Mastacembelidae (Teleostei; Synbranchiformes) subfamily division and African generic division: an evaluation

The family Mastacembelidae is currently divided into two subfamilies, the Mastacembelinae and the Afromastacembelinae. Within the African Afromastacembelinae two genera, Caecomastacembelus and Aethiomastacembelus, are recognized. Both the subfamily and the African generic division were critically evaluated mainly through X‐ray analyses and osteological research. There is no supporting evidence for a split into two subfamilies. In addition, the present diagnosis of the African genera proved to be unworkable, and no convincing phylogenetic evidence supporting this classification is available. Therefore, pending new detailed phylogenetic research, the use of a subfamily classification within the family Mastacembelidae is not recommended. Furthermore, Caecomastacembelus and Aethiomastacembelus are placed in synonymy with the genus Mastacembelus.


Introduction
Mastacembelidae are percomorph fishes present in major parts of tropical and subtropical Africa, the Middle East and South-East Asia, north to China. They are closely related to the families Chaudhuriidae and Synbranchidae with which they form the order Synbranchiformes (Travers 1984a(Travers , 1984b. Mastacembelidae are anguilliform fishes. Some species can attain a maximum length of about 1 m. Very characteristic is the rostral appendage which bears the two tubulated anterior nostrils on each side of a central rostral tentacle. The gill opening is reduced due to a connection of the opercular membrane with the lateral wall of the body. Mastacembelidae have a long series of well-separated dorsal spines, hence their name, spiny eels. They also have a short series of anal spines. Pelvic fins and the pelvic girdle are absent. Most species are characterized by a huge number of small cycloid scales. Characteristic for all African members of the family are the confluent dorsal, caudal and anal fins. Travers (1984b) subdivided the Mastacembelidae into two subfamilies: the Mastacembelinae Travers, 1984 from the Oriental region, and the Afromastacembelinae Travers, 1984 endemic to Africa. Four genera of Mastacembelidae are presently recognized (Travers 1984b(Travers , 1988: two of them, Mastacembelus Scopoli, 1777 and Macrognathus Lacepède, 1800, within the Mastacembelinae, the other two, Caecomastacembelus Poll, 1958 andAethiomastacembelus Travers, 1988, within the Afromastacembelinae (see Figure 1). Kottelat (1991), Kottelat and Lim (1994) and Britz (1996) do not agree with some of the taxonomic changes introduced by Travers (1984b), including the present position of Sinobdella Kottelat and Lim, 1994(5Rhynchobdella Bloch and Schneider, 1801in Travers, 1984b within the Chaudhuriidae, sister family of the Mastacembelidae. Instead, they propose placing S. sinensis (Bleeker, 1870) within the Mastacembelidae (Kottelat and Lim 1994;Britz 1996), in which it may form the sister group of all remaining mastacembelids (Britz 1996).
Also, Johnson and Patterson (1993) found some problems with Travers' (1984b) characterization of the Chaudhuriidae, as they found that some characters seem to be primitive, others may be primitive and others are possibly non-existent. Nevertheless, they regarded Chaudhuria Annandale, 1918 and Rhynchobdella (5Sinobdella) as primitive mastacembeloids.
The criticism was further developed by Britz (1996). He concluded that there are no significant similarities between Sinobdella and the Chaudhuriidae, except for the lost entopterygoid, that would support the inclusion of S. sinensis within the family Chaudhuriidae (for more details see Britz 1996). Britz (1996) mentioned two characters, the special articulation of the ectopterygoid with the lateral ethmoid via a cartilaginous meniscus, and the peculiar elongated shape and dorsally shifted position of the coronomeckelian bone, which are shared between Sinobdella and the mastacembelids but not found in the chaudhurids, synbranchids or any other teleost.

Material and methods
For a complete list of the examined specimens see below. Institutional abbreviations follow Leviton et al. (1985); c/s indicates cleared and stained.
X-rays were made with a Balteau 5-50 kV equipped with a Tubix cell, Type 6LA, Paris (France). Some cleared and stained specimens available at the AMNH, BMNH and Figure 1. Phylogeny of the Mastacembelidae as proposed by Travers (1984b). Numbers are from Travers (1984b) and refer to principal synapomorphies. 58, Mastacembelidae; 59, Mastacembelinae; 61-68, Macrognathus (E, F-J); 90-93, Afromastacembelinae; 94-95, Caecomastacembelus (L, M-S); D, Mastacembelus; K, Afromastacembelus. According to Travers (1984b), within the Mastacembelinae which otherwise have four hypural plates, one species, M. aculeatus, has five hypural elements. This was interpreted by Travers (1984b) as a retention of the plesiomorphic condition. Contrary to Travers (1984b), I never observed five separate hypurals as the hypural elements are always more or less fused to each other (see Table I). Macrognathus pancalus is another exception mentioned by Travers (1984b) as it has only a single upper and a single lower hypural plate which was interpreted by Travers (1984b) as a highly derived condition convergent with that found in some African species.
As a result, there seems to be parallel occurrence within the suborder Mastacembeloidei of a reduction in the number of hypural plates by hypural fusion based on the following evidence: (1) there are only two, or only one [I observed one parhypural and only one large hypural in two cleared and stained specimens of S. sinensis (AMNH 11077 and AMNH 11078); these differences might illustrate some degree of intraspecific variation], hypural plate(s) in Chaudhuria, Pillaia and S. sinensis; (2) there are from five (M. aculeatus) to a single (M. pancalus) (the actual numbers are different: see above) hypural plate(s) in the species of the genus Macrognathus (Mastacembelinae); (3) there is an important intraspecific variation in the number of hypural plates; and (4) there are only two hypural plates (the actual number is different: see below) in the African Mastacembelidae species (Afromastacembelinae). Therefore, a reduced number of hypural plates due to, at least in part, hypural fusion seems to be an independent acquisition in several Mastacembeloidei lineages, i.e. Chaudhuriidae (Chaudhuria caudata Annandale, 1918 andPillaia indica Yazdani, 1972), Sinobdella sinensis, Macrognathus (M. pancalus) and the Afromastacembelinae.
In conclusion, due to the high intra-and interspecific variability and the observed tendencies to hypural fusion in independent lineages, the usefulness of the hypural plate number character for phylogenetic reconstruction and classification within the Synbranchiformes is, at present, unfounded. This character, established by Travers (1984b) to define the Mastacembelinae as a monophyletic assemblage, most probably defines a paraphyletic assemblage.
There is another point that is crucial to the discussion of hypural plates, i.e. that without ontogenetic data, it is hard to determine how many and which hypurals fuse into hypural plates. For instance, it may be that two plates in one taxon represent fusion of 1+2 and 3+4 and that in another the two plates consist of hypural 1 and 2+3. Therefore, homology of elements fusing during ontogeny can only be evaluated through ontogeny (Britz, personal communication). Travers (1984a), in his illustration of the caudal fin skeleton of M. pancalus, labelled the lower hypural plate as H1+2 and the upper plate as H3+4+5+6. A preliminary examination of an ontogenetic series of M. pancalus was undertaken. Early in ontogeny there are only four autogenous cartilaginous precursors for the hypurals. Then the two upper and the two lower hypurals, respectively, fuse together during ontogeny. Also, the lowest hypural fuses TL, total length; n, number examined. The caudal skeletal formulae and caudal soft fin ray number on one single row are not necessarily observed on the same specimens. A plus (+) indicates that the elements are partially fused (most often towards their posterior end). Parentheses ( ) refer to structures that are more tightly fused to each other than to other elements of the caudal skeleton. Numbers between square brackets [ ] indicate that a single structure is composed of several elements that are not completely fused. An asterisk (*) indicates that the elements are partially fused posteriorly only. Caudal skeleton formula: AHS (only reported when partially fused to other caudal elements), autogenous haemal spine; Ph, parhypural; H, hypural; ADS, autogenous dorsal spine. partially with the parhypural. Homology of these four remaining hypural cartilaginous precursors cannot be determined.

Afromastacembelinae
First, the problems concerning the synapomorphies proposed by Travers (1984b) to define the monophyly of the African subfamily are discussed, and secondly the problems with the African generic division will be dealt with. Travers (1984b) considered the subfamily Afromastacembelinae a monophyletic group based on four synapomorphies: (1) the lack of an ascending process on the urohyal, or direct articulation between this bone and basibranchial 1; (2) (a) hypural plates, generally two, (b) tendency for parhypural fusion to ventral edge of lower plate, (c) 8-10 principal fin rays and (d) confluent caudal fin; (3) the scapula foramen not completely bone-enclosed; and (4) a tendency to have noticeably more caudal than abdominal vertebrae (see Figure 1, Nos 90-93).
As formulated by Travers (1984b) a 'tendency' means that not all species included within such an assemblage display this character state. This raises the problem of delimiting monophyletic assemblages. Furthermore, a 'tendency' is not a character.
Criticism can be formulated concerning each of the four synapomorphies presented by Travers (1984b) to support the monophyly of the Afromastacembelinae.
(1) The lack of an ascending process on the urohyal or a direct articulation between this bone and basibranchial 1 is not unique to the Afromastacembelinae; it is also observed in the Asian genus Chaudhuria (Travers 1984b).
Furthermore, Travers (1984a) stated that in a number of Asian mastacembeloids (i.e. some Macrognathus species) also, the urohyal lacks an ascending process although these taxa are distinguished by a direct articulation between the anterodorsal surface of the urohyal and the keel on basibranchial 1.
In addition, the urohyal in Mastacembelus oatesii Boulenger, 1893 [see Figure 2 versus Figure 3 (Mastacembelinae: M. mastacembelus)] also lacks an ascending process [see Figure 4 (Afromastacembelinae: M. nigromarginatus)]. Therefore, following Travers' hypothesis of the phylogeny of the Mastacembelidae, the loss of the ascending process on the urohyal would have occurred independently at least twice: i.e. once within the genus Mastacembelus and once within the Afromastacembelinae. As a result, this character does not provide strong evidence to support the monophyly of the Afromastacembelinae. (2a) Hypural plates, generally two, but three in M. moorii and even four in a single specimen of M. vanderwaali (Travers 1984a). There are more exceptions to the general presence of two hypural plates in Travers' (1984aTravers' ( , 1984b publications. In his illustration of the caudal region of M. vanderwaali (Travers 1984a) one can indeed observe two separate hypurals (H2 and H3+4+5+6). However, he labels a third one as fused with the parhypural (Ph+H1). Therefore, there are three instead of two hypural plates. From Travers' (1984b) comments it can be concluded that in Mastacembelus moorii, M. ophidium and M. paucispinis, and in an undescribed species (subsequently described as A. sexdecimspinus Roberts and Travers, 1986) there should be a total of three hypurals (two upper elements and one lower element). As no supporting illustration was given this could not be verified. However, in the description of A. sexdecimspinus Roberts and Travers (1986) reported that the caudal skeleton is composed of two hypural plates and the parhypural is fused to the hypural one. From the drawings presented by Travers (1984a) it is obvious that two hypurals (or three as the two upper elements are imperfectly fused) were found in M. congicus Boulenger, 1899 and a single hypural was found in M. ellipsifer Boulenger, 1899 and M. aviceps Roberts and Stewart, 1976. A high amount of interspecific, and also intraspecific, variation of the hypural plate number is also described for several Oriental species. As discussed above, this seems to suggest that there are strong tendencies within various groups of the Mastacembeloidei  towards fusion of the hypural plates. As such, the four separate hypural plates character provided by Travers (1984b) does not provide strong evidence for his dichotomy between Mastacembelinae and Afromastacembelinae. As a result this character, as defined by Travers (1984b), does not provide strong evidence to support the monophyly of the Afromastacembelinae.
(2b) Tendency for parhypural fusion to the ventral edge of the lower hypural plate (Travers 1984a). Indeed, fusion can be observed in several species illustrated by Travers (1984a): M. frenatus, M. shiranus and M. vanderwaali (largely fused) and M. aviceps (partially fused at base). On the contrary, there is no fusion in M. congicus and M. ellipsifer. Since Travers (1984b: 136) considers the fusion as a 'tendency' one could expect the actual fusion to be the most derived condition. This is not the case as M. ellipsifer, which shows the derived character state of having only one hypural, has a completely independent parhypural. As a result also this character is weak evidence to support the monophyly of the Afromastacembelinae.
Because of these additional data this 'tendency' proposed by Travers (1984b) as additional evidence for the monophyly of the Afromastacembelinae has become much too weak.
(2d) All Afromastacembelinae have a completely confluent dorsal, caudal and anal fin (Travers 1984b). Nevertheless, as reported by Travers (1984b) in several Asian Mastacembelinae species (e.g. Mastacembelus armatus, M. caudiocellatus Boulenger, 1893, M. circumcinctus Hora, 1924, M. erythrotaenia and Macrognathus maculatus), even if a distinct caudal fin is present, the fin is confluent basally with the dorsal and anal. According to Travers (1984a), in these Asian species a distinct caudal is still discernible as the caudal rays are longer than, and extend beyond the tips of, the last dorsal and anal fin rays.
However, in  identification key to the Mastacembelidae of Burma and Thailand Mastacembelus armatus, M. erythrotaenia and M. favus are diagnosed by dorsal and anal fins which are broadly joined to the caudal fin, and a caudal fin in which the outline is merged with that of dorsal and anal fins. In at least one M. erythrotaenia specimen examined (ROM 53683: 394 mm TL), and also in Sufi's (1956) illustrations of the species, a distinct caudal is not discernible and the caudal rays are not longer than the last dorsal and anal fin rays, resulting in a caudal fin very similar to that found in African species.
In some of the small M. paucispinis (e.g. MCZ 50562) and M. congicus (e.g. MRAC 63215-226 and MRAC 79-001-P-6343-6360) specimens examined the caudal fin rays are also somewhat longer than, and extend beyond the tips of, the last dorsal and anal fin rays, and so a distinct caudal remains somewhat discernible.
These data and observations bridge the supposed gap between the Mastacembelinae species and the Afromastacembelinae even further and again illustrate that this character employed by Travers (1984b) does not provide strong evidence for a dichotomy between Mastacembelinae and Afromastacembelinae.
According to Travers (1984a), among Asian mastacembeloids the caudal fin in Sinobdella, Chaudhuria and Pillaia is exceptional since it is confluent with the dorsal and anal fins, and has only eight or nine rays. This arrangement is similar to that in the African taxa, all of which have a confluent caudal composed, in the majority of species, of 8-10 principal rays (Travers 1984a). The confluence of dorsal, caudal and anal fins seems to reflect homoplasy: i.e. the independent acquisition of a comparable character state in several lineages as a result of convergent evolution, within the Mastacembeloidei. As such, the confluence of the unpaired fins, as defined by Travers (1984b), is not necessarily synapomorphic for the Afromastacembelinae.
According to Travers (1984a) the caudal fin in Chaudhuria caudata is 'distinct from the dorsal and anal fins…'. Indeed, the distinct caudal had already been pointed out in the original description of Chaudhuria caudata by Annandale [(1918: 40) 'The caudal fin is united to the dorsal and anal by a low membrane; its rays are completely segregated']. According to Kottelat and Lim (1994) and Britz (personal communication), who looked at numerous Chaudhuria caudata, all of them have a separate caudal. Travers (1984a) contradicted his own description (Travers 1984a). A new species, Chaudhuria fusipinnis Kottelat and Britz (in Kottelat 2000) has recently been described with the caudal fin confluent with the dorsal and anal fins (Britz, personal communication).
In addition to the apomorphic characters mentioned above, Travers (1984b) also reported a few associated apomorphic caudal features. One of them is the tendency to a fused, short, spatulate and non-ray-supporting haemal spine on the second preural vertebra. However, Travers (1984a) himself reported that in M. congicus the haemal spine is ray-supporting, and extends from an autogenous haemal arch in a manner similar to that of most Asian taxa.
(3) According to Travers (1984b) in all African mastacembeloid taxa the scapula foramen lies across the anterior border of the scapula and is bordered by cartilage at its anterior edge ( Figure 5), whereas in all Oriental species including Sinobdella sinensis, Chaudhuria and Pillaia, the foramen is completely bone-enclosed ( Figure 6).
All figures of the pectoral girdle provided by Travers (1984a) agree with this statement. Nevertheless, in his text Travers (1984a) has a conflicting statement reporting that the character is present only in the majority of the African species. Based on this latter statement it might be possible to find African representatives where the scapula foramen is completely bone-enclosed. Indeed, in one of two cleared and stained specimens of M. marchei and also in specimens of M. ellipsifer (BMNH unregistered) and M. paucispinis (MRAC 178099), the scapula foramen was found to be completely bone-enclosed on both left and right scapula (personal observation). In the other M. marchei specimen the scapula foramen was completely bone-enclosed on one side but not on the other. As a result, in the majority, but not in all African species, the anterolateral part of the scapular foramen is enclosed by cartilage and this character does not support the monophyly of the African Mastacembelidae.
Problems with this character are not limited to African species but were encountered also in Oriental species. In Sinobdella sinensis the scapular foramen has two, well-separated holes, a small one situated within the borders of the scapula and completely enclosed by bone and another one situated towards the anterolateral margin of the scapula and clearly  not completely bone-enclosed. At present it is not clear which one is homologous with the foramen found in the other Mastacembelidae, or if both are (Britz, personal communication). Therefore, its use to support the monophyly of any lineage within Synbranchiformes seems rather unconvincing. Based on these observations it is clear that further ontogenetic studies are necessary.
(4) Tendency to have noticeably more caudal than abdominal vertebrae. In his discussion of the vertebral column Travers (1984b) added that in all mastacembeloids, other than those with low vertebral counts, the caudal vertebrae number is generally greater due to an increase in total vertebrae number, which does not occur equally among caudal and abdominal vertebrae. Indeed, based on Travers' (1984b) data (see Table II) and as stated by himself (Travers 1984b), I confirm that some Afromastacembelinae species have a higher total vertebrae number, which is mainly due to a higher caudal vertebrae number when compared to some Mastacembelinae representatives. As a result, indeed, there is generally a greater difference between the caudal and abdominal vertebrae numbers in the former than in the latter species.
However, for the Mastacembelinae species listed by Travers (1984a), the caudal minus abdominal vertebrae number (CV 2 AV) varies from five up to 17 vertebrae in Macrognathus and from seven up to 12 vertebrae in Mastacembelus (see Table II), while in about half of the African species (17/37 species, i.e. ¡46%) the caudal minus abdominal vertebrae is also (17 vertebrae and in about one-third of the species (12/37 species, i.e. ¡32%) (see Table II) it is (12 vertebrae. Therefore, the phylogenetic value of this character, as defined by Travers (1984b), is questionable. Indeed, some African Mastacembelidae species do not seem to have higher caudal minus abdominal vertebral numbers than all or some of the Macrognathus and Mastacembelus species which clearly illustrates the problem of delimiting the groups defined by Travers (1984b). Indeed, Travers (1984aTravers ( , 1984b reported that low total and caudal vertebral counts were also observed in a number of African species, for instance Mastacembelus albomaculatus, M. micropectus, M. plagiostomus, M. platysoma, M. tanganicae and M. zebratus from Lake Tanganyika, and M. brachyrhinus, M. brichardi, M. crassus and M. aviceps from the Lower Zairean rapids (see Table II). These low vertebral numbers are associated with other trends to reduction seen in fishes and the crevice-living Tanganyika species of the Lower Zaire rapids (Travers 1984b). The endemic Lower Zaire River mastacembeloid fauna shows a reversal of the general trend towards a greater number of caudal vertebrae seen in the African mastacembeloids (Travers 1984a).
Finally, the tendency of having noticeably more caudal than abdominal vertebrae could easily have developed several times, independently, within a single lineage or several lineages. As a result, this character is questioned as evidence supporting the monophyly of African Mastacembelidae (Afromastacembelinae).
Until 1958 all African mastacembelid species were described within the genus Mastacembelus. In 1958 a new, monotypic, spiny-eel genus, Caecomastacembelus, was described by Poll with C. brichardi as its type species, collected at Pool Malebo (formerly Stanley-Pool) downstream of the Kinsuka rapids in the Congo River basin. The most important generic characters were the apparent absence of the eyes and a more or less pronounced depigmentation of the skin. This blindness justified the introduction of a new genus (Poll 1958) in spite of the close relationship with the normally eyed species Mastacembelus brachyrhinus Boulenger, 1899, also collected from the lower Congo River basin. The genus Caecomastacembelus was rejected by Roberts and Stewart (1976). They stated that the loss of the eyes and the usually associated depigmentation are reduction characters which, especially in genera such as Rhamdia and Mastacembelus in which the eyes are normally small and of secondary importance, can occur independently and without other notable changes. They justified their position by showing the impossibility to determine whether different species, which may be attributed to such a genus, did evolve independently. Furthermore, they pointed out that contrary to the objectives of a phyletic classification this would inevitably lead to the recognition of polyphyletic taxa. Therefore, they did not recognize any other genus distinguished solely by the absence of the eyes and a superficial depigmentation.
Contrary to Roberts and Stewart (1976), Travers (1984b) did recognize the genus Caecomastacembelus but changed completely the concept of the genus by including several other species. In fact, Travers (1984b) recognized two genera within the Afromastacembelinae: Caecomastacembelus and Afromastacembelus Travers, 1984 with its type species Mastacembelus tanganicae Gü nther, 1893.
The genus Caecomastacembelus was redefined by Travers (1984b: 144) as '…fishes of small to moderately large size. With no pharyngobranchial 2 toothplate and less than five preopercular sensory canal pores. Species with atrophied eye tissues and one (i.e. type for genus) is anoptic. General morphological simplification (by secondary reduction and loss) occurs in microphthalmic and cryptophthalmic species'. Twenty-two species were tentatively assigned to the genus. Travers (1984b) mentioned that probably five additional species had to be included in the genus but that no material was available for study. Two synapomorphies were listed by Travers (1984b) to support the monophyly of this assemblage, the absence of a toothplate on pharyngobranchial 2 and fewer than five preopercular sensory canal pores (see Travers 1984b: Figure 1, Nos 94, 95).
The genus Afromastacembelus was defined by Travers (1984b: 145) as '…fishes of moderate to large size; occur predomin[a]ntly from eastern half of the continent and include species endemic to Lake Tanganyika. All afromastacembeline species, other than those assigned to Caecomastacembelus, provisionally lumped in this ''catch-all'' assemblage (which may not be monophyletic) pending closer examination of the groups interspecific relationships'. Sixteen species were tentatively placed in this genus. Travers (1984b) added that probably three more had to be included but that material was unavailable for study.
No synapomorphies were listed by Travers (1984b) to support the monophyly of this assemblage. Travers (1988) Travers (1988). Nevertheless, Travers (1992aTravers ( , 1992b presented new diagnoses of the genera (Table III). Travers (personal communication) mentioned that a suite of characters defines them and if used separately they could prove unreliable.
As a result of all these nomenclatorial changes, the generic position of some African spiny-eel species is largely confusing and it is especially unclear in which genus the different species should be placed (see also Seegers 1996).
Two main reasons can be given for the present confusion: (1) the many changes from one genus to the other and vice versa introduced by Travers himself without any clear justification (Table IV) and (2) the diagnosis of a genus (Table III) which in many cases is in contradiction with the character states observed. Seegers (1996) illustrated this problem based on the anal and dorsal soft fin ray numbers reported by Travers (1992b) for Aethiomastacembelus, Caecomastacembelus and the frenatus group. These problems were discussed in detail by Vreven and Teugels (1996). Study of the type material of both genera by Vreven and Teugels (1996) revealed several inaccuracies and contradictions in the diagnosis of both genera (Table V). For instance, the type specimens of C. brichardi have only four predorsal vertebrae whereas according to the diagnosis 8-12 predorsal vertebrae characterize species of the genus Caecomastacembelus. As a result, for this character C. brichardi agrees with the diagnosis of Aethiomastacembelus (four to seven predorsal vertebrae) rather than with that of Caecomastacembelus.
Similarly, there are also problems with the monophyletic characters of the Asian Mastacembelinae (see above) and, as a consequence, the validity of genera as separate Table III. Character statement in the diagnosis of the Afromastacembelinae genera following Travers (1988Travers ( , 1992aTravers ( , 1992b. Aethiomastacembelus Travers, 1988 Caecomastacembelus Poll, 1958 Type species: Mastacembelus marchei Sauvage, 1879 Type species: Caecomastacembelus brichardi Poll, 1958 1. Fewer than 95 anal fin rays and tendency to have more dorsal than anal fin rays 1. Usually more than 100 anal fin rays and 100 dorsal fin rays and tendency to have more anal than dorsal fin rays 2. Ten to 20 principal caudal fin rays 2. Eight to 10 principal caudal fin rays 3. Tendency to have a pointed snout 3. Tendency to have a blunt snout 4. Jaw cleft extending beyond posterior nasal 4. Jaw cleft below or anterior to posterior nasal 5. Origin of first dorsal spine dorsal or just posterior to pectoral fin when flat against lateral wall of body 5. Origin of first dorsal spine posterior to pectoral fin when flat against lateral wall of body 6. Four to seven predorsal vertebrae 6. Eight to 12 predorsal vertebrae 7. Body depth greatest midway along length 7. Body height even for most of length 8. Median fins of even height 8. Median fins low and fleshy, increase in height caudally Table IV. Changes in generic placement of the African Mastacembelidae species since Travers (1984b).

Conclusions
On the subfamily level At present, a sister-group relationship between Mastacembelinae (Oriental region) and Afromastacembelinae (Ethiopic region) in the family Mastacembelidae, as proposed and defined by Travers (1984b), could not be confirmed; the use of the subfamilies is rejected to avoid further destabilization of Mastacembelidae nomenclature. The characters proposed to support their monophyly are unconvincing. Evidence is proposed suggesting that the Mastacembelinae at least is a paraphyletic assemblage. Additional research on the phylogeny of the Mastacembelidae and at a higher taxonomic level of the suborder Mastacembeloidei and of the order Synbranchiformes is certainly necessary.

On the generic level
The nomenclature of the African genera has been highly confusing since the taxonomic changes introduced by Travers (1984b). At present, there is no phylogenetic evidence supporting their validity (monophyly) and there are no straightforward diagnostic character(s) available for their diagnosis.
The present situation is harmful to the stability of the generic nomenclature of the African Mastacembelidae. Therefore, I suggest that the present use of the genera Caecomastacembelus and Aethiomastacembelus should be abandoned and that both genera are here placed in synonymy with Mastacembelus. Macrognathus siamensis (Gü nther, 1861)-BMNH unregistered (1 c/s), specimen previously identified as Macrognathus aculeatus from Thailand but according to Roberts (1980) M. aculeatus is not known from Table V. Diagnostic characters of the type species of Caecomastacembelus and Aethiomastacembelus which are not in accordance with the generic definition as given by Travers (1988Travers ( , 1992aTravers ( , 1992b. Aethiomastacembelus Travers, 1988 Caecomastacembelus Poll, 1958 Type species: Mastacembelus marchei Sauvage, 1879 Type species: Caecomastacembelus brichardi Poll, 1958 1. Fifty-three to 57 dorsal fin rays and 56-61 anal fin rays 2. Eleven to 13 principal caudal fin rays 4. Jaw cleft anterior to posterior nasal 5. Anterior origin of first dorsal spine anterior to posterior end of pectoral fin when flat against lateral wall of body 6. Four predorsal vertebrae 8. Dorsal median fin higher than anal one