BULLETIN OF MARINE SCIENCE, 60(1): 6-22, 1997

GARDEN LEPTOCEPHALI: CHARACTERS, GENERIC IDENTIFICATION, DISTRIBUTION, AND RELATIONSHIPS

P. H. J. Castle

ABSTRACT More than 30 species of garden (AnguilIiformes, , ) are now known. worldwide, including several that are undescribed. The leptocephali of Hetero- Bleeker have been described from the Atlantic and East Pacific, but little is otherwise known of the early life history of these eels. Study of the large collection of garden eel larvae in the Dana collections, Copenhagen (around 600 specimens), and other material, now allows the larval form of Meek and Hildebrand to be identified, and the distribution of several Indo-west Pacific species to be mapped. leptocephali have midlateral melanophores on the myotomes as in the larvae of most genera of , and Paraconger and Chiloconger in the Bathymyrinae. Those of Gorgasia have a different pigmentation pattern consisting mainly of compact melanophores on the myosepta below the midlateral level. A third larval type, l.eptocephalus maculatus Della Croce and Castle, 1966 has scat- tered melanophores over the body surface and is identified with Heteroconger hassi (Klau- sewitz and Eibl-Eibesfeldt, 1959), The pigment pattern of Gorgasia, though distinctive, is similar to that of the bathymyrine Ariosoma and its closer allies, to Benthenchelys and other Ophichthidae and unexpectedly to Muraenesox, Additionally, some skeletal characters and the common use of tail-first burrowing behavior suggest that a more comprehensive study of the relationship between the heterocongrines, the bathymyrines and the Ophichthidae in par- ticular, is warranted,

Garden eels are generally agreed to comprise their own subfamily, the Heter- ocongrinae and have distinctive form and habits that separate them from the two other larger subfamilies of the Congridae, the Bathymyrinae and Congrinae, Smith (1989a) summarises what is known about these eels. They occur almost exclusively in the tropical to subtropical ocean, mostly in relatively shallow water, though they have been collected in water as deep as 200-300 m. They are usually observed in groups of several individuals to hundreds, though they may form aggregations of several thousand. They are most often seen burrowed tail-first in, but semi-emergent from loose, coral "sand" off the outer edge of reefs or shores, in localities of some current, though they are also known from mixed broken rubble and sand. Although some 35 or so garden eel species are known worldwide, some of these have yet to be formally described. Most species occur in the Indo-Pacific, with several in the East Pacific and two or three in the Atlantic. The species are clearly separated into the two genera Heteroconger Bleeker, 1853 with its type species H. polyzona Bleeker from central Indonesia, and Gorgasia Meek and Hildebrand, 1923, with type species G. punctata Meek and Hildebrand from Pa- cific Panama. Several other genera have been described but were judged by Bohlke and Randall (1981) to be synonyms of Heteroconger because the principal distinctions were made on the degree of reduction (or absence) of the pectoral and caudal fins. This character is not consistent enough to be used. A thorough appraisal of species inter-·relationships within this remarkable subfamily and its position relative to the other Congridae, has yet to be made. Garden eels are very slender, long-bodied fishes in which the snout is very short and the mouth is terminal and markedly oblique. The eyes are relatively

6 CASTLE: GARDEN EEL LEPTOCEPHALI 7

Gorgasia pll1lctata

~ ~. .,,;a . (L.J'-= .. _<::---=----~ --- •• •~.--8';;--"=-- --

anterior nostril

upper lip

Figure 1. External differences between Gorgasia and Heteroconger. Ethmoid pore is arrowed. large and face more or less directly forwards. The pectoral fin is short, minute or even absent and the caudal fin is also short or fleshy and pointed. The principal skeletal features are the strong, forwardly-directed jaw support associated with the terminal mouth, and the reduction of the neural arches and spines along most of the vertebral column, except for the first few and the penultimate 10-12. The immediate postcranial region and the end of the tail are concerned, respectively, with the flexing horizontally forwards of the anterior part of the body as it emerg- es from the burrow in which the eel lives, and the action of the caudal region in digging tail-first into sediments. Species of Heteroconger have upper lips that form a single, prominent, conflu- ent flange across the face of the snout that encloses tiny, subtubular, forwardly- directed anterior nostrils and the ethmoid sensory pores (Fig. 1). The snout is very short or pug-nosed and the oblique mouth barely reaches to below the an- terior margin of eye. The teeth form broad bands on all surfaces, except that the posterior half of the maxilla is curved outwards and has teeth that tend to be uniserial and enlarged. There are few head pores and the pectoral fin is a tiny, fleshy flap (or absent in H. polyzona except for the occasional specimen in which it is a small tab of skin). The Heteroconger species are typically spotted, speckled or banded in body coloration. In contrast, species of Gorgasia have upper lips that are separate medially and do not enclose the anterior nostrils and ethmoid pores (Fig. 1). The snout is somewhat longer and less blunt, the gape is oblique but longer, reaching to below middle of pupil, and the teeth form narrower bands or are even uniserial on the vomer. The complement of head pores is mOrecomplete though those of the lateral line proper may be rather widely spaced and fewer than the number of vertebrae. With one or two notable exceptions the body color of the Gorgasia species is dull or nondescript. 8 BULLETINOFMARINESCIENCE,VOL.60, NO.I, 1997

The Heterocongrinae is currently regarded as a highly specialized congrid sub- family lineage, more closely related to the Bathymyrinae (Ariosoma, Bathymyrus etc,) than to the Congrinae (Conger and its many allies). This view was developed more thoroughly by Smith (1989a) from comprehensive studies on adults of all but a few congrid genera, Larval characters were apparently not explored to any extent by Smith (l989a). Though the larvae of Heteroconger had been known for some time (Blache, 1977; Raju, 1974, 1985; Smith, 1989b), those of Gorgasia had not been described under that name, except incidentally and without detailed comment by Castle (1984). As will be made clear later, Gorgasia larvae were in fact described earlier by D'Ancona (1928) from the Red Sea, by Blache (1977) from the East Atlantic and by Smith (1989b) from the western Atlantic, but not identified as such. Raju (1974, 1985) erred in referring his 65-95 mm TL specimens from the Pacific coast of Mexico (SIO 61·-248, 75-249) that had midlateral melanophores, to Gor- gasia punctata. These were apparently Heteroconger larvae with regenerated tails and the metamorphic larva of G, obtusa (Garman, 1899) from Pacific Panama was apparently that of the bathymyrine eel Chiloconger labiatus Myers and Wade, 1941 (D. G, Smith, pers. comm.). The discovery of metamorphic specimens of Gorgasia maculata Klausewitz and Eibl-Eibesfeldt, 1959, allows Gorgasia larvae to be identified, collections of garden eel larvae to be adequately sorted, and an assessment of the contribution that larval characters might make to exploring relationships of the Heterocongri- nae to proceed. Accordingly, this paper distinguishes between the larvae of the two garden eel genera, describes the distribution of larvae of several species in the Indo-Pacific and attempts to evaluate the relationships of the garden eels based on larval and some adult characters,

MATERIALS AND METHODS

Specimens discussed in this paper were derived from a variety of sources. Leptocephali were mostly from the Danish DANAExpedition Round the World 1928-1930 collections now held in the University Zoological Museum, Copenhagen (ZMUC). A few others were from the western South Pacific col- lections of ORSTOM, Noumea. A range of adult specimens and species were loaned mainly from United States institutions. Institutional acronyms, where available, are as given in Leviton et a1.(1985). Vertebral counts were made from radiographs. My observations show that garden eels seldom lose their tail tips, as is very common in eels such as species of the congrid eel genus , so vertebral counts can be considered reliable. There is a likelihood of some small error in estimating total numbers of larval myome:res because of difficulty in counting the last one or two myomeres. This might explain slight differences in mean vertebral counts and myomeres given for the species discussed in this paper,

GENERIC CHARACTERS AND IDENTIFICATION OF THE LARVAE The larvae of Heteroconger (Fig. 2), have a short gut (anus placed near mid- length in full grown individuals) and in this respect are like those of Muraenesox in the Muraenesocidae. In contrast, the anus is subterminal in the Ariosoma group, and well back in other congrids. However, in pigmentation Heteroconger larvae are rather like those of several genera of Congrinae. They have a series of paired, ocellate melanophores along the gut and a midlateral series of spaced, ocellate or rather diffuse melanophores on the surface of the myomeres, with one on each myomere in H. longissimus (Gunther, 1970), in Heteroconger sp. (Blache, 1977) and more or less so in H. luteolus Smith, 1989. These are Atlantic species, but in the Indo-west Pacific species the midlateral melanophores are fewer and much more widely spaced. Metamorphic and juvenile specimens were described from the East Pacific by Raju (1985) under the names H. digueti (Pellegrin, 1923) and CASTLE: GARDEN EEL LEPTOCEPHALI 9

Figure 2. Heleroconger sp., leptocephali (A-C) and metamorphic specimen (D-E). A, 39.3 mm TL, ORSTOM station S2, 8 June 1962; B-C, 92.3 mm TL, ORSTOM station S2, 8 June 1962; D-E, 76 mm TL, ZMUC P312678, Vanadoro Harbour, Philippine Islands, 22 July 1908; F, ZMUC P312678 (same as D and E), pigmentation of posterior part of gut. Scales indicate I mm, scale in A is 5 mm.

H. canabus (Cowan and Rosenblatt, 1974). These had similar pigmentation and matching myomere/vertebral numbers and some also had recognisable external characters of Heteroconger, confirming this identification. Remarkably, considering that species of Gorgasia are widespread in the Indo- west Pacific, as adults, their larvae have not until now been formally recognised. The discovery of several metamorphic specimens having the pigmentary char- acters of the larvae as well as juvenile characters of Gorgasia in the collections of the University Zoological Museum, Copenhagen has enabled its larval form to be confidently identified. These specimens were amongst a number of leptocephali and metamorphics of Gorgasia and Heteroconger sp. collected early this century in the Philippine Islands by the Bureau of Fisheries ALBATROSSPhilippine Ex- pedition 1907-1909. The character of greatest relevance to the identification of the Gorgasia larval type is the presence in the metamorphics of a series of compact, somewhat ver- tically elongated melanophores on many of the myosepta below the midlateral level (Fig. 3D, E; lOA). This pattern, particularly of the melanophore form and position on the myosepta, rather than in their number along the body, closely matches that of a large group of larvae in the DANAcollections. One of the metamorphics also has separate lip flanges not enclosing the anterior nostrils (Fig. 3D), as in juvenile and adult Gorgasia species. Blache's (1977) figures of the species inferomaculatus, from the Gulf of Guinea, reproduced here (Fig. 4A, B), show a similar larval pigmentary pattern, and also an abbreviated caudal fin (Fig. 4B), similar to that of juveniles and adults of several heterocongrine species. Consequently there is little doubt that L. inferomaculatus is a Gorgasia species that should therefore be referred to 10 BULLETIN OF MARINE SCIENCE. VOL. 60. NO. I. 1997

~-;>-=~=..,""..=, =" ,,= ====

~'

___ v : •• ~ =~~====-==:;:::.=/' Figure 3. Gorgasia maculata, leptocephali (A-C) and metamorphic specimen (D-E). A, 19.8 mm TL, ZMUC P312680, DANA station 3884 IV; B-C, 35.8 mm TL, ZMUC P312681, DANA station 3778 I; D-E, 61.5 mm TL, ZMUC P3l2679, Vanadoro Harbour, Philippine Islands, 22 July 1908. Scales indicate 1 mm. as G. inferomaculata. This establishes the presence of Gorgasia in the eastern Atlantic where it has not previously been reported. Smith's (1989b) record of two somewhat similar larvae from the western At- lantic and Atlantic Panama, i.e., as L. inferomaculatus and "Congridae Genus A species ~', suggests that Gorgasia occurs there also though the latter has a pig- mentary pattern more like that of L. inferomaculatus as described by B1ache (1977). The eventual discovery of adults of Gorgasia species on both sides of the Atlantic can be expected. D' Ancona (1928) described several types of leptocephali from the Red Sea as species of Leptocephalus. Included were L. cotroneii, obviously a Gorgasia and L. congroides and L. magnaghii, possibly species of Heteroconger. D' Ancona's figure of myoseptal pigment in L. cotroneii is reproduced here (Fig. 4C). Lepto- cephalus cotroneii has a myomere number which more or less matches the ver- tebral count of Bohlke and Randall, 1984 from the Indo-west Pacific. The single specimen of L. cotroneii does not allow detailed comparison of the two species to determine whether (or not) they are the same. Gorgasia preclara may have been present in the Red Sea in 1923 when L. cotroneii was collected, though John E. Randall informs me that despite extensive recent col- lecting there he has not seen G. preclara, a distinctively-marked species. The recognition of two distinctive larval types of the garden eels agrees well CASTLE: GARDEN EEL LEPTOCEPHALI 11 B

Figure 4. A-B, Leptocephalus inferomaculatus Blache, portion of body at level of kidney (A, 70 mm TL) and tail tip (B, 75 mm TL), figures reproduced from Blache, 1977; C, Leptocephalus cotroneii D' Ancona, portion of body below midlateral line to show myosepta! pigment, 96 mm TL. figure redrawn from D' Ancona, 1928. with the current view that there are just the two genera Heteroconger and Gor- gasia in the Heterocongrinae. However, there is apparently a third major larval type, clearly more similar to that of Heteroconger than to Gorgasia. This is rep- resented by Leptocephalus maculatus Della Croce and Castle, 1966, originally described from the Mozambique Channel, western Indian Ocean (holotype and only specimen 47.9 mm TL, Institute of Zoology, University of Genoa) (Della Croce and Castle, 1966). This has gut pigment of the "conger" -type, i.e., with a more or less paired series of stellate melanophores along the gut from oesophagus to anus and the body surface rather evenly covered with moderate-sized, spaced melanophores (Fig. 5A-C). Midlaterally, the melanophores seem to follow a linear series, as in Heteroconger. Several other larval types have melanophores present over the whole lateral body surface. They are regularly arranged in the chlopsids Kaupichthys hyopro- roides (Stromman, 1896) and Catesbya pseudomuraena (Bohlke and Smith, 1968), the larval ophichthid Leptocephalus mononucleus Blache, 1977 and scat- tered in the ophichthid Apterichtus kendalli (Gilbert, 1891) and in an unidentified muraenid larva described by Castle (1984). However, the pattern in each of these is individually distinctive and unlike that of Gorgasia so there is little likelihood of specimens of L. maculatus being confused with these. Larvae of the "maculatus" -type are well represented in the Dana collections by more than 50 specimens. The only metamorphic specimen available is one from the western Indian Ocean (CAS 35466, ca. 79 mm TL, Fig. 50, E). This has similar lateral body pigment, a blunt snout, and upper lips continuous across the face of the snout, thus identifying it with Heteroconger. The myomere range and mean of these "maculatus" -type larvae almost exactly matches the vertebral range and mean for Heteroconger hassi (Klausewitz and Eibl-Eibesfeldt, 1959) first described as an adult from the Andaman Islands, Indian Ocean. Since then the species has been reported as adults from many localities from the western 12 BULLETIN OF MARINE SCIENCE, VOL. 60, NO.1, 1997

Figure 5. Heteroconger hassi, leptocephali (A-C) and metamorphic specimen (D-E). A, 34.3 mm TL, ZMUC P312682, DANA station 3778 II; B-C, 79.8 mm TL, ZMUC P312683, DANA station 3778 II; D-E, ca. 79 mm TL, CAS 35466, metamorphic. Scales indicate 1 mm, scale in A is 5 mm.

Indian Ocean to southern Japan and Samoa (Fig. 8) and is apparently the most common garden eel in the Indo-west Pacific. Among the approximately 12 species of Heteroconger known from the Indo- west Pacific, only H. hassi and the recently described H. taylori Castle and Rand- all, 1995 from Papua New Guinea and Indonesia have vertebral counts agreeing with those of the CAS metamorphic. Heteroconger taylori happens to be rather similarly colored to H. hassi, except that it lacks the prominent black body patch- es. On the basis of the widespread occurrence of H. hassi, the matching distri- bution of "maculatus" -type larvae, and the match of vertebral and myomere num- bers, L. maculatus is referred to H. hassi. Except for its very distinctive body coloration, H. hassi differs very little externally from its congeners and no great significance is accorded the difference in its larval pigmentary pattern. However, it may be significant that there are skeletal differences between H. hassi and H. longissimus as illustrated by Klausewitz and Eibl-Eibesfeldt (1959) and Smith (1989a) respectively (skull in particular), perhaps sufficient to justify the revival of the prior generic name Xarifania Klausewitz and Eibl-Eibesfeldt, 1959 for H. hassi. This can only be confirmed, or otherwise, by a thorough osteological com- parison of the species.

DISTRIBUTIONOF GARDENEELSANDTHEIRLEPTOCEPHALIINTHE INDO-WESTPACIFIC The DANAcollections include about 600 garden eel larvae from the Indo-west Pacific, ranging from about 10 rom to more than 100 rom TL. At least six species Figure 6. Distribution of leptocephali and adults of Heteroconger sp. Leptocephali (DANAstations 1928-1930): n = 194; adults (USNM 235785): n = 55. appear to be represented and, based on myomere numbers in the larvae and ver- tebral numbers in adults, two can be identified with reasonable certainty. The uncertainty about the others is because there is some near coincidence or overlap of vertebral numbers amongst the adult species known from the area. The distributions of larvae and adults of the two identified species and one unidentified species are summarised here (Figs. 6-8) to demonstrate particularly the apparent importance of the Indonesian area for the Heterocongrinae as a group. However, the distributions may simply reflect the thoroughness (or lack of it) of collecting by divers for adults, and the sampling program of the DANA during

120" ISO- 180" 40'

30' .':

20'

., 10' '. '. -., 0'

10·

tl.:'. Q

30- 30'£ 60' 120' 150'£ 180' Figure 7. Distribution of leptocephali and adults of Gorgasia macu/ata. Leptocephali (DANAstations 1928-1930): n = 107; adults (from various museum holdings): n = 32. Not included is one specimen from Cebu, Philippine Islands and a visual record from eastern Papua New Guinea. 14 BULLETIN OF MARINE SCIENCE, VOL. 60. NO. I, 1997

120'

Figure 8. Distributions of leptocephali and adults of Heteroconger hassi, Leptocephali (DANAstations 1928-1930, ORSTOM 1962): n = 84; adults (from published records and known museum holdings): n = 73, Not included are a museum record from American Samoa and a visual record from Tonga (Malcolm Francis, pers, comm),

1930 for the leptocephali. The DANA collected most intensively in the northeast Indian Ocean during its search for the spawning area(s) of anguillid species, particularly close into the west coasts of Sumatra and Java (Carlsberg Foundation, 1934), to the extent of 100 stations, each of multiple samples. The vessel sampled rather little eastwards in the Coral Sea and Bismarck Sea where adult garden eels would be expected in numbers, and thus a source area for their larvae. A more extensive treatment of the DANA larval collection (descriptions of the various species, identifications, larva distributions and growth) is planned as a separate study.

BRIEF SPECIES ACCOUNTS Heteroconger sp. Figures 2, 6

Taenioconger chapmani (?) Herre, 1923: 152, pI. 3 (original description), holotype, 690 mm TL, adult, Dumaguete, Philippine Is, ca, 1914, This specimen is no longer extant.

Material Examined.-A total of 134 specimens from 52 DANAstations, 13,2-106,0 mm TL, Diagnosis.-Heteroconger larvae with a series of widely spaced, midlateral, dif- fuse melanophores, and with 188-208 myomeres. Remarks.-These larvae could not be identified for certain. Heteroconger chap- mani, known only from the holotype that was destroyed in Manila during the early part of World War II, was a very elongate Heteroconger that probably had a high number of vertebrae. Heteroconger cobra Bohlke and Randall, 1984 also has a high vertebral number (198-204) and could possibly be the adult of these larvae. Victor G. Springer (USNM) collected a large number of a small from near Fiji in 1982 and this may also possibly be the adult of these larvae. In 55 adults from the latter collection, the vertebral number was 195.13 CASTLE:GARDENEEL LEPTOCEPHALI 15

(mean) and 188-208 (range); in the 134 leptocephali, myomeres were 196.08 (mean) and 186-204 (range).

Gorgasia maculata Klausewitz and Eibl-Eibesfeldt, 1959 Figures 3, 7

Gorgasia maculata Klausewitz and Eibl-Eibesfeldt, 1959: 147, figs. 13, 14, 17 (original description), holotype SMF 4619, 528 mm TL, adult, Nicobar Islands, Indian Ocean, 4 September 1958.

Material Examined.-A total of 112 specimens from 46 DANA stations, 13.3-48.1 mm TL. Diagnosis,-Gorgasia larvae with compact, ovoid melanophores on nearly every myoseptum below the midlaterallevel, and with 164-] 83 myomeres. Remarks.-These specimens are identified specifically with G. maculata because of their close agreement in myomere number with the number of vertebrae in the adults. In 28 adults, the vertebral number was 172.64 (mean) and 167-178 (range); in the 112 leptocephali, myomeres were 173.80 (mean) and 164-183 (range).

Heteroconger hassi (Klausewitz and Eibl-Eibesfeldt, 1959) Figures 5, 8

Xarifania hassi Klausewitz and Eibl-Eibesfeldt, 1959: 138, figs 6, 8-10, 15, 16 (original description), holotype SMF 4556, 357 mm TL, adult, Addu Atoll, Maldive Archipelago, Indian Ocean, 25 December 1957. Four paratypes SMF 4557-4560. Leptocephalus maculatus Della Croce and Castle, 1966: 156, fig. IA-C (original description), ho- lotype Institute of Zoology, University of Genoa, 47.9 mm TL, leptocephalus, ANTON BRUUN Station 404 A, Mozambique Channel, 10 October 1964.

Material Examined.-A total of 56 specimens from 48 DANA stations. A metamorphic specimen CAS 35466, ca. 79 mm TL, Grande Comore Id, western Indian Ocean, 19 February 1975. Diagnosis.-Heteroconger larvae with scattered, stellate melanophores over the lateral body surface that midlaterally form a more or less distinct line, and with 165-183 myomeres. Remarks.-In 24 adults, the vertebral number was 169.63 (mean) and 164-175 (range); in the 56 leptocephali, myomeres were 172.39 (mean) and 165-183 (range). As shown in Figure 8, larvae were collected in nearly every area, as were adults. Though there were few larvae less than 30 mm TL, the coincidence of location of larvae and adults suggests that the species probably spends its early life close to the same area as adult distribution.

LARVAL CHARACTERS AND THE RELATIONSHIPS OF THE HETEROCONGRINAE Among the Anguilliformes, the Congridae (with about 25 genera) is second only to the Ophichthidae (with about 55) in number of genera. These two families together make up about half of the currently recognised genera in this large group of fishes. The ophichthids live mainly in soft sediments of tropical near shores, while the congrids live principally on the outer shelf and upper slope. Smith (1989a), having worked extensively with the congrids and other families of eels over many years, proposed a scheme of relationships which recognised, with the addition of the Heterocongrinae, the fundamental dichotomy of the congrids ear- lier delineated in broad terms by Asano (1962). Smith's scheme (Fig. 9) divided the family into the three subfamilies Bathymyrinae (Ariosoma and its allied five genera), Congrinae (Conger and about 18 other genera) and the Heterocongrinae 16 BULLETIN OF MARINE SCiENCE. VOL. 60. NO. i, 1997

Heterocollger

] HETEROeONGRIN" Gorgasia Paracollger

Clziloconger ----- Ariosoma

BATHYMYRINAE

19 genera ] CONGRINAE

Figure 9. Cladogram of the Congridae, based largely on Smith (l989a) but with addition of Poecil- oconger. Details of Congrinae not shown.

(Heteroconger and Gorgasia). The Heterocongrinae appear in this scheme as the sister group of the Bathymyrinae. Smith's (1989a) arrangement is principally based on adult characters with only brief mention of the relevance of characters of the leptocephali. Insofar as only about half of the 25 or so congrid genera are known in their larval form, it would seem quite proper that no significant attempt was made at that stage to incorporate larval characters in any hypothesis of congrid phylogeny. A further point is that there are other types of congrid larvae that do not seem to be referable to any of the known genera, suggesting that there may be additional congrid genera yet to be discovered. Two of these are Leptocephalus stenorhynchus Castle, 1964 (pos- sibly a Kaup, 1859) and L. laureus Castle, 1964 (possibly Promyl- lantor purpureus Alcock, 1890). Currently, I have adults of two congrid species that apparently belong to undescribed genera. The Bathymyrinae of Smith's (1989a) scheme comprise the genera Ariosoma (possibly including Alloconger), Bathymyrus, Parabathymyrus, Paraconger, and Chiloconger. The larval forms of all of these are known. (Gunther 1871) can be added to this group since it has unsegmented fin rays, as do the other Bathymyrinae, but not the Congrinae in which the fin rays are segmented in all genera. Its larval form is unknown, and while it is unwise to predict what characters it might have (take for example the strangely aberrant ophichthid-like larval form of amongst the Congrinae-Smith and Leiby, 1980), it seems likely that it would resemble that of Ariosoma. The relative completeness of knowledge of the larval forms and adults of the bathymyrine-heterocongrine clade, though not of the Congrinae, suggests that it might be worthwhile to consider the relevance of larval characters to the phylog- eny proposed by Smith (1989a). Regrettably, such an analysis is likely to be constrained by the inadequate information available on adult Heterocongrinae, except for just a few species that have been studied. As has already been outlined, Heteroconger leptocephali differ from those of Gorgasia most distinctly in their pigment pattern. The midlateral series of rather diffuse melanophores on the surface of the myotomes, together with the series CASTLE: GARDEN EEL LEPTOCEPHALI 17 along the body wall from oesophagus to anus in Heteroconger (Fig. 2), contrasts strongly with the irregular series of compact melanophores on many of the my- osepta below the midlateral level and the additional supraintestinal series along the posterior part of the gut in Gorgasia (Fig. 3F). Heteroconger larvae thus conform not only with those of genera of Congrinae whose larval forms are known (with various modifications, deletions and additions throughout the genera) but, interestingly, also with those of Paraconger and Chiloconger, both of which are members of the Bathymyrinae (Smith, 1989a), at least according to adult characters. It follows that the most general larval pigment pattern in the family Congridae, and incidentally in a somewhat similar form in the Muraenesocidae, a family that has been allied with the Congridae, is that seen in the genera of Congrlnae, in some Bathymyrinae, and in Heteroconger. This must have evolved before the Bathymyrinae/Heterocongrinae, or more correctly their common an- cestor, which split from the common ancestor of the Congrinae. The pigment pattern should therefore be viewed as derived for the Congridae as a whole. It is noted that in their bulky body form, reaching as large as 400 mm TL, Ariosomal Bathymyrus larvae contrast markedly with those of Gorgasia (100 mm maximum TL). This may possibly be a reflection of Gorgasia species having a short larval life close to shore, while for Ariosoma species and their close relatives it is likely to be longer, and in the open ocean. The pigment pattern on the myosepta of Gorgasia larvae (Fig. lOA) was first judged to resemble that of some Bathymyrinae, e.g., Ariosoma (Fig. lOB), Bath- ymyrus, Parabathymyrus, perhaps Poeciloconger, the ophichthid Benthenchelys (Fig. 1OC),of several other Ophichthidae (Fig. 100 and E) and of Muraenesox (Fig. IOF). Gorgasia also has the supraintestinal series of melanophores like Ario- soma (Fig. 3F). This similarity in pigmentation suggested a closer relationship between these groups than previously suspected. However, as the illustrations of these show, the patterns are rather different in detail, and are therefore judged in each case to be unique features of the groups concerned. If it is accepted that on adult morphology the affinities of the Muraenesocidae are both with specialized congrids such as Gilbert and the Nettastomatidae, the larval pigmen- tation of Muraenesox is seemingly anomalous. Alternatively, are those larval types that possess myoseptal pigment in one form or another like that of Gorgasia, merely retaining in common the ancestral (plesiomorphic) pigmentation and there- fore are of no significant value in reconstructing the phylogeny of these groups of eels? The purpose, in part, of this paper has been to evaluate the significance, if any, of larval characters in determining relationships of the garden eels. The outcome has been that larval pigment patterns add rather little to the broad pattern of relationships that have already been established from study of the adults (Smith, 1989a). In fact, the addition of larval characters to the equation has in some respects complicated the issue. It seems clear that while the leptocephali of Heteroconger possess the rather simple, generalized pigment pattern of most genera of the Congridae, those of Gorgasia have a unique pattern of somewhat greater complexity. On the other hand, in external features, adults display the reverse trend: Heteroconger has several derived features, including the fusion in the midline of the upper lips, thus enclosing the anterior nostrils, the marked foreshortening of the snout and jaws so as to permit (presumably) a fair degree of binocular vision, and the reduction of the cephalic sensory pore system and pectoral fin. Gorgasia displays what appears to be the more primitive state in these characters, including the separate upper lips and anterior nostrils, a longer snout and jaws, more laterally placed 18 BULLETIN OF MARINE SCIENCE, VOL. 60, NO. I, 1997

Figure 10. Melanophore patterns on myosepta. A, Gorgasia maculata, 35.8 mm TL, at level of anus, ZMUC P312681; B, Ariosoma scheelei, 141 mm TL, at myomere 60-61, ORSTOM Diaphus 10, station 267; C, Benthenchelys cartieri, 51 mm TL, at myomeres 25-29, ZMUC; D, Ophichthidae, 70.5 mm TL, at myomeres 66-69, ORSTOM Cyclone I, station 17; E, Ophichthidae, 104 mm TL, at myomeres 60-62, ORSTOM Caride II, station II A. F, Muraenesox bagio, 80 mm TL, at myomeres 73-75, Nobioka Bay, Kyushu, Japan. Scales are 1 mm. CASTLE: GARDEN EEL LEPTOCEPHALI 19

c o ~>

Figure 11. Dorsal and lateral views of neurocrania of A, Heteroconger longissimus (Heterocongri- nae); B, Benthenchelys cartieri (Ophichthidae); C, Ariosol1la balearicul1l (Bathymyrinae); D, Gorgasia maculata (Heterocongrinae); E, Paraconger caudilil1lbatus (Bathymyrinae); F, Myrophis vafer (Ophichthidae). Figures A, C, E from Smith, 1989a; Figure B from Castle, 1972; Figure D from Rosenblatt, 1967; Figure F from McCosker, 1977. Abbreviations: LEP lateral ethmoid process; SO, supraoccipital.

eyes, a more complete set of head pores, and a small but prominent pectoral fin. There is no clear pattern in the degree of reduction of the caudal fin because the whole range is seen across the species of both Heteroconger and Gorgasia. An analysis of skeletal features further adds to the confusion. Rosenblatt (1967), in his osteological study of Gorgasia punctata, concurred with Bohlke's (1957) view that the garden eels should be accorded subfamily status within the Con- gridae. He further noted that the heterocongrines share characters both with the Bathymyrinae, including unsegmented fin rays, reduced caudal, and well ossified lateral line scales, and the Congrinae, including the presence of a supraoccipital 20 BULLET1N OF MARINE SCIENCE, VOL. 60, NO. I. 1997

(SO in Fig. 11), more numerous abdominal vertebrae, and gas bladder free from the parapophyses. Gorgasia, but not Heteroconger, has a lateral ethmoid process (LEP in Fig. 11) i.e., a lateral extension of the intermaxillary-ethmoid that ap- parently assists in bracing the maxilla against the side of the skull. The bathy- myrines have this structure but not the congrines. Smith's (1989a) cladogram took into account the above similarities and differences, amongst others, in placing the Heterocongrinae alongside the Bathymyrinae rather than aligning it with the Con- grinae. Chiloconger has a lateral ethmoid process and other features that show it to be a bathymyrine and Poeciloconger can be added here for similar reasons. Poeciloconger fasciatus Gunther, 1871 in Hawaii is known to bury tail-first into sediments, as does P. kapala Castle, 1990 at the Kermadec Islands, South Pacific. The Poeciloconger species have robust bodies and dig substantial burrows from which only the forepart of the head is emergent and to which they can return after leaving to forage. The species described and figured as Ariosoma anago (Temminck and Schlegel, L842) by Asano (1962) is probably not that species, but apparently behaves similarly. The heterocongrines are generally much longer and more slender bodied than the bathymyrines (an Indonesian specimen of Gor- gasia barnesi recently collected was approximately 1200 mm TL) and many do not have permanently maintained burrows. Their vertebral column is organised differently: the immediate postcranial vertebrae have low neural arches without prominent spines and the caudal skeleton is broad but compact; in the bathymy- rines there are relatively well developed neural spines and a rather slender caudal complex. Presumably these differences relate to the specific ways in which the garden eels and the bathymyrines have separately exploited the tail-first burrowing mode of life. The ophichthids have moved down this adaptive pathway more profoundly. They bury tail-first with very little of their bodies emergent, though Robert J. Lavenberg informs me that he has seen Paraletharchus sp. in the East Pacific behaving much as garden eels do, i.e., with their bodies partly emerged and erect, and in fact in the company of Gorgasia punctata. I have observed the ophichthid, Muraenichthys australis Macleay, 1881 similarly to form groups of semi-emergent individuals on sea grass beds in northern New Zealand. The skeletal features of ophichthids differ considerably from those of heterocongrines and bathymyrines but the extraordinary similarity of the skull of Heteroconger longissimus as il- lustrated by Smith (1989a), to that of Benthenchelys cartieri FowLer, 1934 (Ophichthidae, Benthenche1yinae) (Fig. lIB) rather than to any other ophichthid e.g., Myrophis (Fig. lIF) may be evidence of a close phylogenetic link between these groups of eels. The pigment pattern of Gorgasia and some features of its skeleton are at least suggestive of such a connection with the Ophichthidae. This is not wholly un- expected since the ophichthids were, on adult characters, included as members of the suborder Congroidei by Robins (1989). The mutual possession of the tail-first burrowing habit of adults of most Ophichthidae, the Heterocongrinae, and Ario- soma and Poeciloconger among the Bathymyrinae, coupled with somewhat sim- ilar pigmentary patterns in their larvae is strongly suggestive of a relationship among these eels. Rosenblatt (1967) viewed the behavioral and structural simi- larities of the Heterocongrinae and Ophichthidae as superficial, parallel adapta- tions to a similar mode of life. An alternative view is that these similarities may well have a more deeply-rooted and ancient evolutionary significance. In view of the different, though in this case rather equivocal conclusions that might be drawn from considering larvae and adults, and the substantive division of the eel life history into larva and adult, the question quite fairly arises as to CASTLE: GARDEN EEL LEPTOCEPHALI 21 whether we might expect adults and leptocephali to have followed separate phy- letic pathways. Should we therefore expect to uncover more meaningful evidence for relationships separately among the characters displayed on the one hand in the leptocephali and on the other by the adults? Alternatively, would it not be more fruitful to evaluate characters of the various life history stages as an inte- grated whole? These considerations are, of course, not the sole preserve of the eels, nor indeed of fishes generally, since any organism which passes through its life history in a similar manner must be considered in a similar light.

ACKNOWLEDGMENTS

I thank the following for arranging and helping with the loan of specimens and providing infor- mation on which this paper is based: E. B. Bohlke (ANSP), J. Fong (CAS), M. Francis (National Institute of Water and Atmosphere, New Zealand). R. J. Lavenberg (LACM), M. McGrouther (AMS), N. Mochioka (Fukuoka University), J. Nielsen (ZMUC), J. E. Randall (BPBM). D. G. Smith (USNM) provided valuable information; I especially wish to thank him as a long-time colleague with whom I have had many fruitful discussions on eels and leptocephali. I thank M. J. Miller for commenting extensively on the manuscript. Jo Llewellyn assisted with radiographs and the unenviable task of counting myomeres and vertebrae. The Internal Grants Committee and the Leave Committee, Victoria University of Wellington, provided funds for research support and travel respectively for which I am grateful.

LITERATURE CITED

Asano, H. 1962. Studies on the conger eels of Japan. Bull. Misaki Mar. BioI. Inst., Kyoto Univ. I: 1-143. Blache, J. 1977. Leptocephales des poissons anguilliformes dans la zone sud du Golfe de Guinee. Faune Tropic. 20: 1-381. Bohlke, J. 1957. On the occurrence of garden eels in the western Atlantic, with a synopsis of the Heterocongrinae. Proc. Acad. Nat. Sci. Phil ad. 109: 59-79. --- and J. E. Randall. 1981. Four new garden eels (Congridae, Heterocongrinae) from the Pacific and Indian Oceans. Bull. Mar. Sci. 31: 366-382. Carlsberg Foundation. 1934. Introduction to the reports from the Carlsberg Foundation's Oceano- graphical Expedition Round the World 1928-30. Dana Rep. 1: 1-130. Castle, P. H. J. 1972. The eel genus Benthenchelys (Fam. Ophichthidae) in the Indo-Pacific. Dana Rep. 82: 1-30. ---. 1984. Notacanthiformes and Anguilliformes: development. Pages 62-93 in Spec. Publ. Am. Soc. Ich. Herpet. 1. 760 p. D' Ancona, U. 1928. Murenoidi (Apodes) del Mar Rosso e del Golfo di Aden. R. Com. Talassogr. Ital. 146: 1-146. Della Croce, N. and P. H. J. Castle. 1966. Leptocephali from the Mozambique Channel. Boll. Mus. Inst. BioI. Univ. Genova. 34(21 I): 149-164. Herre, A. W. C. T. 1923. A review of the eels of the Philippine Archipelago. Phil. J. Sci. 23: 123- 236. KIausewitz, W. and 1. Eibl-Eibesfeldt. 1959. Neue Rohrenaale von den Maldiven and Nikobaren (Pisces, Apodes, Heterocongridae). Senckenberg. BioI. 40: 135-153. Leviton, A. E., R. H. Gibbs, Jr., E. Heal and C. E. Dawson. 1985. Standards in ichthyology and herpetology: Part I. Standard symbolic codes for institutional resource collections in ichthyology and herpetology. Copeia 1985: 802-832. McCosker, J. E. 1977. The osteology, classification, and relationships of the eel family Ophichthidae. Proc. Calif. Acad. Sci. ser. 4, 41: 1-23. Raju, N. S. 1974. Distribution, growth, and metamorphosis of leptocephali of the garden eels, Tae- nioconger sp. and Gorgasia sp. Copeia 1974: 494-500. ---. 1985. Congrid eels of the eastern Pacific and key to their leptocephali. NOAA Tech. Rep. 22: 1-19. Robins, C. R. 1989. The phylogenetic relationships of the anguilliform fishes. Pages 9-23 in E. B. Bohlke, ed. Fishes of the western North Atlantic. Part Nine. Volume One: Orders Anguilliformes and Saccopharyngiformes. Sears Found. Mar. Res. New Haven. 1,055 p. Rosenblatt, R. H. 1967. The osteology of the con grid eel Gorgasia punctata and the relationships of the Heterocongrinae. Pacif. Sci. 2 I: 9 1-97. 22 BULLETINOF MARINESCIENCE,VOL. 60, NO.1, 1997

Smith, D. G. 1989a. Family Congridae. Pages 460-567 in E. B. Bohlke, ed. Fishes of the western North Atlantic. Part Nine. Volume One: Orders Anguilliformes and Saccopharyngiformes. Sears Found. Mar. Res. New Haven. 1,055 p. ---. 1989b. Family Congridae: leptocephali. Pages 723-763 in E. B. Bohlke, ed. Fishes of the western North Atlantic. Part Nine. Volume Two: Leptocephali. Sears Found. Mar. Res. New Haven. 1,055 p. --- and M. M. Leiby. ]980. The larva of the congrid eel Acromycter alcocki (Pisces: Anguilli- formes), and the distinction between congrid and ophichthid larvae. Proc. BioI. Soc. Wash. 93: 388-394.

DATE ACCEPTED: February 23, ]996.

ADDRESS: School of Biological Sciences, Victoria University of Wellington, P.O. Box 600, Wellington 6000, New Zealand.