JOURNAL OF MORPHOLOGY 240:169–194 (1999)

Comparative Study on the Cranial Morphology of Gymnallabes typus (Siluriformes: Clariidae) and Their Less Anguilliform Relatives, Clariallabes melas and Clarias gariepinus

ERIK CABUY,1* DOMINIQUE ADRIAENS,1 WALTER VERRAES,1 AND GUY G. TEUGELS2 1Institute of Zoology, University of Ghent, Ghent, Belgium 2Ichthyology, Department, Royal Museum for Central Africa, Tervuren, Belgium

ABSTRACT We compare the cranial morphology of four fish species with an increasing anguilliformism in the following order: Clarias gariepinus, Clarial- labes melas, Gymnallabes typus, and Channallabes apus. The main anatomical- morphological disparities are the stepwise reduction of the skull roof along with the relative enlargement of the external jaw muscles, which occurred in each of them. Gymnallabes typus and C. apus lack a bony protection to cover the jaw muscles. The neurocranial bones of C. gariepinus, however, form a closed, broad roof, whereas the width of the neurocranium in C. melas is intermediate. Several features of the clariid heads, such as the size of the mouth and the bands of small teeth, may be regarded as adaptations for manipulating large food particles, which are even more pronounced in anguil- liform clariids. The jaw musculature of G. typus is hypertrophied and at- tached on a higher coronoid process of the lower jaw, causing a larger adductive force. The hyomandibula interdigitates more strongly with the neurocranium and its dentition with longer teeth is posteriorly extended, closer to the lower jaw articulation. The anguilliform clariids also have their cranial muscles modified to enable a wider gape. The adductor mandibulae and the levator operculi extend more posteriorly, and the anterior attachment site of the protractor hyoidei dorsalis shifts toward the sagittal plane of the head. A phylogenetic analysis of the Clariidae, which is in progress, could check the validity of Boulenger’s hypothesis that predecessors of the primitive fishes, such as Heterobranchus and most Clarias, would have evolved into progres- sively anguilliform clariids. J. Morphol. 240:169–194, 1999. ௠ 1999 Wiley-Liss, Inc. KEY WORDS: Siluriformes; Clariidae; Clarias; Clariallabes; Gymnallabes; Channallabes; catfish

The external morphology of the Clariidae, air. These fishes have their greatest diver- or air-breathing catfishes that live in fresh- sity in Africa, but they also occur in Syria, water, is characterized by an elongated southern Turkey, and Southeast Asia (Teu- body with long dorsal and anal fins. An adi- gels, ’96). Anguilliform clariids such as pose fin, supported by elongated neural Gymnallabes, Dolichallabes, and Channall- spines, is present in Heterobranchus, Dinot- abes, however, have a more restricted distri- opterus, and two Clarias species (Teugels, bution and only occur in Central- and West ’83). The head of clariids is flattened dorso- Africa. ventrally and has small eyes and four Some African clariids can be ordered in pairs of barbels surrounding the terminal an orthogenetic series with an increasing eel-like form: Heterobranchus longifilis, Di- mouth. Clariids have a suprabranchial or- notopterus cunningtoni, Clarias poensis, gan formed by arborescent structures origi- nating from the second and fourth epibran- chials (Moussa, ’56; Greenwood, ’61; Munshi, *Correspondence to:Erik Cabuy, who is now at Laboratory for ’61; Vandewalle and Chardon, ’91). These Histology, University of Ghent, Louis Pasteurlaan 2, B-9000 organs enable them to utilize atmospheric Ghent, Belgium. E-mail: [email protected]

௠ 1999 WILEY-LISS, INC. 170 E. CABUY ET AL. Clariallabes melas, Clariallabes variabilis, situated between Nigeria and Cameroon Gymnallabes typus, Channallabes apus (Teugels, ’86a), which in part corresponds (Boulenger, ’07; Pellegrin, ’27). Heterobran- with the Lower Guinean ichthyofaunal prov- chus is positioned at the beginning of this ince recognized by Roberts (’75). The precise series. Its fusiform body is provided with a habitat of G. typus is not well known, but well-developed adipose fin and unpaired and river and lake banks (Matthes, ’64) as well paired fins (Teugels et al., ’90). A closed roof as marshes (Poll, ’42a; Gosse, ’63; Teugels et of thick bones makes up the neurocranium al., ’92) are places where it can be found. (Gregory, ’33: Fig. 77A; Teugels et al., ’90). Gymnallabes nops is caught in fast-running Channallabes apus, at the other end of the rivers (Roberts and Stewart, ’76). Channal- series, is the most extremely anguilliform labes apus, the only representative of its species. Its dorsal and anal fins are rela- genus, is distributed in the Congo system tively much larger and are confluent with (Teugels, ’86a). They have been caught in the caudal fin, whereas the adipose fin and the river silt or under vegetation in nearly the paired fins are absent. The neurocra- dried marshes by fishermen searching for nium is narrow and the lateral cranial bones worms (Poll, ’59). are strongly reduced (Boulenger, ’07, ’11; Anguilliform clariids occupy an appar- David, ’35; Poll, ’42a, ’77). ently different habitat and would be adapted After Boulenger (’07) and Pellegrin (’27), to a burrowing way of life (Nelson, ’94). many other authors became inspired by this Their anatomy has hardly been studied at regressive evolution among the clariids, as all to the present, and no attempt has been they called it (David, ’35; Poll, ’41, ’42a, ’77; made to give a detailed description. Their Greenwood, ’56, ’61; Lambert, ’60). David external morphology is very similar, but the (’35) has described this regressive trend in cranial skeleton is an important determina- detail in the first comprehensive anatomical tion factor to distinguish them from each and systematic study on clariids. She consid- other. This is clearly shown in the present ered the eel-like clariids as more developed, study when two out of the seven specimens, specialized catfishes. According to Poll (’77), which had originally been determined as the anguilliform body would have evolved Gymnallabes typus, revealed different osteo- many times independently. logical characters. In addition, an intraspe- Boulenger’s orthogenetic series (’07) is cific variation is noted in the remaining five partly represented here by examining four specimens. It is to be expected, therefore, different species: Clarias gariepinus, Clari- that in the future, more or less species may allabes melas, Gymnallabes typus, and Chan- be classified under the genus Gymnallabes. nallabes apus. The genus Clarias covers the This report mainly addresses the skeleton greatest number of species, i.e., 32, of which and cranial muscles because differences re- C. gariepinus has an almost Pan-African lating to them are more pervasive. A basic distribution (Teugels, ’82; Skelton and Teu- comparative osteology and cranial myology gels, ’92). Clariallabes numbers 15 species, between Gymnallabes typus and Clarias of which C. melas is distributed in the Congo gariepinus is given. Both species are at the system (Teugels, ’86a). Clariallabes tends to opposite ends of the series, i.e., anguilliform be associated with swift water and coarse vs. fusiform, respectively. Gymnallabes ty- rocky substrates where it can hide pus is used as a starting point in the descrip- (Winemiller and Kelso-Winemiller, ’96). This tion. Clarias gariepinus is taken instead of genus has not been studied thoroughly and Clarias poensis (ϭ C. camerunensis [Teu- the precise separation between this genus gels, ’86a]) of the series of Boulenger be- and Clarias is, therefore, open to argument. cause much more information on it is avail- The formerly defined subgenus Clarias (Alla- able. Intermediate forms on the basis of the benchelys) (David, ’35), which recently has external morphology between G. typus and been classified under the genus Clarial- C. gariepinus can be easily found among the labes, consists of intermediate forms rang- Clariidae. Clariallabes melas is used here as ing between Clariallabes and Clarias (Teu- such an example; Channallabes apus is dealt gels, ’86b). The genus Gymnallabes is with only in the Discussion. currently known to consist of three species: The aim of the present work is: (1) to give G. typus (Gu¨ nther, 1867), G. alvarezi (Ro- the main functional-morphological dispari- man, ’70), and G. nops (Roberts and Stewart, ties between the anguilliform Gymnallabes ’76). The area of distribution of G. typus is typus, the fusiform Clarias gariepinus, and CRANIAL MORPHOLOGY IN CLARIIDAE 171 the intermediate Clariallabes melas, and (2) on the cranium, hereafter named ‘‘cranium to comment on the functional implications of length,’’ is taken from the rostral tip of the the larger jaw muscles of anguilliform clari- premaxilla to the rostral side of the fourth ids. This study is also meant to encourage parapophysis instead of the caudal end of similar additional observations on Clariidae the supraoccipital process (see Discussion). or of other animal groups to unravel possible The body depth at anus, expressed as % TL, convergences. indicates the degree of anguilliformism of the specimen. Stomach analyses were car- MATERIALS AND METHODS ried out on all specimens except those that The material examined in the present were commercially reared (all C. gariepinus) study belongs to the Department of Ichthyol- or kept in captivity (two of the seven ogy, Koninklijk Museum voor Midden- G. typus). Where not mentioned in the text, Afrika (KMMA) Tervuren, Belgium, and the relative lengths are expressed in relation to University of Ghent. The research on Clar- body length. The material was cleared and ias gariepinus (Burchell 1822), Clariallabes stained for bone following Taylor and Van melas (Boulenger 1887), Gymnallabes typus Dyke (’85). Vertebral counts exclude those of (Gu¨ nther 1867), and Channallabes apus the Weberian apparatus and the compound (Gu¨ nther 1873) was carried out on a total of caudal centrum. Although the Weberian cen- 18 alcohol- preserved specimens. Only speci- trum consists of the fusion of the first five mens of which the most complete data were vertebrae (Rademaker et al., ’89), the num- gathered are listed in Table 1 , i.e., Clarial- bering of the remaining vertebrae in this labes melas: (KMMA 38495–508) 114–186 study starts from 1. The last abdominal ver- mm TL, Kunangu; G. typus: (KMMA 92–083- tebra is interpreted here as the one located P-0035–0036) 217 mm TL, (KMMA 91–067– above the anus. 0135–0136) 238 mm TL, (KMMA RG84–16- The cranial muscles have been studied P-1–2) 167 mm TL, Nigeria; C. apus: (KMMA only by means of dissection on a 167 and 239 90–29-P-151–163) 273–274 mm TL; Zaire. mm TL Gymnallabes typus, a 186 mm TL All other specimens in Table 1 are commer- Clariallabes melas, and a 166 mm TL Clarias cially bred. A dry cranial skeleton of an adult gariepinus. Stated differences between the C. gariepinus (194 mm cranium length) was three species and interpretations about the compared with the juveniles of C. gariepi- function of the muscles are based on these nus. Live specimens of G. typus and C. gari- dissections. The cranial muscles of Carias apus epinus (commercially obtained) were also could not be examined due to the poor state observed in a 90 L aquarium. Specimens of preservation. We have adopted the osteo- were sexed by examination of gonads, logical and myological terms used by Adri- whereby no differences were noted in the aens and Verraes (’96, ’97a,b,c, ’98) and Adri- urogenital papilla. All measurements of fresh aens et al. (’97). They provided a thorough and alcohol specimens are expressed in mil- account of the ontogeny of the cranial osteol- limeters. These consist of the total length ogy and myology of C. gariepinus and can be (TL), preanal length, body depth at anus, consulted for those aspects of the cranium, and cranium length. The distance measured which are not mentioned here.

TABLE 1. Morphometric (in mm) and meristic characters of four clariid species Clarias gariepinus Clariallabes melas Gymnallabes typus Channallabes apus Sex m m mm f mmmmm m f Total length 173 166 117 114 186 217 238 167 214 226 273 274 Preanal length 75 74 41 41 64 65 68 48 58 65 69 63 Body depth at anus 21.4 22 13.4 11.9 20.9 14.2 15 10.5 12 11.5 13.8 13.8 Cranium length 38.0 35.6 18.5 17.6 28.6 23.3 23.7 19.6 23.8 23.0 20.2 19.8 Dorsal fin rays 74 69 96 97 103 102 99 105 109 99 154 155 Anal fin rays 59 58 82 82 86 87 86 87 94 84 51 ϩ 561 133 Caudal fin rays 29 24 21 21 21 16 18 18 18 19 15 15 Pectoral fin rays 8/8 deformed 8/9 7/8 8/8 7/7 7/7 7/7 7/7 7/7 0 0 Pelvic fin rays 6/5 5/5 6/6 6/6 0/6 6/6 6/6 6/6 6/6 6/6 0 0 Vertebrae 59 59 65 69 71 78 76 76 80 78 108 108 Ribs 12/12 12/12 10/11 11/11 10/11 10/9 9/9 9/9 10/9 8/8 18/18 18/18 Branchiostegal rays 9/9 9/9 9/8 9/9 8/9 10/9 11/11 9/9 9/9 10/11 8/8 8/9

1One of the two examined C. apus exhibited two separate ventral median fins. 172 E. CABUY ET AL. Drawings were made with a camera lucida ful tail. The hypertrophied jaw muscles, attached to a stereomicroscope. Areas filled which characterize its peculiar head (Fig. with open circles on the drawings represent 1A,C), are not covered by skull bones as in cartilage. Clarias gariepinus. The terminal, broad mouth is surrounded by large fleshy upper RESULTS and lower lips, which are fused with each External morphology other by a skin fold at the mouth corners. Gymnallabes typus is an agile, brown- The size of the head, in relation to body colored, eel-like fish of which the most strik- length, equals approximately half the size of ing feature is its wide gape and long, power- the head of C. gariepinus and increases

Fig. 1. External morphology of the head of Gymnal- nostril; mnd-b-ex, external mandibular barbel; mnd-b- labes typus. A: dorsal; B: ventral view of the head (167 in, internal mandibular barbel; mx-b, maxillary barbel; mm TL); C: rostral view (217 mm TL). a-n, anterior ns-b, nasal barbel; p-n, posterior nostril. CRANIAL MORPHOLOGY IN CLARIIDAE 173 gradually from G. typus to C. gariepinus species. They have the same length in rela- (Fig. 2A–C). The eyes are small and posi- tion to the skull size, but are slightly broader tioned dorsally when the mouth is closed. in C. gariepinus. The pelvic fins are abdomi- They become gradually larger and more lat- nal in position, ventrally located, and slightly erally positioned in the series from G. typus anterior to the anal fin. to C. gariepinus. The body of Gymnallabes typus is exceed- Cranial skeleton ingly elongate, ovoid behind the head, and The overall skull of Gymnallabes typus is laterally compressed caudally. The body relatively less broad, compared with the depth at the anus is very small. The degree other species. Also, the neurocranium of G. of anguilliformism varies between 5.1% and typus (Fig. 3A) is narrow, whereas that of 6.5%. The other two species in the series Clarias gariepinus (Fig. 3C) is much broader. become gradually less elongated. The degree The relative size of the neurocranial bones of of anguilliformism of Clariallabes melas is Clariallabes melas (Fig. 3B) is in between ϳ11%, whereas that of Clarias gariepinus is those of the other two species, except for 13%. The dorsal and anal fins of G. typus are some rostral bones (mesethmoid, nasal, ant- confluent with the caudal fin. The dorsal fin orbital), which are nearly identical with those starts approximately at a distance from the of G. typus. The relative size of the skull supraoccipital process, which equals the cra- increases: G. typus has approximately an nial length. The beginning of the anal fin is average cranial length of 11% TL, C. melas located at approximately three times the 16% TL, and C. gariepinus 22% TL. Histo- length of the skull starting from the tip of logical observations revealed that some cra- the head. The median fins of C. gariepinus nial bones of G. typus are substantially less are separated from the caudal fin, but less in spongious than those in C. gariepinus. C. melas. They start gradually more posteri- The tubulous nasal bone of Gymnallabes orly in the series. The pectoral and pelvic typus and Clariallabes melas, the anterior- fins are present (except in one C. melas of most bone with a supraorbital canal of the which the right pelvic fin is missing). The lateral line system covering the nasal sac pectoral fins are positioned at the posterior longitudinally, is much more reduced com- ventrolateral side of the head in the three pared with that of Clarias gariepinus. The

Fig. 2. Lateral view of the skeletons of A: Gymnal- between the left and right marked vertebra, equaling an labes typus (214 mm TL); B: Clariallabes melas (114 mm arbitrary number of 47 vertebrae, is approximately the TL); C: Clarias gariepinus (173 mm TL). The body sizes same for the three species, illustrating that the indi- are scaled to equal each other in length. The length vidual vertebrae do not lengthen in this region. 174 E. CABUY ET AL.

Fig. 3. Dorsal view of the skull of A: Gymnallabes hm, hyomandibula; io-II, infraorbital II; io-III, infraor- typus (214 mm TL); B: Clariallabes melas (114 mm TL); bital III; io-IV, infraorbital IV; lac, lacrimal; leth, lateral C: Clarias gariepinus (173 mm TL). The skull of C. ethmoid; meth, mesethmoid; mnd, mandible; mx, max- gariepinus is asymmetrical because the right suprapreo- illa; ns, nasal; op, opercle; par-soc, parieto-supraoccipi- percle is absent. The vacated territory is taken over by tal; pf, posterior fontanel; pop, preopercle; pp-v4, par- the neighboring bones: the fourth infraorbital, the sphen- apophysis of vertebra 4; pp-v5, parapophysis of vertebra otic and mainly the pterotic. The distance between the 5; prmx, premaxilla; pt, pterotic; p-trc, transscapular tip of the premaxilla and the caudal side of the braincase process; pt-scl, posttemporo-supracleithrum; q, quad- (exoccipital) is scaled to the same length. af, anterior rate; sph, sphenotic; spop, suprapreopercle. fontanel; ant, antorbital; apal, autopalatine; fr, frontal; nasal bone is connected to the surrounding able as the four separate bones observed in bones (mesethmoid and lateral ethmoid) with Clarias gariepinus. The fourth infraorbital connective tissue, which allows a dorsoven- of G. typus is the largest and is situated tral expansion of the nasal cavity. posterolaterally of the lateral ethmoid and The mesethmoid of Gymnallabes typus and anterolaterally of the frontal from which it is Clariallabes melas is narrower behind the separated by connective tissue. This bone two rostral wings compared with that of partly covers the jaw muscle anteriorly. It is Clarias gariepinus. Because the lateral eth- sometimes broken in two parts, either on moid of G. typus does not make any contact one or on both sides of the cranium. Charac- with the second infraorbital bone, as is found teristic for the other infraorbitals of G. typus in the other two species, the articulation is that they consist of only a neurodermal ridge is absent. The lateral ethmoid of component, i.e., the part surrounding the G. typus does not make any contact with the sensory canal (Adriaens et al., ’97). In addi- fourth infraorbital bone either and is nar- tion, they lie far from each other with tiny rower posteriorly to its lateral process. fragmented bones bridging the gaps be- The antorbital is a very small bone that tween them. The second infraorbital bone surrounds the base of the nasal barbel ante- lacks a medial ridge to support the eye and riorly with its posteriorly concave shape. It does not articulate with the lateral ethmoid, is situated above the rostral tip of the auto- as in the other two species. palatine, close to the articulation with the The dermal frontals are the largest bones maxilla. The infraorbital canal of Gymnal- of the skull roof in all three species. The labes typus does not run through the antor- frontals of Gymnallabes typus and Clarial- bital bone as in Clariallabes melas and Clar- labes melas reach at their full length nearly ias gariepinus. the same width, whereas those of Clarias The infraorbitals of Gymnallabes typus gariepinus are broader in the middle. The are strongly reduced, but are still identifi- relative width of the frontals gradually wid- CRANIAL MORPHOLOGY IN CLARIIDAE 175 ens in the series from G. typus to C. gariepi- preopercle laterally (see left side of the aber- nus. At midline of the skull, the contralat- rant skull in Fig. 3C). Anterolaterally of the eral frontals abut each other whereas sphenotic in G. typus and C. melas, two long they are separated in the front by the ante- processes descend to interdigitate with the rior fontanel for approximately one-third hyomandibula (Fig. 3A,B). Posterolaterally, (G. typus) or one-half (C. gariepinus) of their a cartilaginous ridge articulates with the entire length. As in most catfishes, this fon- hyomandibula. The sphenotic of C. gariepi- tanel reaches the mesethmoid. The anasto- nus has only one process, which is situated mosis between the supraorbital and the in- on the ventral side (Adriaens and Verraes, fraorbital canals shifts posteriorly from ’98: Fig. 21E). The sphenotic of G. typus and G. typus to C. gariepinus.InG. typus and C. melas carries the otic canal that runs C. melas, the infraorbital canal branches off from mediorostral to mediocaudal, whereas from the supraorbital in the frontal bone. In the sphenotic of C. gariepinus carries the C. gariepinus, this bifurcation takes place in anastomosis between the supra- and infraor- the sphenotic bone (Adriaens et al., ’97: bital canals. Fig. 4A). Similarly, the pterotic of Gymnallabes ty- The sphenotic of Gymnallabes typus and pus curves down, partly forming a dorsal Clariallabes melas differs substantially from attachment site for the jaw muscle. This that of Clarias gariepinus. The sphenotic of bone is nearly horizontal in Clarias gariepi- the former species partly forms the sidewall nus, whereas that of Clariallabes melas is of the neurocranium, whereas that of intermediate in form. Anterolaterally, pter- C. gariepinus lines the skull roof dorsally. otic processes interdigitate with the pos- The sphenotic of G. typus interdigitates ros- terodorsal side of the hyomandibula of trally with the frontal, to which it abuts G. typus (Fig. 3A). The number of processes laterally, and caudally with the pterotic. In decreases from G. typus (3), C. melas (2), to C. gariepinus, this bone also makes contact C. gariepinus (0). The pterotic becomes with the fourth infraorbital and the supra- gradually broader in the series from G. ty-

Fig. 4. Ventral view of the neurocranium of A: Gym- mp, metapterygoid; mx, maxilla; osph, orbitosphenoid; nallabes typus (217 mm TL); B: Clariallabes melas (117 para, parasphenoid; pop, preopercle; pp-v4, parapophy- mm TL). The entopterygoid of C. melas is half the size of sis of vertebra 4; pp-v5, parapophysis of vertebra 5; the bone in the other two specimens. C: Clarias gariepi- prmx, premaxilla; prot, prootic; psph, pterosphenoid; nus (173 mm TL). The prevomeral tooth plates are fused p-trc, transscapular process; pt, pterotic; pt-scl, posttem- with each other in adult C. gariepinus. apal, autopala- poro-supracleithrum; pvm, prevomer; q, quadrate; sph, tine; boc, basioccipital; enp, entopterygoid; eoc, exoccipi- sphenotic; spop, suprapreopercle. tal; fr, frontal; hm, hyomandibula; leth, lateral ethmoid; 176 E. CABUY ET AL. pus to C. gariepinus.. Posterolaterally, con- terodorsal processes articulating with the nective tissue links the pterotic of G. typus pterotic. The first anterodorsal process is to the much more reduced suprapreopercle. always the largest, the second is smaller, On the pterotic of G. typus, a medial ridge is and there may be a smaller third one. The present for the attachment of the jaw muscle. posterodorsal processes grow larger toward The parieto-supraoccipital is situated me- the back. The hyomandibula of C. gariepi- dioposteriorly of the skull roof and is charac- nus (Fig. 5C) has only one anterodorsal pro- terized by a pointed, posteriorly directed cess. The cartilaginous articulatory ridge of process. In some specimens of Clariallabes G. typus runs parallel with the midline of melas and Gymnallabes typus, this bone is the skull, but is obliquely oriented in divided caudal to the posterior fontanel. In C. gariepinus and C. melas (Fig. 4A–C). The one G. typus, the parieto-supraoccipital even hyomandibula shifts gradually anteriorly in consists of two symmetrical bones (Fig. 3A). the series from G. typus to C. gariepinus. The caudal part of this bone gradually The caudoventrally directed opercular pro- lengthens in the series from G. typus to cess of G. typus, which articulates with the C. gariepinus. In dorsal view, the supraoccipi- opercle, is situated more posteriorly. The tal process of G. typus almost reaches the morphology of the quadrate shows no fourth vertebra. The parieto-supraoccipital gradual trend in the three species, as the of C. gariepinus (Fig. 3C) is larger posteri- quadrate of C. melas is larger than in the orly and the supraoccipital process covers others. Its articulatory facet for the lower the sixth vertebra. jaw is situated most anteriorly. This facet is The posttemporo-supracleithrum, which inclined in C. gariepinus and C. melas, but makes contact with the pterotic medially, is not in G. typus. the posteriormost dermal, canal bone. The The pterygoid bones of the three species posttemporo-supracleithrum of Gymnal- differ substantially. In general, the morphol- labes typus is more reduced compared with ogy of these bones in the examined speci- that of Clariallabes melas and Clarias garie- mens is similar to those illustrated in Figure pinus. Its transscapular process makes only 4, except for the specimen of Gymnallabes contact with the rostral side of the fourth typus (Fig. 4A) where the right entoptery- parapophysis. goid was probably broken during ontogeny, The suspensorium of Gymnallabes typus as both parts together make up the entop- and Siluriformes in general consists of six terygoid in the other specimens. The meta- bones: hyomandibula, preopercle, quadrate, pterygoid is defined here as the pterygoid metapterygoid, entopterygoid, and autopala- bone, which interdigitates with the quad- tine (Fink and Fink, ’96). The latter, how- rate posteriorly. The anterior rim of the ever, develops separately from the pterygo- metapterygoid of G. typus (Fig. 4A) is perpen- quadrate in Siluriformes (Arratia, ’92; dicular with the long axis of the suspenso- Adriaens and Verraes, ’97b). The hyoman- rium, in contrast with Clarias gariepinus, dibula of the three species differs mainly in which has an inclined facet (Fig. 4C). The the number of dorsal processes that interdigi- medial ridge of the metapterygoid of G. ty- tate with the neurocranium. The neurocra- pus is concave and is not used for the attach- nial-hyomandibula articulation in G. typus ment of the adductor arcus palatini muscle, and C. melas is situated at the level of the as is the case in the other two species. The posterolateral part of the sphenotic and the metapterygoid of Clariallabes melas (Fig. 4B) anterolateral part of the pterotic (Figs. 3A,B; is remarkable in that it almost reaches the 4A,B). In Clarias gariepinus, the hyoman- lateral wing of the prevomer, thereby leav- dibula articulates with the ventral side of ing less space for the entopterygoid. The these bones (Fig. 4C). Because of the reduc- metapterygoid of C. gariepinus is situated tions of the sphenotic and pterotic in G. typus, laterad of the entopterygoid, unlike in the the dorsal articulation of the hyomandibula other two species. The entopterygoid of is situated more medially and the height of G. typus is posteriorly separated from the this bone is slightly increased. Dorsally, the quadrate by the metapterygoid, except later- hyomandibula of G. typus (Fig. 5A,B) can be ally. The entopterygoid of C. gariepinus is subdivided in three zones: two to three an- laterally separated from the quadrate, but terodorsal processes articulating with the contacts with it posteriorly. In C. gariepinus sphenotic; a cartilaginous ridge also articu- and G. typus, this bone is relatively large, lating with the sphenotic; and three pos- whereas it is small and narrow in C. melas. CRANIAL MORPHOLOGY IN CLARIIDAE 177

Fig. 5. A: Left lateral view of the skull of Gymnal- III; io-IV, infraorbital IV; iop, interopercle; lac, lacrimal; labes typus (214 mm TL). B: Detail of the upper part of mnd, mandible; mp, metapterygoid; mx, maxilla; ns, the hyomandibula of G. typus. C: Clarias gariepinus. nasal; op, opercle; osph, orbitosphenoid; pop, preopercle; ant, antorbital; apal, autopalatine; ch-a, anterior cerato- psph, pterosphenoid; prmx, premaxilla; pt, pterotic; pt- hyal; ch-p, posterior ceratohyal; enp, entopterygoid; hm, scl, posttemporo-supracleithrum; pvm, prevomer; q, hyomandibula; io-II, infraorbital II; io-III, infraorbital quadrate; sph, sphenotic; spop, suprapreopercle.

The entopterygoid forms an insertion area orly directed teeth, almost up to the coronoid for the adductor arcus palatini muscle in process. The dental tooth rows gradually C. gariepinus, but not in the others. shorten in the other two species, but are The relative size of the lower jaw is largest rostrally more numerous in C. melas. for Gymnallabes typus and shortest for Clari- Also, some ventrally situated bones differ allabes melas, but in the latter it is more between the species (Fig. 4A–C). The pre- massive. The coronoid process on the lower maxilla is gradually becoming shorter in jaw of these species is much more developed, the series from Gymnallabes typus to Clar- compared with Clarias gariepinus (Fig. ias gariepinus. The anterior ventral half 6A–C). The mandibular symphyseal surface of it in G. typus and Clariallabes melas is in G. typus and C. melas is also larger than covered with four to five rows of pointed, that in C. gariepinus. An additional differ- curved teeth that are directed posteriorly. ence between the lower jaws of these spe- The premaxillary tooth plate of C. gariepi- cies, is the position of the teeth. The anterior nus is covered more densely and with smaller lower jaw of G. typus is covered with posteri- teeth. The nasal sac of G. typus is much 178 E. CABUY ET AL.

Fig. 6. Medial view of the lower jaws of A: Gymnal- gariepinus, respectively (␣1 ϭ␣2). af-q, articulatory facet labes typus (217 mm TL); B: Clariallabes melas (117 mm of the mandible with the quadrate; ang-c, angulo-splenio- TL); C: Clarias gariepinus (173 mm TL). The lower jaws articulo-retroarticular complex; c-Meck, cartilago Meck- are scaled to the same length. The higher coronoid eli; com, coronomeckelian bone; den-c, dento-splenio- process of G. typus (␤1 Ͼ␤2) results in a larger power mentomeckelium complex; mnd-sym, mandibular arm or input length Li1 and vertical force vector Fcp1 of symphysis; p-co, coronoid process; p-ra, retroarticular the jaw adductor muscle compared to Li2 and Fcp2 of C. process. more supported by this bone than in C. gari- prevomeral tooth plate. Straight conical epinus. The axe-shaped prevomer abuts teeth cover the prevomeral tooth plate of G. the mesethmoid ventrally and interdigitates typus and C. melas. The prevomeral tooth caudally with the parasphenoid through a plate of C. gariepinus is packed more densely long pointed spine. The prevomer of C. garie- with smaller teeth. Here, a distinction can pinus has a bifurcated spine. The pre- be made between granular teeth centrally vomeral tooth plate of G. typus is laterocau- and villiform teeth at the lateral edge. The dally more extended than that of C. parasphenoid, the largest part of the neuro- gariepinus, even touching the entoptery- cranial floor, is broader in C. gariepinus.Itis goid. The left entopterygoid of one examined striking to see that the braincase of C. melas G. typus is even partly covered with teeth, is much broader compared to the other two probably as a result of its fusion with the species. CRANIAL MORPHOLOGY IN CLARIIDAE 179

Gymnallabes typus is further specialized A2A3Ј by the levator arcus palatini and is in having the anterior ceratohyal and the situated against the caudolateral part of the dorsal and ventral hypohyal greatly ex- retractor tentaculi. The A3Љ muscle of Gym- panded, forming a broad surface for the ori- nallabes typus can also be subdivided, but in gin of the hyohyoideus inferior muscle. The a lateral pars superficialis and a medial pars ventral hypohyal of G. typus is posteroven- profunda. The adductor mandibulae A3Љ pars trally more elongated than in Clarias gariepi- superficialis, which lies against the medial nus. The dorsal hypohyal is much smaller side of the levator arcus palatini, forms the than the previous bone but both are larger biggest part of the A3Љ muscle. Posteriorly, than those of C. gariepinus. No major differ- the A3Љ is connected with the ventrolateral ences were observed in the bones of the ridge of the frontal, just under the otic canal, branchial basket and in the pharyngeal tooth the anterolateral side of the sphenotic, and plates of the three species. the lateral side of the pterosphenoid and Cranial myology parasphenoid. On the suspensorium, this part of the A3Љ is attached to the caudal part The adductor mandibulae of Gymnallabes of the quadrate and the anteromedial cavity typus is an enormous muscle, covering nearly of the hyomandibula. The pars profunda is the entire dorsolateral side of the skull. This Ј caudally attached to the anteromedial mem- muscle is formed by an external A2A3 -part branous plate of the hyomandibula, to which and an internal, smaller A Љ-part. One part 3 the retractor tentaculi is connected more of A3, hereafter referred to as A3Ј, is indistin- anteriorly. Rostrally, A3Љ inserts tendinously guishably fused with A2, whereas the medial Љ on the dorsomedial side of the angular com- part of A3, hereafter referred to as A3 , forms Ј a separate muscle (Adriaens and Verraes, plex, medially of the A2A3 tendons. Simi- ’96). larly, the direction of the muscle fibers of the internal adductor mandibulae ranges from a The bipinnate A2A3Ј muscle of Gymnal- labes typus is nonhomogenous in that the horizontal to a vertical position. The same fiber length and fiber angle differ. The fibers bones of insertions of the adductor mandibu- are arranged at mixed angles on the sheet- lae in G. typus are found in Clariallabes like tendon with the posterior fibers being melas. Compared to Clarias gariepinus, the Ј shorter and more inclined. The A A Ј (Fig. adductor mandibulae A2A3 of G. typus is 2 3 greatly expanded. The frontal, pterotic, and 7A) can be subdivided into a dorsal A2A3␣Ј- posttemporo-supracleithrum bones form new part and a ventral A2A3␤Ј-part, which are both attached to a broad aponeurosis con- attachment sites. The insertions and mor- Љ nected to the coronoid process. The upper phology of the adductor mandibulae A3 , how- part, A2A3␣Ј, is attached to the ventral side ever, are more or less equal to those of of the fourth infraorbital, the complete ven- C. gariepinus. However, the deeper part of Љ trolateral ridge of the frontal, the dorsal side the A3 in G. typus, the adductor mandibulae Љ of the sphenotic, almost the fully dorsal side A3 pars profunda, is completely separated of the pterotic, the medial side of the most from the pars superficialis. caudal suprapreopercle, and the rostral part The levator arcus palatini is a flat muscle of the posttemporo-supracleithrum. Some ex- that connects the ventrolateral side of the ternal fibers are attached to connective tis- neurocranium with the suspensorium (Fig. sue under the skin where in Clarias gariepi- 7B) and separates A2A3Ј posteriorly from the Љ nus the fourth infraorbital is situated. A2A3Ј␤ A3 muscle. In Gymnallabes typus, the leva- is posteriorly musculously connected with tor arcus palatini is mainly attached to the the lateral side of the suspensorium: the ventrolateral side of the frontal, but also on lowermost part of the hyomandibula, the the anterolateral part of the sphenotic and preopercle, and the caudal side of the quad- the caudolateral part of the lateral ethmoid. rate. The muscle fibers of the external adduc- The levator arcus palatini of Clarias gariepi- tor mandibulae run in such a direction that nus (Adriaens and Verraes, ’97a: Fig. 7B) they form a transition, starting from the extends more to the front and covers the horizontal fibers of the A2A3Ј␤ into the verti- complete ventrolateral side of the latter bone. cally directed fibers of A2A3Ј␣. With regard to the suspensorium, there is a The deeper part of the adductor mandibu- partly muscular and aponeurotic connection lae, A3Љ (Fig. 7C), is connected with the neu- on the rostral crest of the hyomandibula, rocranium and the suspensorium as well. whereas in C. gariepinus, the attachment is This muscle is dorsocaudally separated from only aponeurotic. The sphenotic of G. typus 180 E. CABUY ET AL.

Figure 7 CRANIAL MORPHOLOGY IN CLARIIDAE 181 forms an additional insertion area, whereas from the posteroventral part of the frontal, an attachment on the fourth infraorbital the lower lateral part of the sphenotic and and the quadrate, as is the case in C. gariepi- the upper lateral part of the hyomandibula. nus, is missing. Contraction of this muscle The muscle fibers run mainly dorsoventrally abducts the suspensorium and causes an and insert tendinously on the dorsal process expansion of the orobranchial cavity. of the opercle, close to the articulation with The adductor arcus palatini, the antago- the hyomandibula. This muscle is shorter nist of the previous muscle, connects the and not attached to the lateral ethmoid as in neurocranium with the dorsal rim of the Clarias gariepinus, where the fibers are more suspensorium (Fig. 7D), lining the mouth horizontally oriented (Adriaens and Ver- cavity laterodorsally. The adductor arcus raes, ’97a: Fig. 7C). palatini lies medially of the caudal part of The adductor operculi (Fig. 7C) is a thicker, the retractor and extensor tentaculi. The shorter muscle connecting the dorsocaudal muscle fibers of the adductor in Gymnal- ridge of the hyomandibula with the opercle. labes typus originate mainly on the lateral In Clarias gariepinus, this muscle lies in a side of the parasphenoid, but also on the cavity of the hyomandibula, which is absent pterosphenoid and the orbitosphenoid. On in Gymnallabes typus. Dorsocaudally this the suspensorium, the muscle is inserted on muscle is also attached to connective tissue, the medial side of the hyomandibula rostral which covers the dorsolateral side of the plate and on the full medial edge of the suprabranchial cavity. The adductor oper- quadrate. The adductor arcus palatini con- culi is covered anteriorly by the dilatator sists of sparsely tendinous parts, which are operculi. The adductor inserts musculously most prominent in Clariallabes melas, run- on the medial side of the opercular dorsal ning parallel with the muscle fibers. The process, caudolaterally of the opercular pro- fibers run in a transversal plane, with the cess of the hyomandibula, as is the case in rostral fibers more obliquely oriented. The C. gariepinus. The muscular insertion of this metapterygoid and entopterygoid are addi- muscle and the tendinous insertion of the tional attachment sites in Clarias gariepi- nus, the former bone only for C. melas. dilatator operculi on the opercle seem to be The opercular muscles comprise three fused with each other. muscles. The flattened dilatator operculi, The levator operculi is the most solid of being the most rostral one, covers the upper the three opercular muscles. In Gymnal- lateral side of the hyomandibula. The ros- labes typus (Fig. 7C), this muscle originates tral part of the dilatator operculi is situated on the caudolateral edge of the pterotic, the complete rostral rim of the posttemporo- between the adductor mandibulae A3Љ and the levator arcus palatini. The dilatator oper- supracleithrum, and partly on the connec- culi of Gymnallabes typus (Fig. 7C) starts tive tissue lining the dorsolateral wall of the suprabranchial cavity. The posttemporo- supracleithrum of G. typus forms a new at- tachment site compared with the other two Fig. 7. Lateral view of the head musculature of Gym- species and results in a more inclined leva- nallabes typus (238 mm TL). A: Skin and barbels are tor muscle. Ventrally, the levator operculi removed. B: External jaw muscle is removed. C: Levator inserts almost on the entire upper ridge of arcus palatini is removed. D: Internal jaw muscle and the opercle. opercular muscles are removed. The caudal part of the retractor tentaculi is cut away. apal, autopalatine; ch-a, Ventrally situated cranial muscles show anterior ceratohyal; enp, entopterygoid; fr, frontal; hm, peculiar differences as well. The hyoid bars Ј hyomandibula; iop, interopercle; m-A2A3 , external part are connected with the lower jaws by a large, Љ of the adductor mandibulae muscle; m-A3 , internal part compact muscle, the protractor hyoidei. At of the adductor mandibulae muscle; m-ad-ap, adductor arcus palatini muscle; m-ad-op, adductor operculi muscle; the ventral face of the head, rostrally of the m-dil-op, dilatator operculi muscle; m-ex-t, extensor ten- isthmus, this muscle is seen as a U-shape taculi muscle; m-l-ap, levator arcus palatini muscle; (Fig. 1B). The ventral side of the lower jaw, m-l-op, levator operculi muscle; mnd, mandible; mnd-b- except caudally, is completely covered by it. ex, external mandibular barbel; mnd-b-in, internal man- dibular barbel; mp, metapterygoid; m-pr-h, protractor The protractor consists of a ventral part hyoidei muscle; m-pr-p, protractor pectoralis muscle; (Fig. 8A), penetrated by the mandibular bar- m-re-t, retractor tentaculi muscle; m-sh, sternohyoideus bels, and a dorsal part (Fig. 8B). Caudally, muscle; mx, maxilla; mx-b, maxillary barbel; op, opercle; the muscle fibers originate on the ventrolat- pop, preopercle; pt, pterotic; pt-scl, posttemporo-su- pracleithrum; q, quadrate; sbo, suprabranchial organs; eral side of the anterior ceratohyal, lateral sph, sphenotic. to the attachment of the hyohyoideus infe- 182 E. CABUY ET AL.

Fig. 8. Ventral view of the head musculature of Gym- adductores muscles; m-intm, intermandibularis muscle; nallabes typus (238 mm TL). A: Skin and barbels are mnd, mandible; mnd-b-ex, external mandibular barbel; removed. B: Ventral part of the hyoid protractor is mnd-b-in, internal mandibular barbel; m-hh-inf, hyohyoi- removed. C: Hyoid protractor and connective tissue be- deus inferior muscle; m-pr-h-d, protractor hyoidei pars tween the jaws are removed. D: Hyohyoideus inferior is dorsalis muscle; m-pr-h-v, protractor hyoidei pars ventra- removed. ch-a, anterior ceratohyal; ch-p, posterior cera- lis muscle; m-sh, sternohyoideus muscle; puh, paruro- tohyal; cl, cleithrum; hh-v, ventral hypohyal; iop, inter- hyal; pvm-tpl, prevomeral tooth plate; r-br-VIII, bran- opercle; m-ad-mnd, adductor mandibulae muscle; m-hh- chiostegal ray VIII. ab, hyohyoideus abductor muscle; m-hh-ad, hyohyoidei CRANIAL MORPHOLOGY IN CLARIIDAE 183 rior. The left and right ventral parts are pared with Clarias gariepinus. The anterior rostrally connected with each other by an suprabranchial structures of G. typus are aponeurosis, which is tightly connected with either absent or consist only of one vesicle, the lower jaws. The dorsal part consists of whereas that of C. melas counts two to three two parts, which cannot be distinguished vesicles. The suprabranchial organs of from each other caudally. It is the medial C. gariepinus are most developed and reach part of the protractor hyoidei dorsalis of the skull more posteriorly. Gymnallabes typus that differs substan- Postcranial skeleton tially with that of Clarias gariepinus.Inthe latter, the contralateral parts of it are con- A comparison of the three species with nected with each other in the medial plane each other reveals that tail length is allome- by an aponeurosis (Adriaens and Verraes, tric. The mean number of caudal vertebrae ’97c: Fig. 6B). In G. typus, however, the two increases from 38 (Clarias gariepinus)to64 parts are not connected with each other, but (Gymnallabes typus). The number of abdomi- are attached to the caudal side of the lower nal vertebrae is approximately the same for jaws. Clariallabes melas and G. typus, but slightly Another hyoid muscle of Gymnallabes ty- more for C. gariepinus. Gymnallabes typus pus, the hyohyoideus inferior (Fig. 8C), shows (Fig. 2A) exhibits 76 to 80 vertebrae (maxi- a different morphology. This muscle covers mum 17 abdominal and 64 caudal verte- the ventral side of the anterior ceratohyal brae). These numbers are lower than those and the ventral and dorsal hypohyal, and of the G. typus specimens (123 mm and 195 consists of a rostral and a caudal part. No mm TL) described by Poll (’77), who counted such a distinction, however, is seen in Clar- 89 vertebrae, 15 of which are abdominal and ias gariepinus (Adriaens and Verraes, ’97c: 74 caudal. The urogenital porus (which is Fig. 8A). The contralateral fibers are medi- situated anteriorly of the pelvic girdle in one ally connected to an aponeurosis. The ros- aberrant specimen) is positioned ventrally tral part in G. typus is thicker and caudo- between vertebrae 14 to 17. The number of laterally not only connected with the abdominal vertebrae of G. typus is variable anteroventral side, but also with the lateral as observed in some other clariids. Clarial- side of the anterior ceratohyal. labes melas (Fig. 2B) exhibits 65 to 71 verte- Finally, the sternohyoideus of Gymnal- brae (maximum 17 abdominal and 54 caudal labes typus (Fig. 8D), which seems to be vertebrae), whereas C. gariepinus. (Fig. 2C) larger than in Clarias gariepinus, connects has 59 vertebrae (56 to 63 [Teugels, ’86b]) the pectoral girdle with the hyoid bars (maximum 21 abdominal and 40 caudal ver- through the parurohyal. The cleithrum bears tebrae) (Table 1). In addition, the abdominal a rostrally directed wing to which the sterno- vertebrae are more compressed in an antero- hyoideus is attached. This muscle is com- posterior direction in C. gariepinus, and least posed of three myomeres separated by two in G. typus. Articulation facets between the myocommata. The rostral myomere is the vertebrae are most pronounced in G. typus. largest, whereas in C. gariepinus, the maxi- The neural and haemal spines of G. typus mum length of the three myomeres is equal are more inclined toward the back with the (Adriaens and Verraes, ’97c: Fig. 8C). Fur- first haemal spine (left and right haemal thermore, the myocommata are positioned spine fused distally) originating between ver- less inclined than in C. gariepinus. tebrae 14 and 17. In C. melas (in two of the No differences were observed in the other three specimens) and in C. gariepinus, these cranial muscles of the three species, such as haemal spines are situated more in the front, the extensor and retractor tentaculi, the hyo- at vertebrae 12 or 13. From the first, second, hyoideus abductor, and the hyohyoidei ad- or third to the ninth, tenth, or eleventh ductores. vertebrae, 8 to 10 pleural ribs connect the lateral parapophyses of G. typus, occasion- Suprabranchial organs ally followed by a rudimentary one, floating The suprabranchial organs (Fig. 7D) con- in the abdominal wall. The parapophyses of sist of an anterior and a posterior structure, the second vertebra are the longest, whereas situated on the second and fourth epibranchi- those of the first vertebra are hardly devel- als, respectively. The anterior structure is oped (Fig. 3A). The parapophyses of the an- always smaller than the posterior one. Both terior abdominal vertebrae of C. gariepinus organs of Gymnallabes typus are much more are much larger than in the other two spe- reduced, but less in Clariallabes melas com- cies. Clariallabes melas has 10 or 11 pairs of 184 E. CABUY ET AL. ribs and C. garienpinus 12. The distance dorsal and anal fin rays are only split at the between the pterygiophores and the verte- caudal part of the fin. These fin rays are not bral column increases, together with an in- divided to their base. creasing bogy height, in the series from G. DISCUSSION typus to C. gariepinus. One or two pterygio- phores are present between successive neu- The differences seen in the three species, ral and haemal spines. The number of pte- Gymnallabes typus, Clariallabes melas, and rygiophores, for instance for every 30 Clarias gariepinus, cannot be attributed to vertebrae, slightly decreases from G. typus the difference in developmental stage, al- to C. gariepinus. though the studied species represent fishes The number of dorsal and anal fin rays of of differing body lengths. The studied fishes Gymnallabes typus does not lie within the showed within their species essentially no specified intervals defined by Poll (’77), but obvious allometric growth in overall skull varies between 99 to 109 dorsal fin rays and morphology. The jaw muscles in a 167 mm 84 to 94 anal fin rays. Only the proximal TL G. typus, for instance, have a similar radials of the pterygiophores of both fins are morphology as in a 238 mm TL adult. There- ossified and osseous distal radials are ab- fore, it appears acceptable to compare spe- sent, except in some G. typus where tiny cies of a different length with each other on bones can be seen. The caudal fin of G. typus the basis of their skeleton and cranial myol- has a maximum total (principal plus procur- ogy. rent) of 19 fin rays. The maximum numbers The impact of the enlargement of the ad- ductor mandibulae and its effect on the over- of dorsal, anal, and caudal fin rays of Clarial- all skull is discussed. It is also considered labes melas observed is 97, 82, and 21, re- whether skull remodeling (narrowing of the spectively. For Clarias gariepinus, these neurocranium as a result of the increase of numbers are 74, 59, and 29. The caudal the jaw adductors) is related with body elon- skeleton of G. typus differs intraspecifically gation. in the number of hypurals fused with each other, ranging from two to all. The parhypu- Significance of larger jaw muscles ral is not fused with the ventral hypurals. in Gymnallabes typus The hypurals of the caudal skeleton of ob- When Gymnallabes typus is put in an served C. gariepinus are not fused with each aquarium without ground cover, it hastily other (except in one specimen), whereas they senses the bottom with its lips and barbels. are in C. melas. While it is swimming, it bends its head and The pectoral fins of Gymnallabes typus pushes its mouth against the bottom as if it have mostly a nonserrated spine and seven were searching for an opening. A burrowing fin rays that articulate with three proximal G. typus, however, has not been reported so radials of which the most anterior is less far in the literature and neither has it been ossified. The number of pectoral fin rays in seen during our observations, either in gravel Clariallabes melas and Clarias gariepinus or in sand. It should be feasible, however, is, respectively, 7–9 and 9–10. Their pectoral bearing in mind that the related but more spines are serrated at both sides. No major elongated Channallabes apus can burrow differences were seen in the pectoral girdles themselves in mud (Lambert, ’60; Matthes, of the three species, except that the girdle in ’64) or in gravel (personal observation). Al- C. melas consists of more massive bones and though the suprabranchial organs of G. ty- that the anterior rim of the cleithrum has a pus are considerably reduced compared with concave shape in G. typus. Similarly, the those of Clarias gariepinus, it still strongly Weberian apparatus in C. melas is more depends on atmospheric respiration in oxy- strongly built. The pelvic fins of all three gen-poor water. It was also seen in the species exhibit six fin rays, which articulate aquarium that G. typus, which were put with the basipterygium of the pelvic girdle; together in a group, seemed to be fond of osseous radials are absent. The morphology each other, rubbing their skins on one an- of the basipterygium is the same in the three other and nesting closely together on the species. The length of the paired and caudal bottom. A totally different behavior was ob- fins, in relation to body length, increases in served when only two individuals were kept the series, whereas the length of the dorsal together. They assumed quite aggressive and anal fins decreases. Paired and un- fights, which involved biting each other’s paired fins have bifurcated rays, but the tails and grasping movements of the mouth. CRANIAL MORPHOLOGY IN CLARIIDAE 185 The food habits of Gymnallabes typus have ond mode of feeding. Gymnallabes typus did not been reported yet. Only Matthes (’64) not manage to tear off pieces out of a big carried out a stomach analysis on one speci- chunk of salmon, which was presented free men of 75 mm. The stomach content con- in the aquarium, by shaking its head. It took sisted of some sand, one ostracod, one hydra- it in its wide-open mouth instead and re- carid, and a few insect larvae. Personal turned to its hiding place where it started stomach analyses on the specimens caught processing the food. This extensive meal re- in their natural habit revealed sand, insect sulted in an enormous swollen belly. Appar- larvae, snail shells, fish scales, termites, a ently, these fishes are well suited to this horsehair worm, and plant material. Al- kind of feeding behavior. The number of though they probably prey on anything that empty stomachs encountered could be ex- is edible, the short gut of G. typus indicates plained by the fact that G. typus do not have mainly a carnivorous diet. According to Mat- to feed continuously, because the large di- thes (’64), other eel-like clariids, such as mensions of the prey enable them to eat a Dolichallabes microphthalmus and Chan- considerable amount of food in a short pe- nallabes apus, are omnivorous. Presenting riod, reducing their need to forage. Another G. typus with live prey, such as mealworms explanation could be that captured speci- (Tenebrio molitor), fly larvae, water fleas mens were kept alive for a while before (Daphnia pulex), earthworms (Lumbricus being put in formalin, so that their stomach terrestris), zebra fish (Danio rerio), but also content was already digested. canned mackerel, was successful. When it Stomach analyses of Clariallabes melas was predating live fishes, voracious G. typus revealed only fish scales, whereas the stom- attained a high initial speed. It seemed that ach of Channallabes apus contained one food with an easily dispersible flavor, such grasshopper and plant material. Clarias as fresh salmon or canned mackerel, was gariepinus in general is also an omnivorous most favored, but whether they locate their scavenger whose food consists mainly of food primarily by cutaneous senses or by fishes, terrestrial invertebrates, aquatic in- olfaction is unknown. Sometimes, G. typus sects, and zooplankton (Groenewald, ’64; takes up sand, which it will spit out later, Clay, ’79). Spin feeding in C. gariepinus,as possibly partly filtered, explaining the pres- mentioned by Helfman and Clark (’86), is ence of some sand in the stomach. When the not a plausible feeding form for fish with a mouth gape is wide, the mouth folds are fusiform body. In general, it can be stated stretched and form a funnel, an advantage that in nature clariids tend to feed on almost for sucking up small invertebrates and pre- any food with a preference for animal mate- venting potential food from escaping. rial (Clay, ’79). The three basic feeding modes of the an- The adult skull of Gymnallabes typus guillids occur also in Gymnallabes typus, (Fig. 3A) shows characteristics that corre- i.e., 1. suction of food; 2. grasp and shake, spond with the other anguilliform clariids which entails grasping of larger items in the such as Channallabes apus (personal obser- mouth while shaking head and body, thereby vation) and Dolichallabes microphthalmus tearing off small pieces; and (3) rotational or (Poll, ’42b). There is no bony protection above spin feeding, which entails grasping of larger the jaw muscles, nor is there one between items in the mouth and spinning on the long the neurocranium and the parapophyses of axis, thereby tearing larger prey items into the fourth vertebra. The nasal sac is partly smaller pieces (Helfman and Clark, ’86). The covered by the nasal bone, the infraorbital first feeding mode in G. typus, suction of bones are small and tubulous, the eyes are food, was observed with Daphnia.Asan reduced, and there is a wide gap between the experiment, a large piece of frozen salmon fourth infraorbital and the suprapreopercle, was tied to a weight in the aquarium. The both of which are small bones. The neurocra- meat could not be torn easily, but G. typus nial bones of Clarias gariepinus (Fig. 3C), in was able to jerk off small pieces of the larger contrast, form a closed, broad roof, whereas item by grasping and shaking its head. Most the neurocranial width of Clariallabes me- of the time, the body swung itself in a las (Fig. 3B) is situated in between. The U-shaped position, before grasping the food. most conspicuous feature of the head of Rotational feeding was not as pronounced as G. typus is its enormous jaw muscles, which was observed in the eel by Helfman and nearly cover the entire dorsolateral side of Clark (’86) and was combined with the sec- the head. It is primarily the external part of 186 E. CABUY ET AL. the jaw muscle that reduces in the series planted at higher angles on the aponeurosis, from G. typus to C. gariepinus (Fig. 9A–C). producing a much higher moment. As a re- The relative enlargement of these muscles sult, the resultant force vector of the jaw in G. typus was possible because the broad- muscle is rotated counterclockwise and re- ened aponeurosis in the bipinnate jaw muscle sults in a relatively larger vertical vector provided a larger attachment side for addi- component compared to C. gariepinus for tional fibers. The increase of the larger jaw the same contraction (Fig. 9C–A). In addi- muscles has an impact on a major part of the tion, this muscle is extremely enlarged pos- skull. The frontals narrow, for instance, and terodorsally of the head, but also extends the fourth infraorbital and suprapreopercle more toward the front of the lateral side of are more reduced and separated from each the frontal. The anteriormost upper and other to make room for the invasion of the lower fibers of the external adductor insert jaw adductors. These smaller jaw muscle- at the highest angles on the tendinous skel- surrounding bones allow a more extensive eton. Here, the high angle fibers may pro- growth of the external jaw muscles during vide a greater holding force. development, but it cannot be deduced from A second indication for a more powerful adult specimens whether the larger jaw bite is the position of the teeth. The exam- muscles induce a reduced neurocranium, or ined specimens illustrate that the teeth of vice versa. The larger jaw musculature is the lower jaw are positioned more posteri- not directly the driving force behind all the orly, thus closer to the lower jaw articu- cranial differences, because the reduction of lation, when the jaw muscles increase the cranial bones also appears in places (Fig. 6A–C). The position of the prevomeral where they are not susceptible to being af- teeth shows a similar trend (Fig. 4C–A). The fected by the larger jaw muscles. For ex- lateral processes of the prevomer of Gymnal- ample, the caudal side of the skull roof in labes typus are extended more posteriorly, G. typus (Fig. 3A) is much shorter than in by which the teeth are positioned more in C. gariepinus (Fig. 3C) and the ethmoid re- the back, even touching the entopterygoid. gion is less covered with bone. Similarly, a Moreover, the left entopterygoid of one exam- relation between the size of the eyes and the ined specimen is partly covered by teeth, jaw muscles seems to be present. With the whereas usually the entopterygoid of cat- enlargement of the jaw muscles, the eye size fishes does not have teeth (Arratia, ’92). as well as the circumorbital bones decrease Teeth in G. typus and Clariallabes melas are and the eyes shift gradually in a more antero- not as densely packed as in Clarias gariepi- medial position (Fig. 3C–A). nus and are much larger, penetrating deeper In Figure 3, the distance between the ros- into the prey. tral tip of the premaxilla and the caudal side A third indication for a stronger bite is the of the braincase (exoccipital), or the rostral size of the coronoid process of the lower jaw tip of the premaxilla and the fourth par- to which the external jaw muscle is at- apophysis (the latter is defined as the neuro- tached. This process is higher in Gymnal- cranial length in this text), is scaled to the labes typus and Clariallabes melas than in same length. In doing so and by comparing Clarias gariepinus (Fig. 6A–C). A higher the three skulls, we see that the positions of coronoid process implicates a longer power the sutural connections between some neuro- arm and produces a relatively greater verti- cranial medial bones (e.g., mesethmoid, cal force vector component of the jaw muscle frontal, parieto-supraoccipital) are approxi- for the same contraction. The elevation of mately the same. Apparently, the enlarge- the coronoid process is not exclusively corre- ment of the jaw muscles does not change the lated with larger jaw muscles, as is clearly positions of the contact zones between these illustrated by the coronoid process of C. me- bones. las, which is similarly developed as in Some features of the head of Gymnallabes G. typus, but the jaw adductors are not. typus indicate a relatively stronger bite po- Furthermore, the coronoid process of G. ty- tential compared with Clarias gariepinus. pus is situated more in the front compared First, G. typus can generate a stronger force with the lower jaw length of C. garienpinus on the lower jaw because of the larger physi- and also causes a greater moment for the ological cross section of the jaw muscles by same contraction force of the jaw muscle. the attachment of additional fibers. More- Differences in some cranial muscles of over, the anterior muscle fibers are im- Gymnallabes typus involved in mouth open- CRANIAL MORPHOLOGY IN CLARIIDAE 187

Fig. 9. Lateral view of the skull and superficial jaw (␣1 Ͼ␣2 Ͼ␣3). io-III, infraorbital III; io-IV, infraorbital adductor and opercular levator muscles of A: Gymnal- IV; m-A2A3Ј, external part of the adductor mandibulae labes typus, B: Clariallabes melas, C: Clarias gariepi- muscle; m-l-op, levator operculi muscle; spop, suprapre- nus. F indicates the approximate resultant force vector opercle. of the external jaw adductor exerted on the lower jaw 188 E. CABUY ET AL. ing have been found as well. These mainly therefore, contraction of this muscle results concern the levator operculi, the dorsal part in a relatively larger retraction of the inter- of the protractor hyoidei, and the sternohyoi- opercle. Additionally, the opercular process deus. They may be principally modified to of the hyomandibula of G. typus is directed enable a larger gape and to open the mouth more ventrally and has a straight facet in- more quickly, but also to compensate the stead of an obliquely oriented process with a greater resistance caused by the larger jaw more rounded facet as in C. gariepinus. The muscles during jaw depression. A first adap- more vertical position of the opercular pro- tation concerns one of the opercular muscles, cess also explains the shorter, more ven- the levator operculi (Fig. 9A). Normally, con- trally directed dilatator and adductor oper- traction of this muscle results in an upward culi muscles of G. typus. rotation of the opercle around its articula- A second adaptation to generate a larger tion with the hyomandibula. This action mouth opening force on the lower jaws is the plays an important role in the opercular attachment of the protractor hyoidei dorsa- mouth opening mechanism (Aerts and Ver- lis, which is situated on the ventral side of raes, ’84). The rotation is coupled with a the head (Fig. 8B). Contraction results in retraction of the interopercle, which in its elevation of the hyoid bars during closing of turn pulls at the retroarticular process of the mouth, but also in a depression of the the lower jaw, which results in a depression lower jaw when the hypaxial and sternohyoi- of the lower jaw. Since the suprapreopercle deus muscles are contracted simultaneously. of G. typus is much smaller than that of The anterior attachment site of the protrac- Clarias gariepinus, the attachment area is tor muscle in the series from Gymnallabes also smaller. This shifts the levator operculi typus to Clarias gariepinus shifts toward the to the posttemporo-supracleithrum, which midline of the skull (Fig. 10A–C). The contra- is located more posteriorly. The muscle fi- lateral dorsal parts of the protractor hyoidei bers are, therefore, more inclined than in C. of G. typus are separately attached to the gariepinus (Fig. 9C). As a result of this pos- caudal side of the lower jaw. In C. gariepi- terodorsal displacement of the opercular le- nus, however, both parts of this muscle are vator attachment site, its main resultant connected with each other in the sagittal force vector is rotated so that it forms an plane and on the caudal side of the lower acute angle with the dorsal rim of the opercle. jaws. Therefore, the resultant force vector of The muscle fibers are also longer than those the protractor hyoidei dorsalis of G. typus is of the other two species, due to the longer rotated more toward the sagittal plane than distance to be bridged. Both features, more that of C. gariepinus. The morphology of the inclined and larger fibers, may imply a gain protractor of Clariallabes melas is interme- in mechanical advantage of the opercular diate in the one dissected specimen. Appar- four bar system. The levator operculi of ently, there is no correlation between the G. typus is almost oriented to the line of mediolateral shift of the anterior protractor action between the lower jaw and the opercle; muscle fibers and the length of the lower

Fig. 10. Schematic ventral view of the dorsal part of lengths of the lower jaws are scaled to the same length. the hyoid protractor muscle of A: Gymnallabes typus, B: ch-a, anterior ceratohyal; hh-v, ventral hypohyal; mnd, Clariallabes melas, C: Clarias gariepinus. F indicates mandible; m-pr-h-d, protractor hyoidei pars dorsalis the approximate resultant force vector of the dorsal muscle; puh, parurohyal. hyoid protractor muscle exerted on the lower jaw. The CRANIAL MORPHOLOGY IN CLARIIDAE 189 jaws, because the jaws of C. melas are shorter results in a larger mouth opening for the than those of C. gariepinus. A correlation, same rotation angle of the lower jaw. The however, may exist between the hyoid pro- position of the lower jaw articulation socket tractor morphology and the size of the jaw of G. typus is situated lower compared with muscles. C. gariepinus. Generating a larger force on Third, the sternohyoideus of Gymnallabes the lower jaw requires also a better-fixed typus (Fig. 8D) seems to be longer as well, articulation with the quadrate. An adapta- causing a larger depression (or retraction) of tion, therefore, is the lower jaw socket of the hyoids. And, finally, the enlargement of G. typus and C. melas, which has a deeper the posterior fibers of the adductor mandibu- curve than in C. gariepinus (Fig. 6A-C). lae reduces the relative stretching of the The described arrangements in Gymnal- individual posterior fibers as the lower jaw labes typus, which result in a relatively is depressed, thus compensating for a larger greater adductive force and a larger gape, gape. may reflect, therefore, dietary specialization Several skeletal differences between the in anguilliform clariids in general. However, three skulls may as well be related to the stomach analyses of G.typus (this study), larger size of the jaw musculature. A first other extreme eel-like clariids (Matthes, ’64), adaptation concerns a stronger interdigita- and Clarias gariepinus (Groenewald, ’64; tion of the suspensorium with the neurocra- Clay, ’79) have revealed no particular di- nium. In the series from Gymnallabes typus etary difference. According to these data, to Clarias gariepinus, the interdigitation be- although restricted for the anguilliform spe- comes gradually less pronounced. The hyo- cies, all of these catfishes seem to be opportu- mandibula of C. gariepinus has only one nistic feeders. This may provide the answer upward process, articulating with the sphen- to the question why skull remodeling has taken place in smaller, but elongated clari- otic and followed by a cartilaginous articula- ids. The relative increase of the gape and tion ridge (Fig. 5C). The hyomandibula of adductive forces may have compensated for G. typus and other anguilliform clariids have the reduction of the skull and enabled the several processes. Both the sphenotic and clariids to retain their feeding preferences. A pterotic of G. typus have a number of lateral larger number of stomach analyses on an- descending processes, which articulate with guilliform clariids, however, is required to the hyomandibula (Fig. 5A,B). These addi- make this a solid argument. tional processes may be an adaptation for the larger jaw adductor muscles. Indeed, the Anguilliform body and skull size vs. larger posterior hyomandibula processes as well as jaw muscles the jaw musculature of Clariallabes melas The bodies of the three clariid species be- are less developed than in G. typus. The long come more elongated in the series from Clar- posterior hyomandibula processes are ias gariepinus to Gymnallabes typus and the obliquely oriented in the direction of the number of the caudal vertebrae increases lower jaw articulation point. These addi- (Fig. 2C–A). It has been statistically demon- tional processes may prevent a forward or strated in Fundulus that the caudal region backward movement of the hyomandibula is the site of modifications in the number of during depression or elevation of the lower vertebrae (Ford, ’33 in Gabriel, ’44), appar- jaws, as well as dislocation through torque ently in the region where the somites last forces. Also, the hyomandibula shifts gradu- form. Elongation means either a reduction ally posteriorly in the series from C. gariepi- in the relative diameter for a given mass, an nus to G. typus (Fig. 4C–A). Moreover, its increase in the relative length, or both, the rostral membranous plate shortens and this latter probably being applicable for the three results in longer jaw muscle fibers because species. As the bodies of the three clariid their insertion area is partly situated on this species become more anguilliform, the dor- plate. The hyomandibula of G. typus is also sal and anal fins gradually enlarge and are higher, articulating closer to the midline of ultimately confluent with the caudal fin. In the skull. A higher suspensorium can enable the latter, more hypurals are fused with a larger lateral displacement of its ventral each other when the body elongates, ending margin and thus an orobranchial volume up in a single plate as in Channallabes apus. increase, than a shorter one for the same The paired fins reduce relative to the total rotation angle (Adriaens and Verraes, ’94) body length (Fig. 2C–A), as was noted by (Fig. 11A–B). In addition, a longer mandible earlier authors (Boulenger, ’07; Pellegrin, 190 E. CABUY ET AL.

Fig. 11. Schematic view of the position of the suspen- typus is relatively larger than hq2 and L2 in C. gariepi- sorium in relation to the skull of A: Gymnallabes typus nus, respectively. This results in a relatively larger and B: Clarias gariepinus. The distance hq1 and L1 in G. mouth opening (hj1 Ͼ hj2).

’27; David, ’35; Poll, ’77), but this does not markable example to illustrate this is Tan- occur in relation to the cranial length ganikallabes mortiauxi, endemic to Lake (Fig. 9A–C). Tanganyika (Poll, ’43). The neurocranium of Anguilliform clariids are equipped with T. mortiauxi is extremely narrow (Poll, ’43: much larger jaw muscles compared with Fig. 1), but its external morphology looks their fusiform relatives, and intermediate like that of C. melas. Based on the approxi- forms such as Clariallabes melas can be mate number of dorsal fin rays, i.e., 70 (Poll, easily found among the Clariidae. Both char- ’43), the number of vertebrae should be much acteristics, anguilliform body versus hyper- lower than in an anguilliform clariid. Other trophied jaw muscles, and thus feeding mode, examples, such as Clarias salae or the may in some way be correlated with each smaller Clarias submarginatus, have a num- other. This study and literature on clariids, ber of vertebrae similar to C. melas, but however, seem to reveal that the relation their lateral skull roof bones are not reduced between the jaw muscle size and the degree (Teugels, ’86b). Channallabes apus also of anguilliformism (or the number of ver- shows that the relation between the degree tebrae) in clariids is not consistent. A re- of anguilliformism and the size of the jaw CRANIAL MORPHOLOGY IN CLARIIDAE 191 muscles is inconsistent or that the size of the ized by a gap between the fourth infraorbital jaw muscles does not gradually increase and suprapreopercle (the supraorbital and when vertebrae are added. Also, two exam- dermosphenotic of Greenwood) (Greenwood, ined clariids (124 and 239 mm TL) originally ’61). determined as Gymnallabes typus, have a The larger jaw muscles of Gymnallabes broader neurocranium than we see in typus and other anguilliform clariids cannot G. typus, almost identical with that of be related to a decrease in absolute skull C. melas, but the number of vertebrae (81) is size alone. Noneel-like clariid species with greater than in G. typus. It is, therefore, not an adult, closed skull roof of Ͻ2 cm corrobo- understood whether a relation exists be- rate this statement. Clarias species, such as tween the increasing jaw muscles and elon- Clarias alluaudi or Clarias albopunctatus, gation in clariids. It can be said that such a e.g., also have a closed skull roof as does relation is not present in other siluroids on Clarias gariepinus (Teugels, ’86b), but are the basis of a superficial survey conducted much smaller in adult life. Therefore, this by Howes (’83), which indicated a mean num- trait cannot be viewed as related to small ber of 40–45 vertebrae among siluroids, size alone. In addition, the degree to which a whereas only some members of the Clari- muscle fiber may be stretched depends on its idae have exceptionally high numbers. How- absolute length, i.e., on the number of sarco- ever, it is certain that eel-like clariids al- meres of which it is composed. Thus the ways have hypertrophied jaw muscles, length of the jaw muscle fibers ina2cmor whereas almost all nonanguilliform clariids slightly smaller anguilliform skull is appar- have much smaller ones. Strikingly, in the ently not limited in restricting the excursion Mastacembelidae or spiny eels the size of range of the jaw adductors, as these fibers the jaw muscles decreases when the body may be even smaller in a similar sized closed elongates, as opposed to what we see in skull roof like that of C. alluaudi. clariids. Microphthalmic and cryptophthal- We could also think that relatively more mic mastacembeloids have expanded their space leaves scope for the development of jaw muscles but have the lowest number of additional structures in larger skulls. The vertebrae (Travers, ’84). meristic characters such as the number of Not only do we see a relative enlargement branchial filaments, the number of gill rak- of the external jaw adductors when the body ers, or the number of olfactory lamellae of becomes more elongate, but the maximum Clariallabes melas are slightly higher than attainable body size decreases as well. The those of Gymnallabes typus, but increase known maximum length for Clarias gariepi- sharply in Clarias gariepinus, as does the nus is 1.5 m (Teugels, ’86a), for Clariallabes skull size of these species. Although the se- melas 260 mm (Poll, ’41), and for Gymnal- ries confirms the statement, meristic charac- labes typus 238 mm. Thus the absolute size ters of the head of Clarias species (Teugels, of the cranium decreases in the series from ’86b) corroborate that there is no relation C. gariepinus to G. typus (Fig. 2C–A). On with skull size. As the absolute size of the first thought, the maximum attainable skull skull decreases, the relative size of the brain- size may be correlated with changes in the case and the semicircular canals (inner ear), arrangement of the jaw adductor muscula- deduced from the size of the prootics, do not ture and skull roof bones. Indeed, readings change. An exception is C. melas, of which in the literature on clariids (e.g., David, ’35; the parasphenoid, forming the anterior Poll, ’41, ’42a,b, ’43, ’57, ’77; Greenwood, ’61; braincase, and the prootics are slightly Teugels, ’86b; Teugels et al., ’90) reveal that broader (Fig. 4B). In addition, the relatively larger jaw muscles only occur in small adult smaller skull of G. typus is less strongly skulls. Teugels (’86b) has noted in his descrip- connected with the postcranial skeleton, tion of each Clarias species whether the making it easier for the head to maneuver. fourth infraorbital and suprapreopercle are The ventromedial ridge of the parieto-supra- separated and reduced. Apparently, mainly occipital of G. typus, on which the epaxial the small fishes (maximum 300 mm) of the muscles are attached, is much shorter than subgenus Brevicephaloides consist of species in C. gariepinus. Moreover, the transscapu- that have separated lateral skull roof bones, lar process abuts only the rostral side of the as a result of larger jaw muscles. An excep- fourth parapophysis (Fig. 3A), whereas in tion to this rule is Dinotopterus. Some adult the other two species, this process reaches it fishes (Ͼ 1 m) of this genus are character- in the middle (Fig. 3B,C). 192 E. CABUY ET AL. The reduction of neurocranial bones in increased elongation in some clariids. Their clariids apparently influences the location of assumption that a separate lineage as such the head sensory canals during develop- would exist is unrealistic. First, selecting a ment. In most Siluriformes, the infraorbital handful of species from a large and complex canal branches off from the supraorbital ca- group of a variety of species is not sufficient nal in the sphenotic bone (Arratia and Hua- to prove such a trend. Second, by construct- quı´n, ’95). Gymnallabes typus and Clarial- ing the orthogenetic series, the importance labes melas have this bifurcation in the fron- of external fish morphology is overempha- tal bone. This is also the case in the two sized. This study has demonstrated, how- examined Channallabes apus, although not ever, the presence of a nongradual trend in in the skull drawn by Poll (’77). The shift of the examined skulls. The cranium of Clarial- this sensory canal bifurcation is related to labes melas is not an entirely intermediate the width of the neurocranium and, there- form between Gymnallabes typus and Clar- fore, to the size of the jaw musculature. The ias gariepinus, some of its skull bones being anterior displacement of the point of branch- more massive. If the presumed trend were ing into the frontal also occurs in a few other consistent, it would have been legitimate to catfishes (see Lundberg, ’82). Yet, the infra- expect an even further reduced skull in the and supraorbital canal fusion does not most elongated Channallabes apus (108 ver- always occur in the frontal when the neuro- tebrae), but the latter’s neurocranium is cranium is narrow. In Diplomystes camposen- slightly broader than that of G. typus.In sis, e.g., the infraorbital still bifurcates addition, the suprabranchial organs of in the sphenotic (Arratia and Gayet, ’95: C. apus are larger and the number of abdomi- Fig. 5). In addition, the preopercular canal of nal vertebrae (23) and ribs (18) is higher G. typus is situated more in the back of the than in G. typus. skull as a consequence of the reduced post- It is suggested here that the orthogenetic temporo-supracleithrum and suprapreoper- series of Boulenger may be rather a reflec- cular bones. The reduction of the neurocra- tion of highly diverse species. Based on the nial bones may also cause the absence of side branches or a decrease in the number of aberrant forms in asymmetry, meristic surface pores of the sensory canals. The ant- counts, and absence of an organ, which were orbitals of C. apus and G. typus do not bear a encountered during this study, the least that sensory canal as in C. melas and Clarias can be said is that the examined clariids can gariepinus. Also, the number of pores of the be regarded as variable specimens. For ex- cranial lateral line system of G. typus is ample, a left-right asymmetry in the num- reduced. For instance, the preoperculo-man- ber of pterygoids is seen in the 217 mm TL dibular canal consists of seven pores, Gymnallabes typus, which also lacks the first whereas C. gariepinus has two pores extra. ceratobranchial. Also, the parieto-supraoc- The reduced neurocranium of anguilli- cipital of the 214 mm TL G. typus is not form clariids has been interpreted as a juve- fused in the middle, as is normally the case. nile trait of their predecessors (Poll, ’42a, Several aberrant characters in the other spe- ’77; Greenwood, ’56). The expression of juve- cies were also revealed. Channallabes apus nile traits in the adult would have been has two separate ventral median fins in- caused by a retarded somatic development stead of one anal fin, Clariallabes melas is (neoteny), which would have played an im- missing a right pelvic fin, and Clarias garie- portant role in the development of morpho- pinus a right suprapreopercle. Fused verte- logical adaptations in anguilliform clariids. brae were present in most of the specimens. It is currently impossible, however, to dis- In order to define some of these characters cuss the implications of paedomorphosis on as an abnormal condition, however, more clariid ontogeny, mainly because of the lack extensive research is required. of information on the embryonic develop- The monophyletic nature of the Clariidae ment of the anguilliform clariid skull. has not yet been demonstrated (Teugels, ’96), and the taxonomy of most genera, the anguil- Interrelationships of anguilliform clariids? liform clariids in particular, needs further The series of Boulenger (’07) and Pellegrin exploration. A phylogenetic analysis of the (’27), illustrated here by four species, has Clariidae, which is lacking in the literature, been interpreted by several other authors could verify the old hypothesis of Boulenger (a.o., Poll, ’41, ’42a, ’77; Greenwood, ’56; Lam- that predecessors of primitive fishes, such as bert, ’60) as an evolutionary trend toward an Heterobranchus and most Clarias, would CRANIAL MORPHOLOGY IN CLARIIDAE 193 have evolved into progressively anguilliform luroidei: Clariidae), and the consequences for the respi- clariids. This is the subject of ongoing re- ratory movements. Neth J Zool 47:61–89. Adriaens D, Verraes W. 1997b. Ontogeny of the chondro- search. Also, the existence of a correlation cranium in Clarias gariepinus (Siluriformes: Clari- between the size of the jaw muscles versus idae): trends in siluroids. J Fish Biol 50: 1221–1257. body shape in clariids could be ascertained Adriaens D, Verraes W. 1997c. Ontogeny of the hyoid once a phylogenetic tree is established. musculature in the African catfish, Clarias gariepinus (Burchell, 1822) (Siluroidei: Clariidae). Zool J Linn Soc 121:105–128. CONCLUSIONS Adriaens D, Verraes W, Taverne L. 1997. The cranial Although only a few specimens have been lateral-line system in Clarias gariepinus (Burchell, 1822) (Siluroidei: Clariidae): morphology and develop- investigated, several conclusions can be ment of canal related bones. Eur J Morphol 35:181– drawn. First, the adductor mandibulae of 208. some anguilliform clariids (e.g., Gymnal- Adriaens D, Verraes W. 1998. Ontogeny of the osteocra- labes typus) is substantially more developed nium in the African catfish, Clarias gariepinus Burchell (1822) (Siluriformes: Clariidae): ossification and the neurocranium is much narrower sequence as a response to functional demands. J Mor- than in the nonanguilliform Clarias gariepi- phol 235:183–237. nus. Second, several morphological adapta- Aerts P, Verraes W. 1984. Theoretical analysis of a tions, both skeletal and muscular, can be planar four bar system in the teleostean skull: the use of mathematics in biomechanics. Ann Soc r zool Belg regarded as being related to a more powerful 114: 273–290. biting potential and a wider gape. A stronger Arratia G. 1992. Development and variation of the sus- bite is performed by a larger physiological pensorium of primitive catfishes (Teleostei: Ostari- cross section of the jaw muscles, a stronger ophysi) and their phylogenetic relationships. Bonn zool Monogr 32:1–148. interdigitation between the hyomandibula Arratia G, Gayet M. 1995. Sensory canals and related and the neurocranium, the prevomeral and bones of Tertiary siluriform crania from Bolivia and mandibular teeth being positioned closer to North America and comparison with recent forms. J the lower jaw articulation, and a higher coro- Vert Paleont 15:482–505. Arratia G, Huaquı´n L. 1995. Morphology of the lateral noid process. The enlargement of the adduc- line system and of the skin of diplomystid and certain tor mandibulae, levator operculi, and the primitive loricarioid catfishes and systematic and eco- sternohyoideus also enables increase of the logical considerations. Bonn zool Monogr 36:5–109. gape. It is not understood, however, whether Boulenger GA. 1887. Ann Mag Nat Hist Lond 19:148. Boulenger GA. 1907. A revision of the African silurid a relation exists between the size of the jaw fishes of the subfamily Clariinae. Proc Zool Soc Lond muscles and the shape of the body. Third, 2:1062–1096. although some morphological features of the Boulenger GA. 1911. Freshwater fishes of Africa. Br intermediate anguilliform Clariallabes me- Mus Nat Hist 2: 217. Burchell W. 1822. Travels in the interior of Southern las fit well into the trend, this does not apply Africa, Vol. I. London: Batchworth Press. to all morphological characters. For in- Clay D. 1979. Population biology, growth and feeding of stance, the Weberian apparatus and the African catfish (Clarias gariepinus) with special refer- braincase are relatively broader, and the ence to juveniles and their importance in fish culture. Arch Hydrobiol 87:453–482. lower jaw is shorter. Moreover, the pterygoid David L. 1935. Die entwicklung der Clariiden und ihre bones of C. melas are different. verbreitung. Rev Zool Bot Afr Bruxelles 28:77–147. Fink SV, Fink WL. 1996. Interrelationships of ostari- ACKNOWLEDGMENTS ophysan fishes (Teleostei). In: Stiassny MLJ, Parenti LR, Johnson GD, editors. Interrelationships of fishes. We thank Prof. Dr. F. Ollevier and Dr. F. New York: Academic Press: p 209–249. Volckaert of the Laboratory of Ecology and Gabriel ML. 1944. Factors affecting the number and Aquaculture (Catholic University of Leu- form of vertebrae in Fundulus heteroclitus. J Exp Zool 95:105–147. ven) and Mr. W. Fleure (Someren, The Neth- Gosse JP. 1963. Sciences. Zoology 116:113–166 erlands) for providing the specimens of Clar- Greenwood PH. 1956. A new species of Clariallabes ias gariepinus. (Pisces, Clariidae), from the Nile. Proc Zool Soc Lond 127:555–564. LITERATURE CITED Greenwood PH. 1961. A revision of the genus Dinotop- terus Blgr (Pisces, Clariidae) with notes on the com- Adriaens D, Verraes W. 1994. On the functional signifi- parative anatomy of the suprabranchial organs in the cance of the loss of the interhyal during ontogeny in Clariidae. Bull Br Mus Nat Hist 7: 215–241. Clarias gariepinus Burchell, 1822 (Teleostei: Siluroi- Gregory WK. 1933. Fish skulls: a study on the evolution dei). Belg J Zool 141:139–155. of natural mechanics. Trans Am Phil Soc 23:75–481. Adriaens D, Verraes W. 1996. Ontogeny of cranial mus- Groenewald AAVJ. 1964. Observations on the food hab- culature in Clarias gariepinus (Siluroidei: Clariidae): its of Clarias gariepinus (Burchell), the African fresh- the adductor mandibulae complex. J Morphol 229: water barbel (Pisces: Clariidae) in the Transvaal. Hy- 255–269. drobiology 23:287–291. Adriaens D, Verraes W. 1997a. Ontogeny of the suspen- Gu¨ nther A. 1867. Gymnallabes, gn (Silur). Ann Mag Nat sorial and opercle muscles in Clarias gariepinus (Si- Hist Lond 20:111. 194 E. CABUY ET AL.

Gu¨ nther A. 1873. Gymnallabes apus. Ann Mag Nat Hist Clarias gariepinus (Pisces, Siluriformes). Can J Zool Lond 12:142–144. 67:2090–2097. Helfman GS, Clark JB. 1986. Rotational feeding: over- Roberts TR. 1975. Geographical distribution of African coming gape-limited foraging in anguillid eels. Copeia freshwater fishes. Zool J Linn Soc 27:249–319. 3: 679–685. Roberts TR, Stewart DJ. 1976. An ecological and system- Howes GJ. 1983. Problems in catfish anatomy and phy- atic survey of fishes in the rapids of the lower Zaire or logeny exemplified by the Neotropical Hypophthalmi- Congo river. Bull Mus Comp Zool 147:239–317. dae (Teleostei: Siluroidei). Bull Br Mus Nat Hist (Zool.) Roman B. 1970. Nuevas especies de peces de Rio Muni 45: 1–39. (Guinea Ecuatorial). P Inst. Biol Apl 49:5–23. Lambert JG. 1960. Famille des Clariidae. Bull Soc Roy Skelton PH, Teugels GG. 1992. Neotype description for Zool Am 19:24. the African catfish Clarias gariepinus (Burchell, 1822) Lundberg JG. 1982. The comparative anatomy of the (Pisces: Siluroidei: Clariidae). Ichthyological Bull JLB toothless blindcat, Trogloglanis pattersoni Eigen- Smith Institute for Ichthyology, Grahamstown, South mann, with a phylogenetic analysis of the ictalurid Africa 56:1–8. catfishes. Misc Publ Mus Zool, Univ MI 163:1–85. Travers RA. 1984. A review of the Mastacembeloidei, a Matthes H. 1964. Les poissons du Lac Tumba et de la suborder of synbranchiform teleost fishes. Part I: Ana- region d’Ikela. Ann Mus Afr Centr Zool 126:1–204. tomical descriptions. Bull Br Mus Nat His 46:1–133. Moussa TA. 1956. Morphology of the accessory air- Taylor WR, Van Dyke GC. 1985. Revised procedures for breathing organs of the teleost, Clarias lazera (C and staining and clearing small fishes and other verte- V). J Morphol 98:125–160. brates for bone and cartilage study. Cybium 9:107– Munshi JS. 1961. The accessory respiratory organ of 119. Clarias batrachus (Linn). J Morphol 109:115–140. Teugels GG. 1982. A systematic outline of the African Nelson JS. 1994. Fishes of the world, 3rd ed. New York: John Wiley & Sons. 600 p. species of the genus Clarias (Pisces; Clariidae), with Pellegrin J. 1927. La disparition des nageoires paires an annotated bibliography.Ann Mus Roy Afr Centr Sci chez les poissons africains du groupe des clariine´s. Zool 236:1–249. Ann Sc Nat (Zool) Paris 10:209–222. Teugels GG. 1983. La structure de la nageoire adipeuse Poll M. 1941. Etude syste´matique et morphologique dans les genres Dinotopterus, Heterobranchus et Clar- d’une collection de poissons de l’Uele (Congo Belge), ias (Pisces; Siluriformes; Clariidae). Cybium 7:11–14. comprenant trois espe`ces nouvelles. Bull Mus Roy Teugels GG. 1986a. Clariidae. In: Daget J, Gosse JP, Van Hist Nat Belg 17:1–18. Den Audenaerde DFET, editors. Checklist of the fresh- Poll M. 1942a. Note sur l’oste´ologie de Dolichallabes water fishes of Africa. Cloffa. 2: 66–101. microphthalmus Poll et remarques sur l’e´volution des Teugels GG. 1986b. A systematic revision of the African Clariidae. Ann Soc Roy Zool Belge 72:222–235. species of the genus Clarias (Pisces; Clariidae). Ann Poll M. 1942b. Description d’un genre nouveau de Clari- Mus Roy Afr Centr, Sci Zool 247:1–199. idae originaire du Congo Belge Rev Zool Bot Afr 36:94– Teugels GG. 1996. Taxonomy, phylogeny and biogeogra- 100. phy of catfishes (Ostariophysi, Siluroidei): an over- Poll M. 1943. Description du Tanganikallabes mor- view. Aquat Living Resour 9: 9–35. tiauxi, gen nov, sp n, de la famille des Clariidae. Rev Teugels GG, Denayer B, Legendre M. 1990. A systematic Zool Bot Afr 37:126–133. revision of the African catfish genus Heterobranchus Poll M. 1957. Les genres des poissons d’eau douce de Geoffroy-Saint-Hilaire, 1809 (Pisces: Clariidae). Zool l’Afrique. Ann Mus Roy Congo belge, Tervuren 8:1– J Linn Soc 98:237–257. 183. Teugels GG, Reid G McG, King RP. 1992. Fishes of the Poll M. 1959. Recherches sur la faune ichthyologique de Cross River Basin (Cameroon - Nigeria), Taxonomy, la re´gion du Stanley Pool. Ann Mus Roy Congo belge, zoogeography, ecology and conservation. Ann Mus Roy Tervuren. 8:75–174. Afr Centr, Sci Zool 266:1–132. Poll M. 1977. Les genres nouveaux Platyallabes et Platy- Vandewalle P, Chardon M. 1991. A new hypothesis on clarias compare´s au genre Gymnallabes GTHR. Synop- the air flow in air breathing in Clarias gariepinus sis nouveau des genres de Clariidae. Bull Classe Sci 5: (Teleostei, Siluriformes). Belg J Zool 121:73–80. 122–149. Winemiller KO, Kelso-Winemiller LC.1996. Compara- Rademaker F, Surlemont C, Sanna P, Chardon M, Vande- tive ecology of catfishes of the Upper Zambezi River walle P. 1989. Ontogeny of the Weberian apparatus of floodplain. J Fish Biol 49:1043–1091.