital and bas- Posterior nares extremely reduced and al- al portion offrontals; another on area ofoper- most invisible, their openings oriented back- cle; one on area of interopercle, and another ward. Distance between anterior nares con- scattered on mesial region of snout (figs. 1 siderably greater than between posterior pair, and 2). Mesial ventral region from beginning and greater than interorbital width. of anal fin to last portion of caudal peduncle Lateral line short (less than half of head with narrow sparse scattering of melano- length) but very conspicuous externally, its phores. Caudal fin with scattered dark chro- shape strongly outlined by covering skin. Ax- matophores at basal portion. Remaining fins illary organ well developed and translucent, transparent, devoid of dark pigment. running just below lateral line and slightly DISTRIBUTION: Known only from the lo- longer than it. Origin of axillary organ im- calities cited for the holotype and paratype. mediately above insertion of pectoral fin, its ETYMOLOGY: The specific name honors Dr. length less than half that of the fin. Michael Goulding, who collected the type Small rayless cutaneous folds running specimens and all known material of Stau- above and below part of caudal peduncle. roglanis, in recognition of his contributions Upper and lower folds with irregular margins to the knowledge of the Amazonian fish fau- and located on different regions ofcaudal pe- na. duncle. Folds extremely thin and low, very REMARKS: From data provided in the orig- transparent and almost invisible under ste- inal description and illustration of Tricho- reomicroscope. No predorsal fold. mycterus santaeritae (Eigenmann, 1918), it is 10 AMERICAN MUSEUM NOVITATES NO. 2950

Su SC SC

A B Fig. 4. Skull of Stauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 30411. A, dorsal view; B, ventral view. Scale = 1 mm. possible that there is a close relationship be- Phylogenetic Relationships of Stauroglanis tween this species and Stauroglanis gouldingi. and Relationships ofthe Sarcoglanidinae with While the two taxa are clearly distinct, judg- Other Trichomycteridae). Another notewor- ing from differences in color pattern, fin-ray thy character present in T. santaeritae is the counts, and number of odontodes, there are very large eye, which once again recalls the some notable resemblances that deserve no- state in Stauroglanis. tice. Eigenmann (1918: 341) described the These characteristics become particularly teeth of T. santaeritae as "in a single series interesting when it is recognized that there or in a very narrow band." This feature con- are no derived characters supporting the trasts with the condition in most species of present generic and subfamilial placement of the genus Trichomycterus, and agrees with T. santaeritae (see section below on the phy- that observed in the dentary and premaxilla letic status of the subfamily Trichomycteri- of Stauroglanis (see Osteology below). nae). Thus, phylogenetic relationships of T. Another remarkable feature is the length of santaeritae, as well as of most other tricho- the nasal barbels, which are very short in both mycterine species, are a completely open S. gouldingi and T. santaeritae. In the great question. majority of Trichomycterus species the nasal Unfortunately, T. santaeritae is so far barbel is much longer than in those two taxa. known only from the holotype, and it was Although not mentioned in the description, not possible to examine the internal char- the bony base of the maxillary barbel illus- acters necessary for an adequate comparison trated for T. santaeritae seems to be very with Stauroglanis. It is thus preferable, for elongate, similar to the condition in Stau- the time being, to maintain T. santaeritae in roglanis and its closest relatives (see sections the genus Trichomycterus. The traits pointed 1989 de PINNA: SARCOGLANIDINE CATFISH I1I

out above may serve as starting points for additional research, but are too scanty to use as a basis for a hypothesis ofrelationship and its consequent nomenclatural alterations. OSTEOLOGY The aim of this section is to provide an account focusing on osteological traits ofsome importance for discussions of relationships, and not an exhaustive description ofthe whole skeleton ofStauroglanis. Consequently, some bones and bony complexes in which no in- formative variation could be detected are omitted or only very briefly mentioned. All internal osteological data are based on the single cleared and stained paratype of Stau- roglanis gouldingi. NEUROCRANIUM The general shape of the neurocranium is shown in figure 4A and B. The mesethmoid 5. Vomer of n. is tilted downward, and the mesethmoid cor- Fig. Stauroglanisgouldingi, gen et sp., paratype, MZUSP 3041 1. Scale = 1 mm. nua are consequently directed ventrad. The frontals are small and slender, not contacting each other along the midline. They are closest otic. The openings are irregular and some- together immediately anterior to the cranial what bilaterally asymmetrical, not closed by fontanel (but do not suture), and get pro- a cartilaginous layer. The single cranial fon- gressively more separate anteriorly. The space tanel is wide and not constricted at any point between the frontals is closed by the under- of its length. Its anterior extremity is bluntly lying posterior portion of the mesethmoid, pointed and the posterior end is rounded. which extends considerably posteriorly. The The lateral ethmoids are very narrow, pro- frontal bears a laterosensory canal segment jecting only slightly beyond the lateral margin along the margin of its posterior half, with of the frontals and inconspicuous in dorsal one pore opening at the anterior extremity view. The pterotic has a conspicuous lateral and a second immediately anterior to the cra- process, which apparently serves as a base for nial fontanel. In dorsal view, the sphenotic the lateral branch of the sensory canal of the has a distinctive and almost right-angled con- pterotic (the process is damaged on the right cavity anteriorly. The anterior canal-bearing side of the cleared and stained paratype). portion of the sphenotic is strongly curved The vomer (fig. 5) has a peculiar rhom- laterally. The very short infraorbital sensory boidal shape, with a necklike constriction di- canal exits at the sphenotic-frontal limit, viding it into smaller anterior and larger pos- opening as a pore locatedjust posterior to the terior portions. Its broadest region is its eye. The supraoccipital is narrow anteriorly posterior part, between the two lateral arms, and the posterior part of the cranial fontanel whose tips are directed backward. The neck- extends along more than half of its length. like constriction ofthe vomer corresponds to The anterior face of the supraoccipital does the site ofarticulation with the palatine. The not contact the posterior margins ofthe fron- parasphenoid is narrow anteriorly and wide tals. posteriorly; its anterior extremity bifurcates There is a very characteristic dorsal open- around the posterior end of the vomer. Pos- ing on each side ofthe neurocranium, formed teriorly, the parasphenoid extends to the an- by concavities in the margins of the frontal, terior third ofthe basioccipital. The posterior supraoccipital, and in particular the sphen- wide portion of the parasphenoid is a very 12 AMERICAN MUSEUM NOVITATES NO. 2950

)sf~ 4,,,.:~~~~~~~7

Fig. 6. Partial lateral view of the skull of Stauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 3041 1. Laterosuperior view, anterior to left. Scale = 1 mm. thin sheet, with the borders difficult to dis- pterosphenoid portion ofthe compound bone cern against the underlying bones. The proot- can be identified as a very narrow ossifica- ic, sphenotic, and pterosphenoid are fused tion, running below and partly hidden by the into a single ossification (as in all trichomyc- canal-bearing border of the posterior part of terids except Trichogenes longipinnis). The the frontal (fig. 6). Although very narrow, the

pal

Fig. 7. Upperjaw and associated structures ofStauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 3041 1. Dorsal view. Scale = 1 mm. 1989 de PINNA: SARCOGLANIDINE CATFISH 13 pterosphenoid portion still has a small but definite anterior cartilaginous facet in its contact with the orbitosphenoid. There are three large lateral openings on OOOK the side of the neurocranium (fig. 6). The posteriormost and largest foramen (trigemi- nofacialis) is irregular, bordered dorsally by the portion of the compound bone corre- Fig. 8. Right premaxilla ofStauroglanis gould- sponding to the pterosphenoid, anteriorly by ingi, n. gen. et sp., paratype, MZUSP 3041 1. An- the orbitosphenoid, and ventroposteriorly by teroventral view. The arrow indicates the premaxillary ventral flat expansion. Scale = 1 mm. the sphenotic-prootic portion of the compound bone. The second foramen (optic) is roughly rounded and is totally bordered by The maxilla is very large, its length almost the orbitosphenoid. The anteriormost open- twice that of the premaxilla (including the ing is elongate in shape, bordered dorsally by lateral process). The proximal extremity of the frontal and ventrally by the orbitosphe- the maxilla articulates closely with the an- noid; it is probably not a foramen, but an terior piece of the palatine, and is ligamen- incomplete union between the orbitosphe- tously connected to the posterior face of the noid and the frontal. premaxilla. At about the same level as the ventral flat expansion ofthe premaxilla, there is a small and low frontoventral flat process JAWS AND RELATED STRUCTURES (figs. 7 and 8) on the maxilla. These flat premaxillary and maxillary processes are similar and lie in par- The premaxilla bears a long, toothless lat- allel planes, but do not articulate closely. The eral process. The process is slightly higher posterior process of the maxilla is propor- than wide and extends along and parallel to tionally enlarged, and isjoined ligamentously the proximal portion of the maxilla. A short to the lower jaw. Below this process there is and stout dorsal process on the premaxilla a small isolated ossificiation not homologous articulates with the neck of the mesethmoid, to any known bone. The surface ofthis piece in the region just posterior to the junction of is rough and its shape is irregular, but it is the ethmoid comua. The premaxillary pro- solid and completely separate from the sur- cess is small and short but well differentiated, rounding bones, being originally present on its dorsal tip is rounded. The premaxillary both sides (it is now lost in the right side of teeth are simple and conical. Teeth insertion the paratype). sites indicate that each premaxilla originally The palatine is elongate and most remark- had between 14 and 17 teeth, disposed in two able in being composed of two pieces. The irregular rows. Teeth insertion sites cover the larger posterior piece is apparently the au- entire margin of the body of the premaxilla topalatine itself, homologous with the auto- (the portion apart from the lateral process). palatine of remaining trichomycterids and Many premaxillary teeth seem to have been other catfishes. The anterior piece seems an lost post mortem, and only between 6 and 9 anterior ossification ofthe cartilaginous head are still attached to each bone. The attach- that covers the anterior face of the palatine, ment of the teeth to the premaxilla is very which is apparently composed of a different loose, with tooth bases often uncalcified. A type ofossification, with a rough surface that small posterior process of the premaxilla ar- contrasts with the smooth texture ofthe pos- ticulates with the anterior piece of the pala- terior piece. The anterior piece is thick and tine. There is a conspicuous ventral flat ex- compact, with a narrow pointed anterior me- pansion on the premaxilla (fig. 8), at the point dial process that partly overlaps the proximal where the lateral process meets the body of end of the maxilla. A thick layer of cartila- the bone. Its shape is roughly semicircular, ginous tissue remains between the anterior and it represents approximately the lateral and posterior palatine pieces, but the result- limit of the distribution of the premaxillary ing articulation seems to be completely im- teeth. movable. The palatine (posterior piece) ar- 14 AMERICAN MUSEUM NOVITATES NO. 2950

A B Fig. 9. Gill arches of Stauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 30411. A, ventral portion under dorsal view, only right paired elements shown. B, dorsal portion, dorsal view of right side. Scales = 1 mm. ticulates medially with the vomer and also two lateral processes articulate with the prox- with the lateral ethmoid. The anterior pala- imal cartilaginous portions ofthe second hy- tine piece articulates with the proximal por- pobranchials. The disposition and orientation of the maxilla and with the posterior tion ofthe two lateral processes do not match process of the premaxilla. perfectly, and the whole basibranchial bony The elongate lacrimal is more than halfthe element looks slightly asymmetrical. It is not length ofthe frontolacrimal tendon bone (the clear whether the single basibranchial bone homologies of this bone are unclear, maybe is a highly modified second basibranchial it is the supraorbital or the antorbital; for (with disappearance ofthe third basibranchi- some discussion see Baskin, 1973: 158; Brit- al) or if it is formed by the fusion between ski and Ortega, 1983: 215; and Arratia, 1987: the primitive second and third basibranchials 96). The lacrimal and the frontolacrimal ten- (the first basibranchial is absent in all cat- don bone are positioned in a straight line, fishes). No sign of a suture can be detected. separated by a small space (smaller than the Regardless, it seems certain that the second length of the lacrimal). The position of the basibranchial is involved in the formation of lacrimal is directly above the lateral limit be- the cross-shaped basibranchial element. The tween the two pieces of the palatine. posterior basibranchial element is wholly cartilaginous. Given its topological relationships to the fourth and fifth ceratobranchials, BRANCHIAL ARCHES it is inferred that it represents the fourth ba- (fig. 9A and B) sibranchial of other trichomycterids. This There are only two basibranchial elements. cartilaginous piece is narrower anteriorly and The anterior element is ossified, and has a has two concavities on each side of its wider very peculiar cross shape, with four articular posterior half, into which fit the proximal tips facets corresponding to the tips of a crucifix. of the fourth and fifth ceratobranchials. The anterior articular facet is the largest. The All three hypobranchials show some ossi- 1989 de PINNA: SARCOGLANIDINE CATFISH 15

h mt

Fig. 10. Suspensorium and opercular apparatus of Stauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 3041 1. Lateral view of left side. Scale = 1 mm. fication. The first hypobranchial is cartilagi- and joins anteriorly the first and second epi- nous only at the articular tips. The second branchials. The fourth pharyngobranchial is has a distal, anteriorly directed conic ossifi- a completely cartilaginous rod joined to the cation, but is largely cartilaginous. The third anterior part ofits enormous bony tooth plate. is similar in general shape to the second, but The fourth pharyngobranchial articulates an- its bony piece is shorter and wider and its teriorly with the third, and posteriorly with cartilaginous portion is larger. the third epibranchial. The fourth epibran- The five ceratobranchials are elongate. The chial articulates only with the tooth plate of first has a conspicuous expanded proximal the fourth pharyngobranchial, not with its portion. Ceratobranchials 2 to 5 are progres- cartilaginous portion. The tooth plate of sively shorter and wider. The fourth has two pharyngobranchial 4 is very large and much short pointed processes (uncinate processes), more conspicuous than the cartilaginous en- the anterior directed forward and the poste- dochrondral portion. Approximately 14 well- rior oriented posteromedially. The fifth bears developed teeth are distributed along the about five teeth disposed irregularly on a small whole ventral extension of the tooth plate of area on a concavity of its proximal portion. the fourth pharyngobranchial. These tooth Gill rakers are completely lacking. plate teeth are progressively larger posterior- There are four epibranchials, the first, sec- ly, where the tooth plate itself is also wider. ond and third are slender and respectively joined to the first, second, and third ceratobranchials. The fourth thicker, dorsally ex- SUSPENSORIUM AND OPERCULAR panded epibranchial articulates with the APPARATUS (fig. 10) fourth ceratobranchial, and is united to the The articular facet for the hyomandibula fifth ceratobranchial by an elongate gently with the neurocranium extends along the lat- curved cartilaginous rod. Epibranchials 1 to eral part ofthe sphenotic and a small anterior 3 are slender and each bears a well-defined portion of the pterotic. The hyomandibula uncinatelike process, which is anteriorly lo- has a fowardly curved conspicuous antero- cated on epibranchials 1 and 2, and poste- dorsal process, the anterior tip of which riorly placed on the third. Epibranchials 1 closely approaches the posterior tip of the and 2 have convergent distal (upper) tips. dorsal process ofthe quadrate, forming a large The third and fourth pharyngobranchials quadrate-hyomandibular fenestra. A very are present. The third is a simple ossified rod, strong, pointed process for muscle attach- 16 AMERICAN MUSEUM NOVITATES NO. 2950

PU1~~ ~ ~ ~ <+ ~ ~~~h P1 ph -~~~~~~~~~~-

Fig. 1 1. Caudal skeleton of Stauroglanis gouldingi, n. gen. et sp., paratype, MZUSP 3041 1. Lateral view of left side. Scale = 1 mm. ment on the hyomandibula is located on the of the palatine closely approaches the dorsal middle ventral region of the bone. This pro- surface (the region above the attachment of cess is obliquely disposed on the hyoman- the metapterygoid) of the dorsal process of dibula and directed markedly anteriorly. A the quadrate, but the two bones do not over- small roundish process, articulating with the lap. A noteworthy character is that the pal- opercle, is present on the middle of the ven- atine is ligamentously linked more closely to tral margin of the hyomandibula. the quadrate than to the metapterygoid. The quadrate is elongate, about the same The preopercle is an elongate bone, dor- length as the longitudinal axis ofthe hyoman- sally concave along its contact with the hyo- dibula. A well-defined, thick process extends mandibula and with a short and very thin lateroventrally from the anterior tip of the anteroventral extension closely approaching quadrate, articulating with the lower jaw. An the dorsal region the interopercle. No sensory enormously developed dorsal process ex- canal is associated with the preopercle. There tends from the anterior dorsal region of the are only two large odontodes at the posterior quadrate. This process is completely flat, process ofthe opercle. These are inserted close shows a plain surface, and projects frontward together and have very broad bases. In spite from the dorsal margin of the quadrate and of the reduced number of opercular odon- then gradually curves backward almost todes, the width ofthe odontode-bearing pos- reaching the tip ofthe anterior process ofthe terior process ofthe opercle is not remarkably hyomandibula. narrow. A small but well-defined concavity on the The interopercle is very reduced when anterodorsal margin of the dorsal process of compared to the opercle. Its posteroventral, the quadrate accommodates the tiny metap- odontode-bearing, process is extremely nar- terygoid. This bone is a simple oval disc, al- row and carries only three small and weak most surrounded by the dorsal process ofthe odontodes (one of each is only partly devel- quadrate. A small but well-defined cartilage oped). The interopercular odontodes are links the metapterygoid with the inner sur- much less developed than those on the oper- face of the concavity forming a synchondral cle and overlap the uppermost branchiostegal joint between the two bones. The synchon- ray. dral articulation between the hyomandibula and the quadrate is robust, provided with a very well-developed interarticular cartilage. CAUDAL SKELETON (fig. 1 1) There is no overlap of bone over the site of No individual hypural elements can be dis- the articulation, i.e., no sutural articulation cerned. The parhypural, hypural 1, and hy- between the two bones. The posterior process pural 2 are represented by a single lower hy- 1989 de PINNA: SARCOGLANIDINE CATFISH 17 pural plate, continuous with the compound fied and one cartilaginous). No end piece (stay) caudal centrum (puI + u1). Hypurals 3, 4, is present in either the anal- or in the dorsal- and 5 are also fused to compose a single upper fin supports. There are two proximal radials hypural plate, which contacts tightly but does in the pectoral girdle, both ossified, with the not fuse with the puI + u . The upper and posterior proximal radial much larger than lower hypural plates tightly contact each oth- the anterior. Distal pectoral-fin radials are er along their proximal limits, but no fusion not present as either bone or cartilage. or superficial ossification can be detected. Each dentary bears about nine loosely at- The profiles of the distal margins of the tached conical teeth, arranged in a single row. upper and lower hypural plates are curved in The nasal bone is absent. a peculiar way, not forming a single, even arch. The two margins turn anteriorly as they approach the midline, and they are most ON THE PAEDOMORPHIC strongly curved at their superior (upper plate) CONDITION OF STA UROGLANIS and inferior (lower plate) distal ends. The me- Some of the features of the skeleton of dial corners of the upper and lower hypural Stauroglanis gouldingi are similar to condi- plates are not at the same vertical. tions commonly found in juvenile catfishes. The uroneural is slender and straight (very The small number, loose attachment, and slightly turned downward distally). It reaches weak calcification of the premaxillary and the dorsal margin of the bony portion of the dentary teeth are commonly seen in young upper hypural plate, but does not cover the trichomycterids ofspecies which do not show distal cartilaginous layer. this feature as adults. The incomplete con- The neural arch of the compound caudal nection between some of the neurocranial centrum is open dorsally (incomplete) and bones of Stauroglanis, such as between the does not form any rudiment ofa neural spine. supraoccipital and frontals, between the two The epural is absent. frontals, or between the frontals and the or- The hypurapophysis is small, located bitosphenoid, are also typical for early on- slightly in advance and below the opening for togenetic stages. The same applies to the gen- the caudal artery. eral absence of sutural joints in the neurocranium. On the other hand, the neurocranial bones do not resemblejuvenile con- OTHER OSTEOLOGICAL CHARACTERS ditions in other aspects. Usually in young There are five abdominal or precaudal ver- stages the bones of the neurocranium lack tebrae (here counted as those without a hemal well-defined borders, with their outer regions spine, not including those of the Weberian gradually becoming a thin cartilaginous (or complex) and 29 caudal vertebrae (those with dermal connective tissue) sheet. In Stauro- a hemal spine, not including the compound glanis, in contrast, such bones have well-de- caudal centrum). The three anteriormost pre- fined borders, lacking the cartilaginous frame. caudal vertebrae bear pleural ribs that are Additionally, Stauroglanis has three ossified similar in appearance; the first pair is only hypobranchials and two ossified proximal slightly shorter and thicker than those fol- pectoral radials, which are not usual in early lowing. The rib-supporting parapophysis of ontogenetic stages of trichomycterids. These the first free vertebra is much larger than those features indicate that the specimens of Stau- of the others. roglanis studied are notjuveniles, despite their The lateral line canal is very short; its length small size (23.2-24.0 mm SL) and lack of is less than that ofa single vertebral centrum. clear evidence of gonadal development. The The canal has a narrow ring of bone sur- apparently juvenile condition of some skel- rounding its anterior end, but is otherwise etal traits ofStauroglanis are thus interpreted totally devoid of a bony covering. as paedomorphic features. The two other de- There are eight dorsal-fin pterygiophores, scribed sarcoglanidines are known from even with five distal radials (four ossified and one smaller examples, Malacoglanis gelatinosus cartilaginous). There are six anal-fin pteryg- 18.2-19.9 mm SL and Sarcoglanis simplex iophores, with three distal radials (two ossi- 19.6 mm SL. The paratype ofM. gelatinosus 18 AMERICAN MUSEUM NOVITATES NO. 2950

(18.2 mm SL) is reported by Myers and These two genera were not included in their Weitzman (1966: 282) as being "an adult fe- list, because their standard lengths exceeded male containing large eggs," supposedly ma- 26 mm. Weitzman and Vari called these cases ture eggs. Thus, at least one member of the "elongate miniatures," to mean forms that sarcoglanidine clade can be confidently said are clearly miniatures in all criteria but the to be sexually mature at a very reduced size. cutoffof26 mm SL. Taking these factors into This constitutes further support for the in- consideration, I think it is reasonable to con- terpretation that the juvenile state of some sider that Glanapteryx and Typhlobelus are of the characters in Stauroglanis and other miniatures, as well as Listrura (a glanapter- sarcoglanidines, as well as their small size, ygine still undescribed by the time Weitzman represents paedomorphic features. and Vari's paper went to press). Thus, the This situation is not uncommon among subfamily Glanapteryginae is composed ex- South American freshwater fishes. Weitzman clusively of miniature species. Stauroglanis, and Vari (1988) recently proposed that the as well as Sarcoglanis and Malacoglanis, fall phenomenon (called miniaturization) occurs below the 26 mm SL limit, and can be con- in at least 85 species, and has evolved on at sidered a miniature according to Weitzman least 34 occasions in the inland fish fauna of and Vari's criterion. Consequently, also the South America. Within neotropical catfishes, Sarcoglanidinae is exclusively composed of miniaturization has taken place in at least 12 miniatures. Baskin (1973) suggested that the separate lineages, pertaining to the families Glanapteryginae and Sarcoglanidinae com- Aspredinidae, Callichthyidae, Loricariidae, pose a monophyletic group. Although his evi- Pimelodidae, Scoloplacidae, and Trichomyc- dence for this hypothesis has been criticized teridae (Weitzman and Vari, 1988). Within elsewhere (Pinna, 1988), his proposal re- trichomycterids, miniature members are mains the only one available for intrafamilial known to occur in the subfamilies Tricho- placement of these two subfamilies. If the mycterinae, Glanapteryginae, Vandelliinae, hypothesis that glanapterygines and sarco- Tridentinae, and Sarcoglanidinae; Weitzman glanidines are each other's closest relatives is and Vari proposed that miniaturization correct, and given that both taxa are com- events evolved at least five times within the posed only of miniature species, it follows family. The authors made this estimate based that a single miniaturization event explains on information that the five subfamilies be- the distribution of paedomorphic features in long to separate and divergent clades within the two-subfamily clade. Another point re- the family. This estimate means that the lates to Trichomycterus santaeritae, listed by members of a given miniature lineage per- Weitzman and Vari as one of the miniature taining to one given subfamily are not the members ofthe subfamily Trichomycterinae. sister group of any other miniature lineage As mentioned above, there is some evidence included in another subfamily, thus requiring indicating that this species may be related to separate miniaturization events. Weitzman the Sarcoglanidinae. If further investigation and Vari's evaluation of the number of in- confirms this, then the miniature condition dependent miniaturizations is based on the of T. santaeritae and of all members of the phylogenetic relationships among the groups clade Glanapteryginae/Sarcoglanidinae is ex- under consideration. Some of the estimates plained by the same event. The other two about trichomycterid interrelationships are trichomycterine miniatures listed by Weitz- altered by conclusions arrived at in the pres- man and Vari (1988) are Trichomycterus ent work, and need further comment. Weitz- johnsoni and Trichomycterus hasemani, man and Vari considered Pygidianops the which are quite similar and probably each only miniature member ofthe Glanapterygi- other's closest relative. Scarcity of material nae, which does not reach 26 mm SL, the of T. johnsoni has prevented an investigation division between miniature and nonminia- of its internal anatomy, but T. hasemani ture species. Nevertheless, as they pointed seems to be more closely related to the out, Glanapteryx and Typhlobelus could also subfamily Tridentinae than to any Tricho- be consided miniatures, because oftheir pae- mycterinae (see discussion below on non- domorphic features and small head sizes. monophyly of Trichomycterinae). The Tri- 1989 de PINNA: SARCOGLANIDINE CATFISH 19 dentinae are also composed exclusively of glanis is completely unknown at present, the miniature species, since Tridens melanops following comments are based almost exclu- (not included in Weitzman and Vari's list) sively on the data available for Malacoglanis. can be considered an elongate miniature. Until further material of Sarcoglanis be- Thus, similar to the situation with T. san- comes available, we must assume that the taeritae and sarcoglanidines/glanapterygines, characters in Malacoglanis are representative miniaturization in T. hasemani (and presum- of the condition in Sarcoglanis. This as- ably also in T. johnsoni) is actually a result sumption does not imply that the state in of miniaturization at the whole tridentine both taxa are identical, but only that the de- clade. If all the above putative hypotheses, rived characters shared by Stauroglanis and which undoubtedly require additional cor- Malacoglanis are also shared in some form roboration, are correct, then Weitzman and by Sarcoglanis. This assumption presumes Vari's consideration of five independent that Sarcoglanis and Malacoglanis compose miniaturization events within the Tricho- a monophyletic group, with Stauroglanis as mycteridae is an overestimation, and three its sister taxon (see below). This working as- such events would seem to have occurred: at sumption seems appropriate, because Sar- the Glanapteryginae/Sarcoglanidinae clade coglanis and Malacoglanis share various ex- (including T. santaeritae); at the Tridentinae ternal unequivocal synapomorphies, which clade (including T. hasemani and T. john- make it unlikely that Stauroglanis is the sister soni); and inside the Vandelliinae (in Para- group to only one ofthem. In agreement with vandellia). Independence of events in van- this, no putatively derived character that delliines and tridentines derives from the could be checked in all three genera was found hypothesis of Baskin (1973), which places to be exclusive to Stauroglanis plus only one Vandelliinae (with only some miniature of the other two. members) as the sister group to Stegophilinae A separate ossification of the anterior car- (a subfamily without miniatures), and Tri- tilage of the palatine seems to be present in dentinae as the sister group to Stegophilinae Malacoglanis (fig. 12). This anterior piece in plus Vandelliinae clade. Considering min- Malacoglanis seems to differ from that de- iaturization at the base of the clade Tri- scribed above for Stauroglanis in its more dentinae/Stegophilinae/Vandelliinae is less rounded shape and more irregular profile. One parsimonious, because it requires one further and noteworthy distinction can be ob- miniaturization plus two subsequent re- served in the articulations shown by this an- versals: one in stegophilines, and at least terior section ofthe palatine. In Malacoglanis another in the nonminiature vandelliines (as- it apparently articulates only with the prox- suming that nonminiature vandelliines form imal portion of the maxilla (fig. 13), while in a monophyletic group). Separate miniatur- Stauroglanis, as noted, it articulates also with izations, in contrast, require only two steps: the premaxilla (fig. 7). The relative position one event in the tridentine clade and another of the bone here called lacrimal is also dis- inside the Vandelliinae. tinct, in Stauroglanis it extends directly above the lateral margin between the two palatine PHYLOGENETIC RELATIONSHIPS pieces, but in Malacoglanis it is located more OF laterally and does not cover the palatine com- STAUROGLANIS plex (fig. 13). In spite of these small distinc- Osteological data for the two previously tions, the state of the anterior palatine ossi- described Sarcoglanidinae are limited to fication in both taxa is similar, and homology Malacoglanis gelatinosus, from unpublished of the structure is a perfectly acceptable pro- anatomical sketches of Baskin (1973) and posal according to structural and topological from drawings provided by Dr. Scott Schae- criteria. Such a separate anterior palatine os- fer, both based on the single cleared and sification is not present in any other member stained paratype. The illustrations of Baskin of the Trichomycteridae. A similar structure show the neurocranium (dorsal and ventral occurs in the pimelodid Rhamdia (Cussac and views) and the suspensorium/opercular ap- Maggese, 1988, referred to this bone as "ar- paratus. As the internal anatomy of Sarco- ticular ossicle," whose presence was con- 20 AMERICAN MUSEUM NOVITATES NO. 2950

firmed also in the Rhamdia species here examined), and in the Sisoridae examined. Considering the many branching points cer- tainly separating Rhamdia and sisorids from sarcoglanidines in phylogeny, and further the absence ofthe structure in at least most other catfishes, it is certain that the anterior palatine ossification has evolved independently in these groups. This ossification can thus be considered a synapomorphy for the Sarco- glanidinae. The premaxilla in Malacoglanis bears a very long and slender lateral process, extending half the length of the maxilla (fig. 13). This process is very similar in shape and position to that described above for Stauro- glanis (fig. 7), showing equivalent positional Fig. 12. Dorsal view of left palatine of Mala- relationships to the surrounding structures. coglanis gelatinosus. Drawing inverted. Anterior In Malacoglanis this process is, however, piece of palatine stippled. Redrawn and modified much more elongate than in Stauroglanis. after Baskin (1973). Such a long premaxillary process is unique within all trichomycterids, and is absent in Nematogenys, in all remaining members of the Loricarioidea and in at least most re-

Fig. 13. Relationships among upperjaw elements and other skull structures ofMalacoglanis gelatinosus. Dorsal view of left side, drawing inverted. Dense stippling represents the premaxilla, light stippling represents the palatine. Redrawn and modified after Baskin (1973). 1989 de PINNA: SARCOGLANIDINE CATFISH 21

o0 10,;

Fig. 14. Suspensorium and opercular apparatus of Malacoglanis gelatinosus. Lateral view of right side, drawing inverted. Stippling represents the quadrate. Redrawn and modified after Baskin (1973). maining catfishes. In Stegophilinae and Tri- son to accept such a proposal. There exists dentinae there is a laterally elongate premax- no additional character supporting the clade illa that could be interpreted as continuous Sarcoglanidinae plus Tridentinae plus Ste- with a lateral process. Nevertheless, while the gophilinae, and the only available phyloge- premaxilla and its process in the Sarcogla- netic hypothesis (Baskin, 1973) indicates that nidinae are clearly distinct portions, in the tridentines and stegophilines form a mono- Stegophilinae and Tridentinae no such sep- phyletic group with the Vandelliinae, a aration exists. In the latter taxa, the entire subfamily with a distinct premaxillary mor- bone looks like a greatly elongate premaxilla, phology. In view of the above, the presence where no distinct process can be identified. of a long distal premaxillary process is given In addition, the premaxillary process is eden- as a synapomorphy at the level of the Sar- tulous in the Sarcoglanidinae (in Malacogla- coglanidinae. nis and Sarcoglanis the entire premaxilla The conspicuous shape of the quadrate of lacks teeth), while the elongate premaxilla of Stauroglanis (fig. 10) occurs also in Mala- the Stegophilinae and Tridentinae bears teeth coglanis (fig. 14). In both taxa, the anterior along its entire ventral surface. These struc- portion of the quadrate bears a large dorsal tural distinctions do not preclude the possi- process that curves posteriorly and nearly bility that the lateral premaxillary process of reaches the anterior portion of the hyoman- sarcoglanidines is homologous with the distal dibula, forming a quadrate-hyomandibular portion ofthe premaxilla in stegophilines and fenestra. Although a dorsal process ofthe an- tridentines, and thus a synapomorphy at a terior portion of the quadrate occurs also in level higher than that of the single subfamily many other trichomycterids, the large size Sarcoglanidinae. Nevertheless, these struc- and posterior extension of the structure are tural dissimilarities would require consider- unique to sarcoglanidines. As the state ofthe able further modification in one or both states quadrate in sarcoglanidines is also distinct for them to match a common pattern ex- from that in the Nematogenyidae, remaining pressable as a synapomorphy (in tridentines loricarioids, and other Siluroidei, it is pro- and stegophilines there is not even a properly posed as a synapomorphy for the subfamily. identifiable lateral premaxillary "process"). A superficially similar fenestra occurs in the On phylogenetic grounds, there is also no rea- suspensorium ofthe tridentine Tridensimilis. 22 AMERICAN MUSEUM NOVITATES NO. 2950

panded toward the lateral tip (fig. 15A, B). The widening results from expansion in an anteroposterior axis. In Trichogenes and most other trichomycterids, as well as in Nema- Ac57z -%7B togenys and most other catfishes, the mes- Fig. 15. Diagrammatic representation of the ethmoid comua narrow uniformly from the anterior portion of mesethmoid. Dorsal view. A, base to the tip, without widening at their dis- Stauroglanis gouldingi, B, Malacoglanis gelati- tal portion (fig. 1 6A-D). The condition ofthe nosus. (B) is based on a sketch provided by Dr. mesethmoid cornua in Stauroglanis and Scott Schaefer. Malacoglanis can thus be considered a synapomorphy for the two taxa. In some Ste- Nevertheless, the fenestra in this taxon is gophilinae (but not all, see fig. 1 7B, D), as in formed exclusively by an expansion of the Pseudostegophilus and Homodiaetus, the hyomandibula, which has the form of a pos- mesethmoid cornua also looks expanded dis- teriorly oriented dorsal process similar to the tally in general appearance (fig. 17A, C). process on the quadrate of sarcoglanidines. Nevertheless, the condition in these stego- Thus, the structures in the two taxa are not philines differs structurally from that in the the same at all, being formed as extensions comparable region of sarcoglanidines. In of different bones. The conditions in sarco- Stauroglanis and Malacoglanis the distal glanidines and tridentines cannot be consid- widening is the result of expansion in the ered homologous, consequently not valid as anteroposterior axis. In the abovementioned evidence for their relationship. stegophilines, in contrast, the widened out- Stauroglanis and Malacoglanis share a dis- line apparent in dorsal view results from a tinctive morphology of the anterior portion flat expansion on the ventral margin of the ofthe mesethmoid. The mesethmoid cornua cornua. This expansion is directed posterior- in the two taxa are distinctly wider distally ly, and its distal outline can be seen in dorsal than basally, their dorsal profile looks ex- view. As an effect of projection, the distal

A B

C D Fig. 16. Diagrammatic representation of the anterior portion of mesethmoid. Dorsal view. A, Ne- matogenys inermis (MZUSP 36956); B, Trichomycterus quechuorum (AMNH 20351); C, Hatcheria macraei (MZUSP 35687); D, Eremophilus mutisii (MZUSP 35409). 1989 de PINNA: SARCOGLANIDINE CATFISH 23

A _5- - - -~ B

> t~~~~~.._, 1 _ _

L;Ar L) Fig. 17. Diagrammatic representation of the anterior portion of the mesethmoid in some Stego- philinae. Dorsal view. A, Pseudostegophilus nemurus (MZUSP 30427); B, Ochmacanthus alternus (MZUSP 30467); C, Homodiaetus anisitsi (MCP uncat.); D, Haemomaster venezuelae (AMNH 10194). profile of the ventral expansion may be mis- glanis and Malacoglanis among trichomyc- taken for the actual posterior profile of the terids (their numbering not followed): (1) cornua, which gives a false anteroposteriorly compact, little-elongated body form; (2) very expanded appearance to the distal part ofthe large, relatively deep head; (3) long adipose comua. Hence, the expanded distal portion fin (high and conspicuous in one species) ex- ofthe mesethmoid cornua in sarcoglanidines tending from end ofdorsal fin base to almost and some stegophilines are not homologous. or quite to caudal fin, and not involved with Another noteworthy feature in the meseth- any procurrent dorsal supporting elements of moid ofStauroglanis and Malacoglanis is the the latter fin; (4) extremely large, broad pec- convex shape of its mesial anterior border toral fins, their length far greater than that of (fig. 1 5A, B). In most other trichomycterids, the large head; tips of pectoral rays not fila- and also in nematogenyids and in at least mentous but all tips project far beyond mem- most other siluroids, the mesial anterior bor- brane; (5) upper jaw toothless; dentary teeth der of the mesethmoid is concave, often uniserial, long, cylindrical, conical, and forming an inward angle at the site where the hooked inward at tip; and (6) a conspicuous, comua meet (fig. 16A-D). The convex con- saclike, fat-filled, adipose organ on each side dition in Stauroglanis and Malacoglanis immediately above pectoral fins. (more notable in the latter) can thus be given The six characters listed above are good as a further synapomorphy indicating that the diagnostic features, but not all of them are two forms compose a monophyletic group. evidence for the monophyly of Sarcoglanis The above five characters are proposed as + Malacoglanis. A discussion ofeach ofthem synapomorphies for Malacoglanis, Sarco- is necessary to decide which ones represent glanis, and Stauroglanis, being evidence for synapomorphies. these taxa forming a monophyletic Sarco- 1. Compact Body Form: This character- glanidinae. istic is very evident in Sarcoglanis, but not Myers and Weitzman (1966) listed six in Malacoglanis. The general body shape of characters that uniquely characterize Sarco- Malacoglanis is similar to that in some ofthe 24 AMERICAN MUSEUM NOVITATES NO. 2950 less-elongate species f Trichomycterus and data, may come from two sources, structural Scleronema, as well as in some Stegophilinae and/or phylogenetic. At present there are no and Tridentinae. The compact body form of detailed anatomical studies comparing the Sarcoglanis is actually very conspicuous, and structures of the adipose fin of Sarcoglanis contrasts with the state in all trichomycterids and Malacoglanis to that of remaining cat- and most other catfishes. The body shape of fishes. Based solely on superficial appearance, Sarcoglanis can, with justification, be regard- there is no evidence against their homology. ed as apomorphic. Nevertheless, as it is On phylogenetic grounds, the interpretation unique to this genus, it is to be interpreted of the adipose fin in these taxa as either a as an autapomorphy. Baskin (1973) ex- neomorph or a symplesiomorphy requires a pressed this character in terms of vertebral resolved hypothesis of relationships among counts. The compact shape of Sarcoglanis the outgroups to Sarcoglanis and Malacog- and Malacoglanis would be related to a very lanis that is presently unavailable. Thus, low number ofvertebrae, 30 to 3 1, not count- whether the structure in Sarcoglanis and Ma- ing those seven involved in the Weberian lacoglanis is a primitive adipose fin or a sec- complex and compound caudal centrum ondary acquisition cannot be decided at pres- (Baskin did not specify to which genus each ent. Nevertheless, it is worth noting that an count referred). These reduced vertebral adiposelike fin occurs also in Stauroglanis (in counts are unique within trichomycterids. The a very reduced state) and in some other mem- closest conditions are found in Trichomyc- bers of the Trichomycteridae, namely Pygi- terus hasemani, which can have as few as 32 dianops and Typhlobelus (the two genera of vertebrae, and T. duellmani which can have the subfamily Glanapteryginae) and also in as few as 33 (see Arratia and Menu-Marque, Scleronema (a Trichomycterinae). In all these 1984). Although the difference is small, there taxa the fin is not supported by procurrent is no overlap of counts and the reduced ver- caudal rays (lepidotrichia), but instead only tebral number in Sarcoglanis and Malaco- by thin actinotrichial filaments that can be glanis can be considered a synapomorphy for faintly discerned in transmitted light under the two taxa. high magnification. This structure is there- 2. Large and Deep Head: From the orig- fore in agreement with that known for ordi- inal descriptions, Sarcoglanis seems to have nary adipose fins. The structural equivalence a very deep head, more so than any other does not preclude the adipose fin being a sec- trichomycterid. In Malacoglanis, however, ondary, hence derived, acquisition in some this characteristic is not conspicuous and does trichomycterids. But if this is the case, the not differ notably from that in many other presence of an adiposelike fin would be syn- trichomycterids. An adequate and precise apomorphic not for Sarcoglanis and Mala- evaluation of this character would demand coglanis, but for a larger assemblage includ- exactly quantified comparative morphomet- ing some other trichomycterid taxa. ric data for trichomycterids in general-which 4. Large and Broad Pectoral Fins, Pectoral is still unavailable. In nonrigorous terms, it Rays Projecting Far Beyond Fin Membrane: seems that the condition of this trait is ob- This trait is actually two distinct and inde- viously unique in Sarcoglanis, but in Mala- pendent characters, both derived. The first, coglanis there occurs overlap with other taxa. the length of the pectoral fins, lacks a precise 3. Long Adipose Fin: The presence of an quantitative evaluation among trichomyc- adipose fin in Sarcoglanis and Malacoglanis terids. Nevertheless, the pectoral fins in both can be interpreted in two ways. Ifthe adipose Malacoglanis and Sarcoglanis are much fin in these two taxa is considered homolo- longer than the head, a condition apparently gous with that found in other loricarioids and not found in other members of the family. remaining catfishes, it is so a symplesiomor- As the plesiomorphic condition is a shorter phic feature. On the other hand, ifthe adipose pectoral fin, found in Trichogenes, most other fin is seen as a secondary acquisition (i.e., as members of the family, and in the Nemato- a neomorphic character), it can be regarded genyidae, the elongate condition in Sarco- as derived. Support for one or the other in- glanis and Malacoglanis is synapomorphic. terpretation, in the absence of ontogenetic The second character, the projection of the 1989 de PINNA: SARCOGLANIDINE CATFISH 25 pectoral-fin rays beyond the fin membrane to The statement about the dentary teeth ofSar- the extent seen in Sarcoglanis and Malaco- coglanis and Malacoglanis does not represent glanis, is unique within the family. It con- a synapomorphy for the two taxa, as Baskin trasts with the state most commonly found (1973) has already noted. Uniserial dentary among other loricarioids and remaining cat- teeth are found also in Stauroglanis, in the fishes, where pectoral-fin rays show little or Glanapteryginae and in some Vandelliinae. no projection beyond the fin membrane. In Dentary teeth that are long, cylindrical, con- the species ofScleronema examined, the pec- ical, and hooked inward at the tip are found toral-fin rays project considerably beyond the in most other trichomycterids and remaining fin membrane. In Stauroglanis, in spite ofthe catfishes, and are not, therefore, apomorphic poor condition of the fins in the available for sarcoglanidines. material, it can be seen that the pectoral-fin 6. Conspicuous, Saclike, Adipose Organ rays also project considerably beyond the fin Above Pectoral Fin: From the original de- membrane. The pectoral-fin rays in Sclero- scriptions, it seems that this adipose organ is nema and Stauroglanis project to an extent conspicuous in both Malacoglanis and Sar- that, aside from Sarcoglanis and Malaco- coglanis, but more so in the latter. This adi- glanis, is unique within the family. Never- pose organ is unique within the family, but theless, the state in the latter two genera is unfortunately little detail is known about its much more advanced than in either Sclero- structure. Even if the organ is homologous nema or Stauroglanis, and is evidence that with the axillary gland found in most tricho- Sarcoglanis and Malacoglanis are sister mycterids and several other catfishes (a sim- groups. ilarly poorly known structure), it can be re- 5. Upper Jaw Toothless; Dentary Teeth garded as uniquely derived for Sarcoglanis Uniserial, Long, Cylindrical, Conical, and and Malacoglanis. Its structure, size, and Hooked Inward at Tip: This character of shape differ markedly from those seen in the Myers and Weitzman, once again, can be axillary gland of remaining members of the considered two distinct characters, one con- family. cerning the premaxillary teeth and the other There are thus five synapomorphies indi- the dentary teeth. The total absence of pre- cating monophyly of the group formed by maxillary teeth is not found in any other Sarcoglanis and Malacoglanis. I propose here member of the Trichomycteridae. The clos- that Stauroglanis is the sister group of those est condition is that of the glanapterygine two genera, and that the derived characters Typhlobelus, which has only a single tooth uniting Stauroglanis and Malacoglanis will on each premaxilla. Nevertheless, all other also be found in Sarcoglanis, when further glanapterygines have higher counts of pre- material representing the genus becomes maxillary teeth. Given the available corrob- available (fig. 18). oration that the Glanapteryginae is a mono- At present there are only two characters phyletic group (Pinna, 1988), it is more that can be given as autapomorphic for Stau- parsimonious to consider the premaxillary roglanis. These are the dorsal lateral openings tooth reduction in Typhlobelus an indepen- on the neurocranium and the form of the dent derivation. Among loricarioid out- vomer (see Osteology). Both traits are unique groups, toothless premaxillae are found only within trichomycterids and have not been re- in some members ofthe Callichthyidae. Con- corded in any other catfish. Within the fam- sidering, nevertheless, that callichthyids are ily, only in Sarcoglanis is the state of these more closely related to the Scoloplacidae, structures unknown. Some other characters, Loricariidae, and Astroblepidae (Baskin, such as the cross-shaped basibranchial or the 1973; Howes, 1983; Schaefer and Lauder, unnamed bone below the maxilla, are also 1986), all with numerous premaxillary teeth, derived and unique within the family. Never- it is more parsimonious to consider the loss theless, as these structures are completely un- of teeth in some callichthyids as a separate known in both Sarcoglanis and Malacogla- event. Consequently, the toothless premax- nis, it is at present uncertain whether they illa of Sarcoglanis and Malacoglanis is fur- represent autapomorphies for Stauroglanis ther evidence for their being sister groups. or synapomorphies for the Sacroglanidinae. 26 AMERICAN MUSEUM NOVITATES NO. 2950 Stauroglanis Malacoglanis

12 11

8 7 6

5 4 3

Fig. 18. Cladogram of the Sarcoglanidinae and summary of proposed characters. Characters numbered are: 1, separate ossification of the anterior cartilage of palatine ("anterior palatine piece"); 2, premaxilla with long and slender lateral process, extending until midway the length of the maxilla; 3, quadrate with large dorsal process distally directed backward; 4, mesethmoid cornua expanded distally; 5, mesial anterior border of mesethmoid convex (all above characters predicted but not observed for Sarcoglanis); 6, pectoral fins large, much longer than head length; 7, pectoral-fin rays projecting well beyond fin membrane; 8, absence of premaxillary teeth; 9, conspicuous, saclike adipose organ above pectoral fin; 10, 30 or 31 vertebrae (not including those ofthe Weberian complex and compound caudal centrum); 11, lateral openings on dorsal side of neurocranium, framed by the supraoccipital, frontals, and sphenotics; 12, vomer with posterior portion expanded laterally, lateral tips directed backward. RELATIONSHIPS OF SARCOGLANIDINAE WITH 1987), but this raises no doubt about the de- OTHER TRICHOMYCTERIDAE rived polarity of the large maxilla in some trichomycterids. Diplomystidae are the sister There is one evident apomorphic character group of all remaining catfishes, and the shared by Sarcoglanidinae and some other structure and size of their maxilla, as well as trichomycterids: their enormously enlarged the presence of maxillary teeth, represent a maxilla, which is much larger than the pre- retention of the plesiomorphic condition of maxilla (figs. 7 and 19). Such a condition con- the maxilla within Ostariophysi (Fink and trasts with that seen in most other members Fink, 1981; Grande, 1987). The situation in of the family and with that in the Nemato- diplomystids contrasts with that seen in all genyidae and most other catfishes. The ple- other catfishes (except the fossil Hypsidori- siomorphic state within Siluroidei is a small dae, the next sister group to catfishes; see maxilla, with the same dimensions or smaller Grande, 1987), in which the maxilla is tooth- than the premaxilla (fig. 20A-C). The Diplo- less, reduced in size and not related to the mystidae have the maxilla larger than the mouth gape (it is, instead, related only to the premaxilla (Fink and Fink, 1981; Arratia, palatine-maxillary mechanism of barbel 1989 de PINNA: SARCOGLANIDINE CATFISH 27

Fig. 19. Diagrammatic representation of the maxilla and premaxilla of Malacoglanis gelatinosus. Dorsal view ofleft side, drawing inverted. Maxilla stippled. Redrawn and modified after Baskin (1973). movement; see Gosline, 1975). The mor- group ofthe Trichomycteridae. According to phology of the maxilla seen in the nondiplo- the same hypothesis, the Nematogenyidae is mystid (and nonhypsidorid) catfishes thus the sister taxon to the clade Trichomycteri- represents a uniquely derived character in- dae/Callichthyidae/Scoloplacidae/Loricari- dicating their monophyly (Fink and Fink, idae/Astroblepidae. Nematogenys has a small 1981). The state of the maxilla in the maxilla, the plesiomorphic condition within Sarcoglanidinae does not resemble the prim- siluroids (fig. 20A). Trichogenes, the sister itive condition seen in diplomystids, it does group of all remaining trichomycterids (Pin- not participate in the gape of the mouth, is na, in prep.) also has the plesiomorphic con- not a compressed bone, and is totally related dition for the maxilla (fig. 20B), similar to to the mechanics of the maxillary barbel that ofNematogenys. In this scheme, the oc- movement. The maxilla of sarcoglanidines currence of an enlarged maxilla within the approaches the condition in diplomystids only Trichomycteridae has to be considered a pu- in its relative size being larger than the pre- tative synapomorphy for the taxa sharing the maxilla. If it is taken into consideration that trait. This reasoning involves a scheme of the Sarcoglanidinae are part of a monophy- outgroup comparison which in general terms letic Trichomycteridae, that the Trichomyc- follows Maddison et al. (1984), and can be teridae form a monophyletic group with the diagramatically appreciated in figure 21. Fol- other loricarioid families, and further that the lowing the steps of the algorithm of Maddi- nondiplomystid catfishes form a monophy- son et al. (1984) and considering the non- letic group, then it becomes evident that the Trichogenes Trichomycteridae as an ingroup, large relative size of the maxilla in sarco- the state at the outgroup node (i.e., the ple- glanidines is a secondary modification. This siomorphic state for the ingroup) is to have holds true even though a large maxilla is found a small maxilla. The equivocal state assess- in diplomystids and most other noncatfish ment at the nodejoining trichomycterids with ostariophysans. Considering the large max- remaining loricarioids becomes decisive at illa of sarcoglanidines as a posterior modifi- the outgroup node, when Trichogenes is in- cation (i.e., as an apomorphic character) is cluded in the analysis. Consequently, the large more parsimonious than considering it as a maxilla seen in a subgroup of this particular symplesiomorphy with diplomystids. A more ingroup can be considered a synapomorphy delicate situation relates to the condition in for it. In view ofthe above, the enlarged max- other loricarioids. All callichthyids and scol- illa ofsarcoglanidines is an apomorphic char- placids examined have a maxilla much larger acter. It should be noticed that this polarity than the premaxilla. In loricariids and as- assessment does not change if the Nemato- troblepids the maxilla is also large, with the genyidae is placed as the sister group to same length as, or larger than, the premaxilla. trichomycterids (the traditional view before According to the presently available phylo- Baskin, 1973). Actually, given the position genetic hypothesis of loricarioid phylogeny of Trichogenes, nontrichomycterid out- (Baskin, 1973; Howes, 1983; Schaefer and groups cannot completely change this polar- Lauder, 1986), the Loricariidae, Astroblepi- ity assessment, regardless of their interrela- dae, Scoloplacidae, and Callichthyidae form tionships (see Rule 3 of Maddison et al., a monophyletic group, which is the sister 1984). This observation is particularly rele- 28 AMERICAN MUSEUM NOVITATES NO. 2950

studied in depth is probably monophyletic. Species of Scleronema share a conspicuous derived dermal opercular flap (a character traditionally used to diagnose the genus) not present in other trichomycterids nor in outgroups. Although presently supported only by this single unequivocal derived character, the monophyly of Scleronema stands as an A acceptable proposal, since there is no evidence for an alternative hypothesis. The maxilla of Scleronema is similar, in relative size to the premaxilla and other upper jaw elements, to that seen in the Sarcoglanidinae (fig. 22A). As occurs in the latter taxon, the I~~~~~~~~~~~~~~~~~~~~~~~~~~~::::.. huge dimensions of the maxilla in Sclero- nema can be detected externally, as a large B and conspicuous bony base for the maxillary barbel. This character was noted by Myers and Weitzman (1966), who first suggested that Scleronema might be related to Sarcoglanid- inae. Another form sharing the derived enlargement of the maxilla is a species presently in- pm cluded in the genus Trichomycterus, namely T. boylei (Nichols). This species was rede- scribed by Arratia and Menu-Marque (1984), who provide much anatomical data not con- C tained in the original description. In contrast to the condition in the Sarcoglanidinae and Schleronema, the enlarged maxilla of T. boylei is not evident externally. The condition in T. boylei (fig. 22B) is much less conspicuous than in Scleronema and sarcoglanidines, but even so it is distinctly enlarged relative to the plesiomorphic state outlined above (compare figs. 20 and 22). Thus, the Sarcoglanidinae, Scleronema, and D T. boylei share a uniquely derived enlarge- Fig. 20. Diagrammatic representation of the ment ofthe maxilla indicating that they form maxilla and premaxilla of: A, Nematogenys inermis (MZUSP 36956); B, Trichogenes longipinnis a monophyletic group. This assemblage, (MZUSP uncat.); C, Eremophilus mutisii (MZUSP placing together one subfamily and two non- 35409); D, Trichomycterus quechuorum (AMNH congeneric members of another subfamily, 20351). Dorsal view, right side. Maxilla stippled. might seem at first hardly acceptable, if one assumes that the nomenclatural status reflects some phylogenetic information. Never- vant when it is remembered that a few other theless, as shown in the next section, proposal catfishes (e.g., Sisoridae) also have the max- of a group in disagreement with the present illa larger than the premaxilla. Trichomycterinae does not require any pu- This apomorphic enlargement ofthe max- tative synapomorphy to become homoplasti- illa of sarcoglanidines is found also in some cally distributed, because monophyly of this members ofthe subfamily Trichomycterinae, subfamily is not supported by any such evi- presently referred to distinct genera. The first dence. of these is Scleronema, which although not As noted in the previous section, the con- 1989 de PINNA: SARCOGLANIDINE CATFISH 29

'0 ft wCf C S 0§ S X~~~~~~~~~~~~~~~~~~~~~~~C tC '0~~~~~~~~~~~~~~~~~~~ ~~~~0~0 Fig.2.D0nt its0cfV. ) co ) A. c5 A.: I

.. /~~~~~~~~~~

~~~~~~~~~x outrou nod

ca~~~~riode~ ~~ representssalailStpetrnlsmTrichomycter iae

Lorica rioidea

Fig. 21. Diagram explaining the polarity assessment of the large maxilla of the Sarcoglanidinae and its closest relatives, following the procedures of Maddison et al. (1984). X represents large maxilla, Y represents small maxilla. Stippled triangles (meaning polytomies at their insertion nodes) represent unresolved relationships and/or components which are not relevant for the present discussion. For further explanation see text.

vex, rounded anterior mesial border of the mesethmoid is a synapomorphy joining Stauroglanis with other sarcoglanidines. In-

pm deed, the trait in Stauroglanis and Malaco- glanis is developed to a degre unique for these two taxa. Nevertheless, the same trait is present, in a much less evident condition, in Scle- ronema (fig. 23A). This character, therefore, provides further corroboration that Sclero- nema is the sister group of the Sarcoglanid- inae. Accordingly, presence of a convex mesial anterior border of the mesethmoid is interpreted as a synapomorphy at the clade Scleronema/Sarcoglanidinae, and the more B extreme state of this character, in turn, is Fig. 22. Diagrammatic representation of the synapomorphic for a less inclusive clade maxilla and premaxilla of: A, Scleronema minu- (Stauroglanis plus remaining sarcoglani- tum (MCP 9315); B, Trichomycterus boylei dines). In the larger of the two specimens of (MZUSP uncat.). Dorsal view, right side. Maxilla T. boylei examined (40.5 mm SL), the ante- stippled. rior mesial margin ofthe mesethmoid is also 30 AMERICAN MUSEUM NOVITATES NO. 2950

may have as few as eight rays (including T. boylei), and Malacoglanis has seven, which makes the difference too small for a meristic feature known to show considerable intraspecific variation. Scleronema may have very reduced anal-fin ray counts, but considerable variation occurs, and counts ranging from six A B to eight have been observed. Another char- Fig. 23. Diagrammatic representation of the acter Baskin (1973) used to corroborate anterior portion ofmesethmoid in: A, Scleronema monophyly of sarcoglanidines plus glanap- minutum (MCP 9315); B, Trichomycterus boylei terygines was the "opercular and interoper- (MZUSP uncat.). Dorsal view. cular odontodes reduced or absent." This trait is undoubtedly apomorphic at this level within the family, and is not seen either in Scle- convex, in the same style seen in Scleronema ronema or T. boylei. Nevertheless, reduction and sarcoglanidines (fig. 23B). In the smaller of odontodes has occurred independently in specimen (31.8 mm SL), the anterior margin Sarcoglanidinae and in Glanapteryginae. The of the mesethmoid is straight, not convex as monophyly ofListrura plus remaining Glan- in Scleronema and sarcoglanidines but also apteryginae is a well corroborated hypothesis not concave as in the primitive condition. (Pinna, 1988), but both species of this genus Possibly the convex state of the mesethmoid have numerous opercular and interopercular becomes evident only in fully grown individ- odontodes. The state in Listrura thus retains uals. The smallest Scleronema specimen ex- the plesiomorphic condition within tricho- amined (19.1 mm SL) has no evident convex mycterids, and remaining glanapterygines margin, while the larger individuals have it. have lost odontodes (at least in the adult Even though not fully conclusive at present, stage). This means that, once monophyly of the presence ofa convex anterior mesial mar- Glanapteryginae is accepted, reduction of gin of the mesethmoid at least agrees with a odontodes has occurred inside the subfamily hypothesis which places T. boylei in a mono- (more precisely at the Glanapteryx + Typh- phyletic group with Scleronema and the Sar- lobelus + Pygidianops clade), and not at a coglanidinae. higher level. As a further consequence, the Baskin (1973) provided five characters to reduction of odontodes seen in Sarcoglanid- support the Sarcoglanidinae plus the Glan- inae, a reduction which is less marked than apteryginae as a monophyletic group. Some in Glanapteryginae, is an independent event. of these characters, if accepted as valid evi- This conclusion involves a scheme of parsi- dence, would put into doubt the proposed mony, in which acceptance of homoplasy in monophyly of the group Sarcoglanidinae + the reduction of odontodes is more parsi- Scleronema + T. boylei. This occurs because monious than postulating homoplasy in all Baskin's characters proposed to unite sar- other known putative synapomorphies sup- coglanidines with glanapterygines are absent porting other components of relationship. in Scleronema and T. boylei, thus implying The relationship proposed here implies that the charactersjoining these two taxa with considerable alterations in the presently ac- sarcoglanidines are homoplastic. cepted classification, and would require Scle- Three of Baskin's characters, "opercular ronema and T. boylei to be transferred to the bone with a long posterior process," "dorsal subfamily Sarcoglanidinae. However, con- membrane present," and "reduced number sidering the still poorly known anatomy of of premaxillary teeth," have been discussed Scleronema, Sarcoglanis, and Malacoglanis, by Pinna (1988), who concluded that they are it is better to wait for a more complete set of not valid evidence for the monophyly ofSar- data, before making the necessary alterations coglanidinae + Glanapteryginae. Another of of nomenclature. Furthermore, removing T. Baskin's characters, "anal rays fewer than boylei from the Trichomycterinae would also eight," as Baskin himself observed, is not re- imply its removal from Trichomycterus, liable evidence. Some other trichomycterids which could be a premature alteration. While 1 989 de PINNA: SARCOGLANIDINE CATFISH 31 phylogenetic relationships among T. boylei, southeastern Brazilian species, all having Scleronema and sarcoglanidines remain un- normal pelvics (e.g., see illustration of resolved, there can be no adequate generic Trichomycterus brasiliensis in Britski et al., placement for this species. 1986). Scleronema comprises S. opercula- Baskin (1973) proposed that Sarcoglanid- tum, S. minutum, and S. angustirostris, inae, Glanapteryginae, and Trichomycteri- species in need of a careful taxonomic revi- nae comprise a monophyletic unit, referred sion. Hatcheria (see Arratia and Menu- to by him as the Trichomycterinae-group. His Marque, 1981), Bullockia, and Rhizosomich- proposal relies on a single character, the pres- thys are presently monotypic. ence of a frontolacrimal tendon bone. As The taxa united as Trichomycterinae show mentioned above, this structure has already a general resemblance to each other which been proposed as homologous with the su- understandably led early taxonomists to con- praorbital and also with the antorbital, and sider them as a natural group. However, as its homology to structures in other catfish discussed below, this subfamily is not mono- groups seems problematical. Schaefer (1987) phyletic. The problems involved with the reported a structure similar in shape and po- recognition of Trichomycterinae are related sition in the family Loricariidae. A bone oc- to interpretation of data, and were not ame- cupying the same position (but with a distinct nable to critical evaluation by early system- shape) is present in Trichogenes longipinnis, atists working with the group. This fact was the plesiomorphic sister group to the whole not due to inaccuracies in their observations, family Trichomycteridae (Pinna, in prep.). If nor to lack of morphological data, but in- the structures in loricariids, Trichogenes, stead, was a direct consequence of the limi- trichomycterines, sarcoglanidines and glan- tations of the systematic methods available apterygines are accepted as homologous, then at the time. This was part of a general ten- their occurrence in the three latter taxa does dency to place highly derived forms in sep- not indicate their monophyly, because it is a arate high-level groups and to unite gener- symplesiomorphy at that level. alized forms in a taxon of equivalent rank. In present terms, the adoption ofa group such PHYLETIC STATUS OF THE as Trichomycterinae is related to acceptance SUBFAMILY TRICHOMYCTERINAE ofsymplesiomorphies as potential indicators of monophyly, which is further related to a The genera Trichomycterus, Hatcheria, definition of monophyly that is not strict. Scleronema, Bullockia, Rhizosomichthys, and A survey of the pertinent literature pro- Eremophilus have been traditionally placed vides a number ofcharacters previously used together to form the subfamily Trichomyc- to delimit or diagnose the subfamily Tricho- terinae. Trichomycterus is by far the most mycterinae. These characters are evaluated speciose genus of the group (and of the fam- below relative to their value as potential in- ily), consisting of over 100 nominal species dicators of phylogenetic kinship. The char- and large numbers ofundescribed forms. Er- acters are extracted mainly, but not exclu- emophilus includes at present two species, E. sively, from works with revisionary aims, mutisii from Colombia, and E. candidus from namely Eigenmann (1918), Myers (1944), and southeastern Brazil (E. camposi is actually a Arratia et al. (1978). These works summa- member of the Glanapteryginae, now placed rized most data relevant to the group, and in the genus Listrura). The two species of attempted wider comparative concerns. In the Eremophilus are brought together based on two former papers, characters provided in a single character, the absence of pelvic fins, keys were also scrutinized, since authors of and it is probable that E. mutisii and E. can- that time used to express their justifications didus do not form a monophyletic group. Pel- for particular groupings in the form of keys. vic fins have been lost at least two other in- Each character treated below is numbered and dependent times within the family, in the is immediately followed by a short commen- Glanapteryginae and in the Tridentinae (in tary evaluating its generality. Miuroglanis; see Baskin, 1973). Further- 1-Nasal Barbal Present: A nasal barbel is more, E. candidus is very similar to some present in all members of the Glanapterygi- 32 AMERICAN MUSEUM NOVITATES NO. 2950 nae and Sarcoglanidinae, in Tridentopsis (a also found in most other subfamilies and re- genus of the Tridentinae), and (in variable maining catfishes, including nematogenyids degrees of development) in the Nematogen- and diplomystids) is plesiomorphic, and yidae (but the condition in the latter taxon hence no evidence for their monophyly. may be nonhomologous; see Arratia, 1987: 7-Mouth Subterminal, not Suckerlike: 91). Presence of a nasal barbel cannot con- The Vandelliinae, Stegophilinae and Triden- sequently be considered a synapomorphy at tinae all have inferior, highly specialized the level of the Trichomycterinae. mouths. What characterizes the Tricomyc- 2-Two Barbels at Angle ofMouth: Nearly terinae is their lack of these buccal special- all trichomycterids possess this trait, in no izations. A similar subterminal and unspe- way unique to the Trichomycterinae. This cialized mouth is found also in Trichogenes, character is probably a synapomorphy at the sarcoglanidines, glanapterygines, nemato- level of the whole family, considering that it genyids and most other catfishes. Thus, the is not found in other members of the Lori- condition in trichomycterines is a primitive carioidea and remaining catfishes (the con- character and does not indicate their exclu- dition in Callichthyidae is nonhomologous; sive common ancestry. Only ifthe suckerlike see Baskin, 1973: 42 and Arratia, 1987: fig. oral disc of some other loricarioids is con- 40C). sidered the primitive state for loricarioids, 3-Gill Membranes Free or Narrowly could the unspecialized mouth oftrichomyc- Connected With Isthmus: This is the plesio- terines be considered a secondary reversal and morphic condition for the gill membranes at thus a derived trait. But no evidence exists the level of the family, found in Trichogenes that a suckerlike mouth is the primitive con- and in the Nematogenyidae. Free or narrowly dition for loricarioids. Actually, the available united branchial membranes furthermore also hypothesis for loricarioid interrelationships found in glanapterygines and sarcoglani- (Baskin, 1973; Schaefer and Lauder, 1986) dines. indicates exactly the opposite. 4- No Mental Barbels: Absence of mental 8 -Caudal Fin Rounded, Truncated, or (chin) barbels is characteristic of all tricho- Emarginate: The shapes specified above vary mycterids. This character is possibly apo- enormously within most subfamilies of morphic at the family level, hence invalid to Trichomycteridae, and do not delimit any unite members of a subgroup. grouping in agreement with the current 5-Opercle and Interopercle with Odon- Trichomycterinae. todes ("Spines"): Mere presence of opercular and interopercular odontodes is a symple- The following characters are osteological, siomorphy, since these are found also in other and were presented by Arratia et al. (1978) loricarioids. The profound modifications of as diagnostic for their subfamily Pygidiinae shape seen in the opercle and interopercle of (= Trichomycterinae). all trichomycterids is almost always associ- 9-Well-Developed Palatine Connected ated with the presence of a definite restricted with Endopterygoid, Premaxilla, and Some- site of odontode attachment. These modifi- times the Maxilla: Arratia (in litt.) informs cations are unique within catfishes and con- me that there was a printing mistake in the stitute an unequivocal synapomorphy for description of this character in Arratia et al. joining all trichomycterids. Thus, they do not (1978), and that the original sentence should represent evidence for monophyly ofthe Tri- be "well developed palatine connected with chomycterinae only. endopterygoid, maxilla and sometimes the 6-Anal Fin Short: This rather vague char- premaxilla." The comments below refer to acter was evidently intended to contrast Tri- the corrected sentence. A well-developed and chomycterinae with Tridentinae, which pos- wide palatine is characteristic of almost all sess high number of anal-fin rays (more than trichomycterids, with the exception of the 15). While the state in tridentines can be con- glanapterygine Typhlobelus (which has all jaw fidently regarded as apomorphic, the condi- structures enormously modified). The con- tion in trichomycterines (8 to 12 rays, a count nection of the palatine with the maxilla is a 1989 de PINNA: SARCOGLANIDINE CATFISH 33 common feature of most catfishes, and is in- trichomycterids. It is also present in the Sar- volved with movement ofthe maxillary bar- coglanidinae and Glanapteryginae (see dis- bel. The connection of the palatine with the cussion above on relationships of sarcoglan- premaxilla, although not widespread among idines with other trichomycterids). catfishes, is found also in most other tricho- 11 -Quadrate, Metapterygoid, Hyoman- mycterids. The connection ofthe palatine with dibula, and Preopercle Incompletely Fused: the pterygoid bones is found in many other There is some disagreement about the iden- trichomycterid subfamilies, such as Stego- tity of the pterygoid bones of trichomycter- philinae and Vandelliinae, and hence is not ids. Arratia et al. (1978) suggested that the unique to Trichomycterinae. The connection metapterygoid is represented by the bony ofthe palatine to the suspensorium in tricho- plate that partly surrounds the quadrate pos- mycterids is undoubtedly a remarkable fea- terodorsally, so that the hyomandibula and ture, deserving careful comparative investi- metapterygoid are fused into a single bone, gation. Nonetheless, it is clear that this the anterior portion of which corresponds to character cannot be a synapomorphy for the the metapterygoid and the posterior portion members ofthe Trichomycterinae, since it is to the hyomandibula. This view was appar- not unique to them. ently supported by the presence of a crest 10-Supraorbital Present: As noted above extending along the bone. This crest seems (see Osteology and Relationships of Sarco- to have been interpreted as the primitive su- glanidinae with other Trichomycteridae), tural limit between the hyomandibula and there is no agreement on the nature of this the metapterygoid, and would result from bone, also proposed as an antorbital (Britski their fusion (see Arratia et al., 1978: 164, fig. and Ortega, 1983) or an ossified frontolacri- 2B). An alternative view adopted here (also mal tendon (the "fronto-lacrimal tendon Arratia, 1987, and Baskin, 1973) is that the bone" of Baskin, 1973). It should be noted whole bony plate that articulates with the that absence ofthe supraorbital has been giv- neurocranium is solely the hyomandibula. en as a synapomorphy for all catfishes (Fink The crest seen on the outer surface ofthe bone and Fink, 1981). If this bone is really a su- is a process for muscle attachment. The an- praorbital (an alternative which would re- teriormost separate suspensory element, ar- quire altering the level of universality of the ticulating ventrally with the quadrate and lig- character "absence of supraorbital" for silu- amentously joined anteriorly to the palatine, roids), its common possession by some is considered the only pterygoid element. trichomycterids would not imply their com- Whether this element is homologous with the mon ancestry, since this structure would be metapterygoid or with the mesopterygoid of a symplesiomorphy. A supraorbital is present other catfishes is uncertain, as is the identity in many Characiformes and remaining Os- ofthe pterygoid bones of siluroids in general tariophysi. Evidently the presence of "su- (for some discussion see Arratia, 1987; praorbital" in some trichomycterids can be Howes, 1983; Gosline, 1975; Fink and Fink, interpreted as a secondary acquisition, thus 1981). According to Arratia (personal com- a derived character. But such a proposal would mun.), this pterygoid element of trichomyc- necessarily rely on specific resolved relation- terids is formed ontogenetically in an endo- ships ofthe group, which are not available at chondral fashion, which hence supports present. A similar reasoning applies if the homology with the metapterygoid. Ifso, only bone is proposed as an antorbital. The sug- the partial fusion of the quadrate, hyoman- gestion that the bone is a frontolacrimal ten- dibula, and preopercle is left to be analyzed don bone implies that it is a neomorphic fea- as a possible diagnostic character for the ture, thus exclusively derived for the taxa Trichomycterinae. I was not able to locate sharing it (a view expressed in Baskin, 1973). evidence offusion between these bones in any Whatever be the case, the presence of a fron- trichomycterids examined. The quadrate, tolacrimal tendon bone (or a secondarily ac- hyomandibula, and preopercle are frequently quired supraorbital or antorbital) is not ex- joined in a very tight union, but it is hard to clusive to the Trichomycterinae among say whether there is fusion involved with this 34 AMERICAN MUSEUM NOVITATES NO. 2950 union. Arratia (in litt.) stated that limits latus. The situation in these species (includ- among the hyomandibula, quadrate and pre- ing T. quechuorum, as examined here in the opercle are indistinguishable in adult speci- same specimen used by Lundberg and Bas- mens of Trichomycterus aerolatus, T. chil- kin) is similar to that described by Arratia toni, T. rivulatus, T. maete, T. mendozensis, (1983) for T. mendozensis and Hatcheria T. chungaraensis, Hatcheria macraei, and macraei. In these cases, only a suture remains Bullockia maldonadoi. Unfortunately there between the tightly united parhypural and are no equivalent data for other trichomyc- hypurals 1 + 2, a condition contrasting to terid subfamilies, thus no comparison ofthis that seen in nematogenyids. In the latter fam- detail is possible. Under gross examination ily, the parhypural is well separated from hy- ofcleared and stained preparations available, purals 1 + 2 by a space, which separates the however, it seems that the state of the joint two units until their attachment to the com- between the three bones in trichomycterines pound centrum (see Lundberg and Baskin, does not differ from the condition in other 1969; Arratia, 1983). The condition of the subfamilies. parhypural in Nematogenys reflects the prim- 12-Occipital Crest Lacking: The medial itive state for catfishes (Lundberg and Baskin, dorsal crest on the supraoccipital, present in 1969) and the fusion (or near fusion) seen in most catfishes, is absent in all members of trichomycterines is consequently a derived the Trichomycteridae, not only in the Tricho- trait. Nevertheless, similar patterns of the mycterinae. Thus, this character is apo- caudal skeleton occur in most, perhaps all, morphic at the level ofthe family, not of the remaining members ofTrichomycteridae and subfamily (see Baskin, 1973). Loricarioidea, as well as in numerous other 13-Pectoral Girdle with One Radial: catfishes. I have not found a single derived Trichomycterines have two pectoral proxi- pattern of hypural fusion shared exclusively mal radials, one ossified and the other car- by the members of the Trichomycterinae. tilaginous. From the shape and position of Perhaps the occurrence of separate hypurals the element illustrated by Arratia et al. (1978), 3, 4, and 5 in some specimens ofsome mem- it is clear that the authors refer to the ossified bers of Trichomycterinae can be found only proximal radial, which is the larger of the in that subfamily among trichomycterids. two. Arratia et al. (1978) were obviously con- Nevertheless, the separate condition of the trasting the situation in Trichomycterinae to upper hypurals is found in numerous other that in Nematogenyidae (their Nematogen- catfishes, and is the primitive situation for yinae), where three ossified proximal radials these structures (Lundberg and Baskin, 1969). are present. Although all members ofTricho- Besides being a plesiomorphic trait, the sep- mycterinae for which data are available (rep- arate condition of the upper hypurals in resentatives of all genera but Rhizosomich- trichomycterines is a difficult character to thys) have only a single ossified proximal ra- evaluate phylogenetically, due to its large in- dial, this character is not exclusive to that traspecific variation (Arratia, 1983). group, but is also found in other members of the family (e.g., most Vandelliinae and some As seen above, none ofthe characters used Stegophilinae). Thus, this character is not ex- until the present to separate the Trichomyc- clusive to Trichomycterinae and is not evi- terinae from the remaining members of the dence for their monophyly. family can be considered a synapomorphy for 14-Parhypural, Hypural 1 and Hypural 2 the group. My studies have failed to disclose Fused; Hypurals 3, 4, and 5 with Different a single derived character exclusive to the Degrees of Fusion: The parhypural, hypural Trichomycerinae, in spite of specific efforts 1, and hypural 2 are fused into a single unit in this direction. This situation by itselfdoes in all members of Trichomycterinae, a con- not directly imply that the subfamily is non- dition contrasting to the separate parhypural monophyletic, but only represents a lack of seen in the Nematogenyidae. Lundberg and evidence for its monophyly. Positive evi- Baskin (1969) report a separate parhypural dence for nonmonophyly of the Trichomyc- for Eremophilus mutisii, Vandellia cirrhosa, terinae can come only from the demonstra- Tricomycterus quechuorum, and T. vermicu- tion that some forms previously included in 1989 de PINNA: SARCOGLANIDINE CATFISH 35

A B - 1 Fig. 24. Diagrammatic representation ofdorsal view ofskull of: A, Tridentopsis sp. (MZUSP 36303); B, Trichomycterus hasemani (MZUSP uncat.). Hatching represents the cranial fontanel. the group are actually more closely related to that of tridentines is present also in T. hase- other subfamilies. From this point of view, mani (fig. 24B), and this constitutes a syn- the nonmonophyletic status of the Tricho- apomorphy for the two taxa. Elsewhere with- mycterinae can already be considered evi- in trichomycterids, a comparable situation is dent. As seen above, there is evidence indi- found only in the vandelliine Paravandellia cating that Scleronema, T. boylei, and perhaps (senior synonym of Branchioica), as shown T. santaeritae, all of them included in the in figure 25. But the situation in Paravan- Trichomycterinae, are actually more closely dellia is an independent event. In other van- related to the Sarcoglanidinae. To these ex- delliines the cranial fontanel is completely amples I would add Trichomycterus hase- closed (except in Paracanthopoma, which has mani (and possibly also T. johnsoni, a dis- a complex and unique skull roof structure) tinct but seemingly very similar species, so and this is also the situation in all stegophi- far too poorly known to allow an adequate lines (considered by Baskin, 1973, as the sis- comparison). Baskin (1973) proposed that the ter group ofvandelliines). Consequently, it is enormously expanded cranial fontanel (which more parsimonious to consider two separate practically leaves the neurocranium without events of fontanel expansion (one in triden- a roof) seen in the members of the Triden- tines and another in Paravandellia) than to tinae (see fig. 24A) constitutes a synapomor- put a single fontanel expansion at the base of phy for this subfamily. I agree with him about the tridentine/stegophiline/vandelliine clade this character; such an expanded fontanel is with subsequent reversals in stegophilines and not seen in other loricarioids or in most other non-Paravandellia vandelliines. Placing T. catfishes. An expanded fontanel similar to hasemani as the sister group exclusively of 36 AMERICAN MUSEUM NOVITATES NO. 2950

that in tridentines than to that in Paravan- dellia. In T. hasemani and tridentines, the distal portions of the anterior margin of the supraoccipital are gently turned laterally. In Paravandellia, in contrast, the profile of the anterior margin of the supraoccipital is in a single arch. Although it cannot be determined whether the similar supraoccipital profile in T. hasemani and tridentines represents a derived condition (because there are no close outgroups with expanded fontanels to compare), the similarity in shape is at least an argument in favor of the homology of the fontanel expansion in the two taxa, when compared to Paravandellia. An interesting observation at this point is that Stauroglanis would fit into the Tricho- mycterinae quite well, according to the available diagnostic characters for the group. Not- withstanding, as seen, Stauroglanis is rather related to the Sarcoglanidinae. A similar example is the genus Listrura, which according to traditional criteria would easily fall into the Trichomycterinae (actually one of the species, L. camposi, was originally described within a trichomycterine genus, Eremophi- lus), but is actually more closely related to the Glanapteryginae (Pinna, 1988). All other subfamilies ofthe Trichomycter- idae are probably monophyletic (some evidence is provided in Baskin, 1973), in agreement with the naturalness envisioned by early authors for the same groupings. This agreement is due to their possession of some externally evident and distinctive specializations, which made each of them seem to fit "natural" assemblages. It is not surprising that together with the recognition ofobvious- ly aberrant groups, a group was also erected to place together the forms lacking those ev- Fig. 25. Diagrammatic representation of dor- ident specializations. This was apparently sal view of skull in Paravandellia sp. (MZUSP what has led to the establishment of the uncat.). Hatching represents the cranial fontanel. Trichomycterinae. After its adoption, the composition of the group remained unques- Paravandellia would be also unparsimoni- tioned, a situation which largely persists to- ous, because it would require reversal in the day. Obviously the nonmonophyly of the several characters joining Paravandellia with Trichomycterinae does not preclude the ex- other vandelliines in the lineage leading to T. istence of very large monophyletic subunits hasemani (for some of these characters, see within the subfamily, but it only means that Baskin, 1973). It is interesting to observe that the subfamily as presently composed is not the shape of the anterior border of the su- monophyletic. praoccipital (i.e., the posterior profile of the It is possible that the forms now united as fontanel) in T. hasemani is more similar to trichomycterines represent various sister 1989 de PINNA: SARCOGLANIDINE CATFISH 37 groups to different subfamilies and groups of derived characters exclusivelyjoining astrob- subfamilies, and will be shown in the future lepids with loricariids (and also those joining to compose widely distant lineages. The nu- the two families with callichthyids) (Schaefer merous species at present placed as tricho- and Lauder, 1986) are also homoplasies. This mycterines will possibly be scattered would result in an enormously unparsimo- throughout the family, at various levels, in a nious arrangement of loricarioid catfishes. complex phylogenetic pattern. Naturally the The expanded shape of palatine of astroble- fragmentation of the Trichomycterinae in- pids actually resembles that in all tricho- volves resolution of all these relationships, a mycterids, and this condition contrasts with task that undoubtedly will demand detailed the more slender palatine of remaining lori- long-term studies. carioids and catfishes. Nevertheless, given the A situation analogous to that outlined available corroboration for the other com- above can be seen within trichomycterines ponents of relationship, this similarity must themselves. Each genus included within the be the result ofconvergence. I did not observe subfamily is characterized by some distinc- a concave palatine in any trichomycterid ex- tive and apparently apomorphic character, amined, but I did not examine the species except the genus Trichomycterus. This genus available to Howes (T. rivulatus). In any event, has been, and continues to be, diagnosed ex- if T. rivulatus and astroblepids share a de- actly by the lack ofspecializations character- rived concave palatine, this character is also izing the remaining genera of the subfamily, more parsimoniously interpreted as conver- a situation remarkably similar to that seen at gent. the higher level of its subfamily. It is no co- The situation outlined above does not mean incidence that the genus Trichomycterus that descriptions of new taxa in the genus comprises the largest species assemblage of Trichomycterus should be viewed as an error the family, a set of numerous superficially or in any sense undesirable. While there are similar forms lacking evident specializations. no resolved relationships that could provide Since the characters elucidative for relation- better generic arrangements, it is completely ships are probably internal anatomical de- justified to make known the numerous species tails, they have remained unnoticed under that are yet undescribed, even if their place- the ordinary depth of character analysis. As ment in Trichomycterus is assumedly pro- a consequence, an enormous number of visory. This situation is not uncommon in species have come to be assigned to the "un- neotropical freshwater fish systematics and specialized" genus. Like the subfamily has been consciously dealt with in other Trichomycterinae, the genus Trichomycterus groups, for example the Curimatidae (e.g., (even if T. boylei, T. santaeritae, T. hase- Vari, 1987). But being aware ofthis situation mani, and T. johnsoni are withdrawn) is makes evident the relevance ofincluding new probably nonmonophyletic. untraditional characters as part of normal Howes (1983: 331) suggested that the ex- taxonomic descriptions, as has been done in panded and dorsally concave palatine of as- trichomycterids, for example, by Arratia and troblepids could be a synapomorphy joining Menu-Marque (1984). this family with Trichomycterus rivulatus, It is probable that the situation verified in which would be reason to question mono- the Trichomycterinae and Trichomycterus is phyly of Trichomycterus. Although I ob- also the case in various other groups of silu- viously agree with him about nonmonophyly roids. A remarkably similar case has been of Trichomycterus, I do not agree with his reported in the neotropical family Loricari- justifications for this conclusion. Considering idae. Schaefer (1987) noticed that the lori- one species of Trichomycterus as the sister cariid subfamily Hypostominae has been tra- group ofastroblepids would imply nonmono- ditionally diagnosed by plesiomorphic phyly not only of the genus, but also of the features, and that no putative synapomorphy Trichomycteridae. This would implicate that is known to indicate monophyly ofthe group. all numerous synapomorphies for the family Similar situations can also be observed in (Baskin, 1973) are homoplastically distrib- taxa ranked as high as family. An enormous uted. Further, it would also imply that all and diverse group such as the Pimelodidae 38 AMERICAN MUSEUM NOVITATES NO. 2950

is not supported by a single synapomorphy Britski, H. A., and H. Ortega (see Lundberg et al., 1988), and monophyly 1983. Trichogenes longipinnis, novo genero e ofthe family remains an assumption not sup- especie de Trichomycterinae do sudeste ported by any evidence. Pimelodids seem to do Brasil. Revista Bras. Zool. 1(3): 211- be defined by nothing but symplesiomorph- 216. Britski, H. A., Y. Sato, and A. B. S. Rosa ies and lack of specializations characterizing 1986. Manual de identificagio de peixes da other neotropical catfish families. regiao de Tres Marias (com chaves de This state of affairs is actually part of a identificacao para os peixes da bacia do more general problem related to the inade- Sao Francisco). Brasilia: Camara dos quacy oftraditional taxonomic schemes when Deputados, Coordenagco de Publica- viewed under a phylogenetic perspective. A c6es-CODEVASF, Divisao de Pisci- broader overview of this subject is given by cultura e Pesca, 2nd ed. Weitzman and Fink (1983), in their discus- Cussac, V. E., and M. C. Maggese sion of classificatory schemes of New World 1988. Ontogeny of the maxillary barbels and Characiformes. palatine-maxillary mechanism in the catfish Rhamdia sapo (Pisces, Pimelodi- dae). Rev. Brasil. Biol. 48(2): 195-201. REFERENCES Dingerkus, G. and L. D. Uhler 1977. Enzyme clearing of alcian blue stained Arratia, G. whole small vertebrates. Stain Technol. 1983. The caudal skeleton of ostariophysan 52: 229-232. fishes (Teleostei): intraspecific variation Eigenmann, C. in Trichomycteridae (Siluriformes). J. 1918. The Pygidiidae, a family of South Morphol. 177: 213-229. American catfishes. Mem. Carnegie 1987. Description of the primitive family Mus. 7(5): 259-398. Diplomystidae (Siluriformes, Teleostei, Fink, S. V., and W. L. Fink Pisces): morphology, taxonomy and 1981. Interrelationships of the ostariophysan phylogenetic implications. Bonner Zool. fishes (Teleostei). Zool. J. Linn. Soc. 72: Monogr. 24: 1-120. 297-353. Arratia, G., and A. Chang Gosline, W. A. 1975. Osteologia de Pygidium areolatum Va- 1975. The palatine-maxillary mechanism in lenciennes, 1848 (Peces, Siluriformes, catfishes, with comments on the evo- Trichomycteridae). Mus. Nac. Hist. Nat. lution and zoogeography of modem si- Chile, Publ. Ocas. 18: 3-12. luroids. Occas. Pap. Calif. Acad. Sci. Arratia, G., and S. Menu-Marque 120: 1-31. 1981. Revision of the freshwater catfishes of Grande, L. the genus Hatcheria (Siluriformes, 1987. Redescription ofHypsidorisfarsonensis Trichomycteridae) with commentaries (Teleostei: Siluriformes), with a reason ecology and biogeography. Zool. Anz. sessment of its phylogenetic relation- 207(1/2): 88-111. ships. J. Vert. Paleontol. 7(1): 24-54. 1984. New catfishes of the genus Trichomyc- Howes, G. terus from the high Andes of South 1983. The cranial muscles of loricarioid cat- America (Pisces, Siluriformes) with re- fishes, their homologies and value as marks on distribution and ecology. Zool. taxonomic characters (Teleostei: Silu- Jahrb. Syst. 111: 493-520. roidei). Bull. Br. Mus. Nat. Hist. (Zool.). Arratia, G., A. Chang, S. Menu-Marque, and G. 45(6): 309-345. Rojas Lundberg, J. G., and J. N. Baskin 1978. About Bullockia gen. nov., Trichomyc- 1969. The caudal skeleton ofthe catfishes, or- terus mendozensis n. sp. and revision of der Siluriformes. Am. Mus. Novitates the family Trichomycteridae (Pisces, Si- 2398: 49 pp. luriformes). Stud. Neotrop. Fauna En- Lundberg, J. G., 0. J. Linares, M. E. Antonio, and viron. 13: 157-194. P. Nass Baskin, J. N. 1988. Phractocephalus hemiliopterus (Pime- 1973. Structure and relationships of the lodidae, Siluriformes) from the upper Trichomycteridae. Unpublished Ph.D. miocene Urmaco formation, Venezue- Dissertation, City University of New la: a further case of evolutionary statis York, New York. and local extinction among South 1989 de PINNA: SARCOGLANIDINE CATFISH 39

American fishes. J. Vert. Paleontol. 8(2): design: evolutionary morphology of 131-138. feeding mechanisms in loricarioid cat- Maddison, W. P., M. J. Donoghue, and D. R. fishes. Syst. Zool. 35: 489-508. Maddison Stewart, D., R. Barriga, and M. Ibarra 1984. Outgroup analysis and parsimony. Syst. 1987. Ictiofauna de la cuenca del Rio Napo, Zool. 33: 83-103. Ecuador Oriental: lista anotada de es- Myers, G. S. pecies. Politecnica-Revista de Infor- 1944. Two extraordinary new blind nematog- macion Tecnico-Cientifica, Quito, 12(4): nath fishes from the Rio Negro, repre- 9-63. senting a new subfamily of Pygidiidae, Tchernavin, V. with a rearrangement of the genera of 1944. A revision of some Trichomycterinae the family and illustrations ofsome pre- based on material preserved in the Brit- viously described genera and species ish Museum (Natural History). Proc. from Venezuela and Brazil. Proc. Cal- Zool. Soc. London 114: 234-275. ifornia Acad. Sci. 23: 591-602. Vari, R. P. Myers, G. S., and S. H. Weitzman 1987. Two new species ofcurimatid fishes (Os- 1966. Two remarkable new trichomycterid tariophysi: Characiformes) from Rio catfishes from the Amazon Basin in Grande do Sul, Brazil. Proc. Biol. Soc. Brazil and Colombia. J. Zool., London, Wash. 100(3): 603-609. 149: 277-287. Weitzman, S. H., and W. L. Fink Pinna, M. C. C. de 1983. Relationships ofthe neon tetras, a group 1988. A new genus of trichomycterid catfish of South American freshwater fishes (Siluroidei, Glanapteryginae), with (Teleostei, Characidae), with comments comments on its phylogenetic relation- on the phylogeny of New World char- ships. Rev. SuisseZool. 95(1): 113-128. aciforms. Bull. Mus. Comp. Zool. Schaefer, S. A. 150(6): 339-395. 1987. Osteology of Hypostomus plecostomus Weitzman, S. H., and R. P. Vari (Linnaeus), with a phylogenetic analysis 1988. Miniaturization in South American ofthe loricariid subfamilies (Pisces: Sil- freshwater fishes: an overview and dis- uroidei). Contrib. Sci. 394: 31 pp. cussion. Proc. Biol. Soc. Wash. 101(2): Schaefer, S. A., and G. V. Lauder 444-465. 1986. Historical transformation of functional Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are available free of charge. Address orders to: American Museum of Natural History Library, Department D, Central Park West at 79th St., New York, N.Y. 10024.

THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER.