CHAPTER 14 Interrelationships of Aulopiformes CAROLE C. BALDWIN and G. DAVID JOHNSON Department of Vertebrate Zoology National Museum of Natural History Smithsonian Institution Washington, D.C. 20560 I. Introduction In 1973, Rosen erected the order Aulopiformes for all non-ctenosquamate eurypterygians, that is, the Ini- omi of Gosline et al. (1966) minus the Myctophiformes (Myctophidae and Neoscopelidae). Rosen's aulopi- forms included 15 families (Alepisauridae, Anotopter- idae, Aulopidae, Bathysauridae, Bathypteroidae, Chlorophthalmidae, Evermannellidae, Giganturidae, Harpadontidae, Ipnopidae, Omosudidae, Paralepidi- dae, Scopelarchidae, Scopelosauridae, and Synodon- tidae) and 17 fossil genera, a morphologically diverse gro^u^p of benthic and pelagic fishes that range in habi- tat from estuaries to the abyss. Rosen (1973) diagnosed the Aulopiformes by the presence of. an elongate uncinate process on the second epibranchial (EB2) bridging the gap between a posterolaterally displaced second pharyngobran- chial (PB2) and the third pharyngobranchial (PB3). He noted that paralepidid fishes lack this distinctive configuration of EB2 and thus questioned their place- ment in the order. Subsequently, R. K. Johnson (1982) recognized that certain paralepidids (Paralepis, and Notolepis) have an enlarged EB2 uncinate process but questioned Rosen's use of this feature to diag- nose aulopiforms because he believed the same condition occurs in neoscopelids. Instead he sug- gested that the modification is a primitive iniome condition and that the small EB2 uncinate process of myctophids is secondarily derived. R. K. Johnson (1982) resurrected a more traditional view of iniome relationships in which Rosen's (1973) aulopiforms and myctophiforms are united in the order Myctoph- iformes. In 1985, Rosen altered his concept of a monophy- letic Aulopiformes, noting that Aulopus shares several derived features with ctenosquamates, most notably a median rostral cartilage. Hartel and Stiassny (1986) considered a true median rostral cartilage a character of acanthomorphs and concluded that the morphol- ogy of the rostral cartilage is highly variable below that level. Nevertheless, Stiassny (1986) supported Rosen's (1985) view of a paraphyletic Aulopiformes, proposing that Chlorophthalmus, Parasudis and Aulopus form the sister group of ctenosquamates based on an elevated, reoriented cranial condyle on the maxilla and concurrent exposure of a "maxillary saddle" for reception of the palatine prong. G.D. Johnson (1992) discussed the shortcomings of Rosen's (1985) analysis and observed that neither Rosen nor Stiassny (1986) mentioned the distinctive gill-arch configuration origi- nally described by Rosen (1973) as unique to aulopi- forms. He added an additional gill-arch character to Rosen's (1973) complex, the absence of a cartilaginous condyle on PB3 for articulation of EB2, and concluded that a suite of gill-arch modifications constitutes a complex specialization supporting the monophyly of Rosen's (1973) Aulopiformes. In addition, Johnson (1992) offered further evidence (absence of the fifth upper pharyngeal toothplate and associated third in- ternal levator muscle) for the monophyly of Rosen's (1973) Ctenosquamata, which include myctophids, neoscopelids and acanthomorphs, but not aulopi- 1NTERRELATIONSHIPS OF FISHES 355 Copyright ? 1996 by Academic Press, Inc. All rights of reproduction in any form reserved. 356 CAROLE C. BALDWIN AND G. DAVID JOHNSON Gosline et al. (1966): Order Inioml Myctophoidea Aulopidae Bathysauridae Synodontidae Harpadontldae Bathypteroidae Ipnopidae Chlorophthalmidae Notosudidae (=Scopelosauridae) Myctophidae Neoscopelidae Alepisauroidea Paralepldidae Omosudldae Alepisauridae Anotopteridae Evermannellidae Scopelarchidae Rosen (1973): Order Auloplformes, new name Suborder Aulopoidei, new name Aulopidae Bathysauridae Bathypteroidae Ipnopidae Chlorophthalmidae Notosudidae (=Scopelosauridae) Suborder Alepisauroidei [15 fossil genera] Superfamily Synodontoidea, new usage [2 fossil general Synodontidae Harpadontldae Giganturidae (? + Rosauridae) Superfamily Alepisauroidea Paralepldidae Omosudldae Alepisauridae Anotopteridae Evermannellidae Scopelarchidae Sulak (1977): Benthic Myctophiformes: Aulopidae Aulopus (including Hime, Latropiscus) Synodontidae Subfamily Harpadontlnae Harpadon (incl. Peltharpadoril Saurida Subfamily Bathysaurinae Bathysawus (incl. Macrtstium) Subfamily Synodontinae Syrwdus (incl. Xystodus) Trachinocephalus Chlorophthalmidae Subfamily Chlorophthalminae Chlnrophthalmus Parasudis Bathysauropsts (incl. Bathysaurops) Subfamily Ipnoplnae Tribe Ipnopini Ipnops (incl. Ipnoceps) Tribe Bathypteroinl Bathypterois (incl. Benthosaurus) Tribe Bathymicropini Bathymlcrops Bathytyphlops (incl. Macristiella) R. K. Johnson (1982): Myctophiformes: Aulopoids Aulopidae Myctopholds + Chlorophthalmoids Myctophoids Myctophidae Neoscopelidae Chlorophthalmoids Notosudidae Scopelarchidae Chlorophthalmidae Ipnopidae Synodontoids + Alepisaurolds Synodontoids Synodontidae Harpadontldae Bathysauridae Alepisaurolds Paralepldidae Anotopteridae Evermannellidae Omosudldae Alepisauridae FIGURE 1 Four previously hypothesized classifications of aulopiform or myctophiform fishes. forms (see also Stiassny, this volume). Johnson et al. (1996) argued that Aulopus is not closely related to ctenosquamates but is the cladistically primitive mem- ber of their Synodontoidei, a lineage that also includes Pseudotrichonotus, Synodus, Trachinocephalus, Saurida, and Harpadon. Finally, Patterson and Johnson (1995) provided corroborative evidence from the intermuscu- lar bones and ligaments for Rosen's (1973) Aulopl- formes, in the extension of the epipleural series anteri- orly to the first or second vertebra. Various schemes of relationships among iniomous fishes have accompanied confusion about the recogni- tion of a monophyletic Auloplformes (Fig. 1). Gosline et al. (1966) recognized two "suborders": myctophoids (Aulopidae, Bathysauridae, Synodontidae, Harpa- dontldae, Bathypteroidae, Ipnopidae, Chlorophthal- midae, Notosudidae [= scopelosaurids of Marshall, 1966?see Paxton, 1972; Bertelsen et al, 1976], Myc- tophidae, and Neoscopelidae); and alepisaurolds (Par- alepldidae, Omosudldae, Alepisauridae, Anotopteri- dae, Evermannellidae, and Scopelarchidae). Rosen (1973) added synodontids and harpadontids (his syn- odontoids) and 17 fossil genera to the Alepisauroidei, described a new suborder, the Aulopoidei, for Aulopi- dae, Bathysauridae, Bathypteroidae, Ipnopidae, Chlorophthalmidae, and Notosudidae and, as noted, 14. Interrelationships of Aulopiformes 357 restricted the Myctophiformes to myctophids and neoscopelids. Sulak (1977) examined aspects of the osteology of the benthic "myctophiforms" and envisioned them forming two divergent lineages exhibiting progres- sively greater differentiation from the basal aulopid body plan, an expanded Synodontidae that included bathysaurids, synodontids, and harpadontids, and an expanded Chlorophthalmidae for chlorophthalmids (including Bathysauropsis) and ipnopids (including bathypteroids). To examine a previously proposed relationship be- tween the Evermannellidae and Scopelarchidae (e.g., Marshall, 1955; Gosline et al., 1966), R. K. Johnson (1982) studied the distribution of selected characters among iniomes. He did not present a formal classifi- cation but described three perceived iniomous clades. One comprised only aulopids, a second was equiva- lent to Rosen's (1973) alepisauroids minus scopelar- chids, and the third included myctophids, neoscopel- ids, chlorophthalmids, ipnopids, notosudids, and scopelarchids. R. K. Johnson's (1982) phylogeny cor- roborated Sulak's (1977) placement of bathysaurids in the synodontid + harpadontid lineage, but he noted that only two clades resulting from his analysis, the myctophoids (Myctophidae, and Neoscopelidae) and the alepisauroids (Paralepididae, Anotopteridae, Evermannellidae, Omosudidae, and Alepisauridae) were well supported. Okiyama (1984b) examined R. K. Johnson's (1982) hypothesis in light of evidence from aulopiform lar- vae. He did not produce an independent hypothesis of relationships but noted that his data offer little sup- port for a notosudid + scopelarchid + chlorophthal- mid + ipnopid lineage; rather, in his similarity matrix, scopelarchids share the most derived features (two) with evermannellids. Larval morphology also does not support a close association between bathysaurids and the synodontid + harpadontid lineage, but, as Okiyama (1984a) noted, Bathysaurus larvae are highly specialized. To demonstrate the potential systematic value of the intermuscular ligaments and bones in teleostean fishes, Patterson and Johnson (1995) investigated au- lopiform interrelationships based on this skeletal sys- tem. Their data provided support for a monophyletic Synodontoidei (sensu Johnson et ah, 1996) and a sister- group relationship between evermannellids and scopelarchids. Novel relationships depicted in their strict consensus of 24 equally parsimonious trees in- clude the following: a clade comprising all aulopiform taxa except ipnopids (represented by Bathypterois in their analysis) and Parasudis; sister-group relation- ships between Chlorophthalmus and synodontoids, notosudids and the evermannellid-scopelarchid lineage, and bathysaurids and giganturids; and a paraphyletic Paralepididae, with Paralepis forming the sister group of a monophyletic clade comprising Omo- sudis and Alepisaurus. Patterson and Johnson (1995) noted that the paraphyly of the Paralepididae sug- gested by their data may be artificial, a result of the greatly reduced number of intermuscular elements in Macroparalepis. No other comprehensive studies of aulopiform rela- tionships have been undertaken, and thus consider- able conflict about the evolutionary history of aulopi- form fishes existed when we initiated this study, the goal of which was to hypothesize a phylogeny of ex- tant aulopiform genera based on cladistic analysis of a wide range of morphological data. R. K. Johnsons's (1982) cladistic analysis of iniome relationships used commonality rather than outgroup comparison to as- sess character polarity, and we thus found that many of his polarity decisions were reversed in our analysis. Patterson and Johnson's (1995) phylogeny is of limited value because it was constructed on the basis of a single complex. Despite their shortcomings these pub- lications, as well as those of Rosen (1973), Sulak (1977) and Okiyama (1984b), proved useful in this study, and we derived many informative characters from them. II. Methods Osteological abbreviations are listed in Appendix 1, and a full list of materials examined is given in Appendix 2. Terminology for bones of the pelvic gir- dle follows Stiassny and Moore (1992), and that for the intermuscular bones and ligaments follows Patterson and Johnson (1995). In all line drawings, scale bars represent 1 mm, and open circles indicate cartilage. A. Data Analysis Character data were analyzed using heuristic meth- ods in Swofford's (1991) PAUP Version 3.0, and char- acter distributions were explored using MacClade Version 3.04 of Maddison and Maddison (1992). Ctenosquamates, represented by the cladistically primitive Myctophidae, Neoscopelidae, Metavelifer, and Polymixia (Stiassny, 1986; G. D. Johnson, 1992; Johnson and Patterson, 1993), were considered the first outgroup, and stomiiforms, represented by the cladistically primitive Diplophos (Fink and Weitzman, 1982), the second. The analysis included all aulopi- form genera except the notosudid Luciosudis; the re- cently described ipnopid, Discoverichthys (Merrett and 358 CAROLE C. BALDWIN AND G. DAVID JOHNSON Nielsen, 1987); and the paralepidids Dolichosudis, Mag- nisudis, and Notolepis. All characters were weighted equally, and all multi- state characters were treated as unordered unless oth- erwise noted. Steps in the transformation of a single character are denoted by subscripts following the character number (e.g., 12 is state 2 of character 1). Many characters have more than one equally parsimo- nious reconstruction, and we optimized ambiguous characters on the tree using ACCTRAN, a method that favors reversals over parallel acquisitions when the choice is equally parsimonious (Farris, 1970; Swof- ford and Maddison, 1987). Ambiguous character states resolved using ACCTRAN are denoted in Dis- cussion (Section VI) with an asterisk, e.g., (34^). Character data also were analyzed using Hennig86 (Farris, 1988) and the results exported to Clados Ver- sion 1.2 (Nixon, 1992) for construction of a tree on which characters and states are indicated (Fig. 6). There are some discrepancies in the distribution of character states between PAUP-MacClade and Hen- nig86-Clados, primarily because (1) for ambiguous characters optimized with e.g., ACCTRAN, Mac- Clade recognizes that ambiguity may still exist at cer- tain nodes, whereas Clados forces a resolution at all nodes; and (2) PAUP-MacClade allows polymor- phisms in terminal taxa, whereas Hennig86-Clados does not. Character states on the tree (Fig. 6) that appear as synapomorphies in Clados but not Mac- Clade are marked with a large dot; they are not dis- cussed in the text, which is based on the PAUP- MacClade results. B. Taxonomy Parin and Kotlyar (1989) resurrected the aulopid genus Hime Starks (type species A. japonicus Gunther) for Pacific aulopids based on a difference in the length of the dorsal-fin base between Atlantic and Pacific species but used length of the anal-fin base as a taxo- nomic feature within Hime. We find the evidence for generic distinction unconvincing and thus follow Mead (1966a) in recognizing a single genus, Aulopus, for all aulopid species. We place Harpadon and Saurida in the Synodontidae as did Sulak (1977), Omosudis in the Alepisauridae, and Anotopterus in the Paralepididae (see Discussion). "Scopelarchoides" herein refers to S. signifer which, ac- cording to R. K. Johnson (1974a), may be an incorrect generic assignment for that species. He hypothesized that S. nicholsi (the type species of Scopelarchoides) and S. danae are more closely related to Scopelarchus than to other species of Scopelarchoides but retained Scopelar- choides for S. signifer pending further investigation. Early in our study it became apparent that Bathy- sauropsis gigas (Kamohara) is not closely related to B. gracilis Regan and B. malayanus (Fowler). Bathysaurop- sis gracilis is the type species of Bathysauropsis Regan, and thus all reference to Bathysauropsis is to B. gracilis and B. malayanus. A new genus, Bathysauroides, is erected for Bathysauropsis gigas. III. Bathysauroides Gen. Nov. Diagnosis?An aulopiform distinguished from all other genera by the following combination of charac- ters: a low number of caudal vertebrae (5-7, or ca. 11-15% of total vertebrae in Bathysauroides gigas), slightly elliptical eyes with an anterior aphakic space and gill rakers present as toothplates. Type species?Bathysaurops gigas Kamohara 1952. Etymology?From the Greek bathys, deep, and sauros, lizard, in reference to the deep habitat and superficial resemblance to lizardfishes. Gender?Masculine. Justification?Our hypothesis of cladistic relation- ships among aulopiform genera (Fig. 2) is best re- flected by removing Bathysauropsis gigas from Bathy- sauropsis Regan and placing it in a distinct genus. In addition to the diagnostic characters listed above, Bathysauroides gigas can be distinguished from its for- mer congeners based on the following features identi- fied in this study or taken from the original description of Bathysaurops gigas (Kamohara, 1952): palatine with more prominent teeth than premaxilla; epipleurals ex- tending anteriorly to the 1st vertebra (vs 2nd); epineu- rals on about the 3rd through 17th vertebrae originat- ing on centrum (vs neural arch); 16-17 pectoral-fin rays (vs 22-24); basihyal with two rows of large teeth (vs no basihyal teeth); pectoral fin extending to vertical through middle of dorsal-fin base (vs beyond base of dorsal); anus much closer to pelvic fins than to anal fin (vs closer to anal fin); and adipose fin inserting above anterior part of anal-fin base (vs well behind anal base) IV. Monophyly of Aulopiformes We agree with Rosen (1973) that a lateral displace- ment of the proximal end of PB2 and a concomitant elongation of the uncinate process on EB2 to bridge the large gap between EB2 and PB3 are derived for aulopiforms (Character 1, Fig. 3). We disagree with R. K. Johnson's (1982) assessment of an elongate EB2 uncinate process as a primitive iniome condition be- cause the first and second outgroups for iniomes are 14. Interrelationships of Aulopiformes 359 Diplophos Myctophidae Neoscopelus Metavelifer Polymixia Aulopus Pseudotrichonotus Synodus Trachinocephalus Harpadon Saurida Bathypterois - Bathymicrops Bathytyphlops Ipnops Scopelosaurus Ahliesaurus Bathysauropsis Chlorophthalmus Parasudis - Omosudis - Alepisaurus Paralepis Arctozenus Lestrolepis Stemonosudis Lestidiops Lestidium Uncisudis Macroparalepis Sudis Anotopterus Coccorella Odontostomops Evermannella Scopelarchus Scopelarchoides Benthalbella Rosenblattichthys-1 Bathysauroides Bathysaurus Gigantura SYNODONTOIDEI CHLOROPHTHALMOIDEI ALEPISAUROIDEI ] GIGANTUROIDEI FIGURE 2 Proposed phylogenetic relationships among aulopiform genera based on strict consensus of nine equally parsimonious trees (length = 364, CI = 0.55, RI = 0.80). acanthomorphs and stomiiforms, neither of which has an EB2 uncinate process. We also disagree with his interpretation of the unbranched anterior portion of the EB2 of Neoscopelus as an elongate uncinate process. There is nothing in the EB2 morphology of Neoscopelus (Rosen, 1973, fig. 71) to suggest that it is configured differently from that of myctophids and stomiiforms?that is, the cartilaginous tip is some- what expanded such that it articulates with both PB2 and PB3 (Rosen, 1973, figs. 18-22 and 69-70). Fur- thermore, like those two groups, the EB2 of Neosco- pelus articulates with a cartilaginous condyle on PB3, the absence of which is another aulopiform synapo- morphy (Character 2; Johnson, 1992). Rosen (1973, figs. 14-16) questioned an aulopiform affinity for paralepidids because of (1) the primitive, salmoniform-like appearance of the dorsal gill arches of juvenile Paralepis speciosa and (2) the absence in adult Paralepis and Lestrolepis of the long EB2 uncinate process and laterally displaced PB2 characteristic of other aulopiforms. The first is invalid because Rosen's (1973, fig. 16) "juvenile Paralepis" is not a paralepidid. We examined the specimen upon which his descrip- tion and illustration were based (AMNH 17232) and 360 CAROLE C. BALDWIN AND G. DAVID JOHNSON PB3 FIGURE 3 Ventral view of dorsal gill arches from left side of (A) Chlorophthalmus atlanticus, USNM 339774 and (B) Synodus variegatus, USNM 339776. concluded on the basis of meristic and other features that it is an argentinoid, probably Bathylagus. Several features characteristic of bathylagid (and not aulopi- form) gill arches are evident in Rosen's fig. 16: PB2 is broad anteriorly rather than tapered, UP4 is absent, UPS (labelled UP4 or UP5 by Rosen) is extremely re- duced to a small ovoid plate, and there is a long levator process on EB4. Note also that the muscle labelled "RAB" by Rosen inserts on EB4 rather than the pha- ryngobranchials, indicating that it is the oesophageal sphincter, not the retractor dorsalis. Rosen's (1973) "juvenile Paralepis" also has an uncinate process on PB3 for articulation with the uncinate process of EB2, a primitive teleostean feature lacking in aulopiforms (Johnson, 1992). Rosen's (1973) second claim, that paralepidids lack a laterally displaced PB2 and concomitant elongation of the EB2 uncinate process is not true of Paralepis, 14. Interrelationships of Aulopiformes 361 Arctozenus, Anotopterus, or Sudis. In those taxa, as in other aulopiforms, the uncinate process of EB2 (which is cartilaginous in Paralepis and Arctozenus) spans the gap between PB3 and the posterolaterally displaced PB2. In other paralepidids examined (Macroparalepis, Uncisudis, Lestidium, Lestidiops, Stemonosudis, and Les- trolepis), the uncinate process of EB2 is parallel and closely applied to the main arm of EB2, which un- doubtedly explains why Rosen overlooked it. The con- figuration of the dorsal gill arches of those paralepid- ids involves several diagnostic modifications that we discuss in more detail in a later section (see charac- ter 9). Additional evidence corroborating the monophyly of Rosen's (1973) Aulopiformes is found in the pattern of the intermuscular bones (Patterson and Johnson, 1995). The group is uniquely characterized by having attached epipleural bones extending forward to at least the second, and frequently the first, vertebra (character 54). Epipleurals are most commonly re- stricted to midbody as they are in stomiiforms, myc- tophiforms, and Polymixia (the only acanthomorph with epipleural bones). Our analysis also indicates that another feature of the intermusculars, the dis- placement of one or more of the anterior epipleurals dorsally into the horizontal septum (character 55), a feature used by Patterson and Johnson (1995) to indi- cate relationships within the Aulopiformes, is best interpreted as a synapomorphy of the group. Another aulopif orm character is their lack of a swim- bladder (character 112; see Marshall, 1954, 1960; Mar- shall and Staiger, 1975). Many deep-sea fishes lack a swimbladder, but the presence of a swimbladder primi- tively in stomiiforms (including Diplophos) and ctenos- quamates (most myctophids and neoscopelids, 1am- pridiforms, and polymixiids)?see Marshall (1960), Woods and Sonoda (1973)?suggests that loss of the swimbladder in aulopiforms is independent of losses in other teleosts. R. K. Johnson (1982) hypothesized three losses of the swimbladder among iniomes: in au- lopids, in the chlorophthalmoid lineage of his mycto- phoid + chlorophthalmoid clade, and in the ancestor of his alepisauroid + synodontoid lineage. Rosen's (1973) hypothesis of a monophyletic Aulopiformes requires a single loss in the ancestral aulopif orm. We agree with R. K. Johnson's suggestion that peritoneal pigment in larvae may be diagnostic of Rosen's (1973) aulopiforms (character 116). Larvae of Diplophos, myctophiforms, and primitive acantho- morphs lack peritoneal pigment, as do several aulopi- forms (notosudids, some ipnopids, and the scopelar- chid Benthalbella), presumably secondarily. Larvae of Bathysauropsis and Bathysauroides are unknown. Finally, we have found new evidence for aulopi- form monophyly in the morphology of the pelvic gir- dle. Primitively in euteleosts, the pelvic plates often approach one another or abut medially in the region of the medial processes (Stiassny and Moore, 1992), as in Diplophos and myctophiforms (Fig. 4A), but the medial processes are never fused. Uniquely in aulopi- forms, the medial processes of the pelvic girdle are long broad plates that are joined medially by cartilage (character 87, Figs. 4B-4D, and 5). Stiassny (1986) rejected Rosen's concept of Aulopi- formes, arguing that three genera of that group (Aulo- pus, Chlorophthalmus, and Parasudis) form the sister group of ctenosquamates based on a particular type of association between the maxilla and the palatine (her fig. 5). This single feature (character 44) does not outweigh the branchial, intermuscular, swimbladder, larval pigmentation, and pelvic girdle evidence that unites aulopiforms. Furthermore, placement by John- son et al. (1996) of the Aulopidae as the sister-group of other synodontoids and our placement of the Chlorophthalmidae as the sister group of other chloro- phthalmoids are in direct conflict with Stiassny's (1986) hypothesis. V. Character Analysis Our hypothesis of the relationships among aulopi- form genera (Fig. 2) was derived from the data matrix in Table 1. The tree represents a strict consensus of nine fully resolved trees (each 364 steps in length, CI=0.55, RI=0.80 in the PAUP analysis), all of the ambiguity occurring within the Paralepididae and Scopelarchidae. The Hennig86 analysis yielded the same trees, although there were small differences in tree statistics. Based on our analysis, we divide aulopif orm genera into four clades: Synodontoidei (Aulopidae, Pseudo- trichonotidae, and Synodontidae), Chlorophthal- moidei (Chlorophthalmidae, Bathysauropsis, Notosud- idae, and Ipnopidae), Alepisauroidei (Alepisauridae, Paralepididae, Evermannellidae, and Scopelarchi- dae), and Giganturoidei {Bathysauroides, Bathysauri- dae, and Giganturidae). In the following comparison of phylogenetically informative characters among aulopiforms, character numbers refer to those in the matrix (Table 1) and on the Glados-derived tree (Fig. 6). A. Gill Arches 1. Second Epibranchial Uncinate Process (Fig. 3)?As discussed above (in Monophyly of Aulopiformes) the presence of an uncinate process on EB2 articulating with PB3 characterizes all aulopiforms except Bathy- pterois and some paralepidids. In Bathypterois (Fig. 7B), 362 CAROLE C. BALDWIN AND G. DAVID JOHNSON LPP LPP FIGURE 4 Ventral view of pelvic girdle of (A) Myctophum obtusirostre, AMNH 29140SW, (B) Chlorophthal- mus agassizi, USNM 159385, (C) Bathypterois pectinatus, FMNH 88982, and (D) Scopelosaurus hoedti, USNM 264256. the EB2 uncinate process falls well short of PB3, but PB2 is posterolaterally displaced as it is in other aulopi- forms. In certain paralepidids (Macroparalepis, Unci- sudis, Lestidium, Lestidiops, Stemonosudis, and Lestrolepis [Fig. 8B]), PB2 is reoriented and the resulting configu- ration of EB2 and its uncinate process is very different from that of other aulopiforms. We describe this con- dition more fully in Character 9 below and, to avoid duplicating what we interpret as a unique specializa- tion of paralepidids, we do not assign a different state to that condition here. Other features clearly place Bathypterois and all paralepidids deep within the Au- lopiformes, and thus the variation in the EB2 uncinate process in those taxa is derived relative to the primi- tive aulopiform condition. (10) = EB2 uncinate process absent (lj) = EB2 uncinate process present and enlarged; 14. Interrelationships of Aulopiformes 363 MPP FIGURE 5 Ventral view of pelvic girdle of (A) Pseudotrichonotus altivelis ZUMT 59882 (redrawn from Johnson et al., 1996), (B) Scopelarchoides signifer, USNM 274385, (C) Evermannella indica, USNM 235141, and (D) Lestrolepis intermedia, USNM 290253. The dorsally projecting autogenous pelvic cartilages in Evermannella are not shown because they are obscured by the pelvic girdle; in Lestrolepis, these cartilages have been manually displaced from their dorsally directed orientation for illustration. PB2 displaced posterolaterally (except in some paralepidids) (12) = EB2 uncinate process present, but not en- larged; PB2 displaced posterolaterally 2. Cartilaginous Condyle on Dorsal Surface of Third Pharyngobranchial?Aulopiforms lack a condyle on PB3 articulating with EB2 (Johnson, 1992). This con- dyle is a primitive euteleostean condition and is pres- 364 CAROLE C. BALDWIN AND G. DAVID JOHNSON & u > S ? a S ? a o u ^ S 11 ?f 73 JA A O >, s ft 8" 3 a. Q S o n g o g fi fi o cd CD n 01 50% head length 48. Premaxillary Fenestra (Fig. 13)?R. K. Johnson (1982) considered the presence of a fenestrated pre- maxilla as a synapomorphy of paralepidids and Ano- topterus, noting that although Rosen (1973) viewed the feature as diagnostic of his alepisauroids, it does not appear to be present in alepisaurids, evermannellids, and other iniome fishes. We concur and note that all paralepidids examined have at least a partial fenestra in the anterior end of the premaxilla. The fenestra is usually complete in larger specimens. (480) = No premaxillary fenestra (48j) = Anterior premaxilla with fenestra 49. Palatine Articulates with Premaxilla (Fig. 13)?In paralepidids, the palatine terminates anteriorly in a long process that articulates by connective tissue with the medial surface of the premaxilla, just posterior to the premaxillary fenestra. Various associations be- tween the palatine and maxilla (when present) exist among aulopiforms and outgroups, but the palatine typically terminates near the point where it articulates with the maxilla and does not extend anteriorly to meet the premaxilla. The palatine also articulates with the premaxilla in Harpadon, but this condition differs from that in paralepidids in that the premaxilla replaces the maxilla as the site of articulation with the palatine and eth- moid because the maxilla is present only as a small remnant. (490) = Palatine without process for articulation with premaxilla (49x) = Palatine with long process for articulation with premaxilla 50. Lacrimal Oriented Horizontally on Snout?The lacrimal typically borders the orbit anteriorly. In para- lepidids, the second infraorbital is in the position nor- mally occupied by the lacrimal, and the lacrimal is located horizontally on the snout, well rostral to the orbit. Our identification of this bone as the lacrimal rather than the antorbital is based on its relatively large size and association with the upper jaw; as is typical of the lacrimal, this bone extends along a por- tion of the upper border of the maxilla. In our small specimens of Sudis, the position of the lacrimal cannot be determined. (500) = Lacrimal bordering orbit anteriorly (50j) = Lacrimal anterior to orbit, oriented horizontally. 51. Maxilla Reduced?Johnson et al. (1996) dis- cussed the problems associated with Sulak's (1977) interpretation of a reduced maxilla in Harpadon and Saurida, especially his assessment of a division of the maxilla into anterior and posterior elements. They concluded that a reduced maxilla is a synapomorphy of synodontids and harpadontids and suggested that the reduced maxilla in Bathysaurus, which exists as a remnant anteriorly, is not homologous with the poste- rior remnant in Harpadon. The condition in Bathy- saurus, however, may be homologous with the poste- rior maxillary remnant of Gigantura, as our phylogeny supports Patterson and Johnson's (1995) hypothesis of a sister-group relationship between Bathysaurus and Gigantura (see character 117 below and Discussion); nevertheless, we conservatively consider anterior and posterior maxillary remnants as distinct. The maxilla of most alepisauroids is slender relative to that of other aulopiforms, but usually retains the same shape in that the posterior end is expanded. We do not recognize the alepisauroid condition as a separate state except in Anotopterus, in which the max- illa is a very slender, strut-like bone with no poste- rior expansion. (510) = Maxilla well developed with posterior end expanded (51a) = Maxilla intact but slender, posterior end not expanded (512) = Maxilla present as posterior remnant (513) = Maxilla present as anterior remnant D. Cranium We did not examine the cranial morphology of au- lopiforms in detail but describe certain aspects of the ipnopid cranium below. 52. Frontal Expanded Laterally over Orbit?The skull of ipnopids is dorsoventrally compressed, and the frontals extend laterally over the eyes in Bathypterois, 382 CAROLE C. BALDWIN AND G. DAVID JOHNSON Bathymicrops, and Bathytyphlops and beneath the large photosensitive plates in Ipnops. In other aulopiforms and outgroups, the frontals lie completely between the orbits. (520) = Frontal not expanded laterally (52j) = Frontal expanded laterally 53. Sphenotic Process?The sphenotic is modified in all ipnopids such that a process of the bone extends anteriorly beneath the greatly expanded frontal. In its most extreme form, the process extends forward to the lateral ethmoid, as in Bathytyphlops. It extends about halfway to the lateral ethmoid in Bathymicrops, and only a little forward in Ipnops. It is least developed but present as a small rounded extension beneath the frontal in Bathypterois. In other aulopiforms and outgroups, the sphenotic abuts the frontal but has no anteriorly directed process. (530) = Sphenotic without anterior process (53x) = Sphenotic with anterior process E. Intermuscular Bones and Ligaments Unless otherwise noted, all characters described in this section are from Patterson and Johnson (1995). Our survey of aulopiform intermusculars is more ex- tensive than that of Patterson and Johnson, but it is still incomplete (Table 1). Further investigation is needed. 54. Epipleurals Extend Anteriorly to First or Second Vertebra?In synodontoids and chlorophthalmoids, epipleurals extend anteriorly to the second vertebra (V2); in alepisauroids and giganturoids, they extend to VI. When present, epipleurals begin on V3 in the outgroups. Metavelifer and all acanthomorphs except Polymixia lack epipleurals, a condition we code as a separate state for this character but as "missing data" for other epipleural characters in this section to avoid erroneously inflating tree length and modifying other tree statistics. Patterson and Johnson (1995) hypothe- sized that the anterior extension of epipleurals to V2 is a synapomorphy of Aulopiformes, but their consen- sus tree indicates that the extension of epipleurals to VI (from V2) evolved independently in the ancestor of the Evermannellidae + Scopelarchidae and the re- maining alepisauroids. Our phylogeny suggests that the extension of epipleurals to VI occurred once, in the ancestor of alepisauroids + giganturoids. (540) = Epipleurals originate on V3 (54x) = Epipleurals originate on V2 (542) = Epipleurals originate on VI (543) = Epipleurals absent 55. One or More Epipleurals Displaced Dorsally Into Horizontal Septum?The presence of one or more ante- rior epipleurals displaced dorsally into the horizontal septum is a synapomorphy of aulopiforms. Aulopus and Gigantura have a single epipleural in the horizon- tal septum, but most other aulopiforms have more than one displaced. Bathypterois, Bathymicrops, and Parasudis have all epipleurals beneath the horizontal septum, and those taxa fall outside of a clade compris- ing the remaining aulopiforms in Patterson and John- son's (1995) tree constructed solely on the basis of intermuscular characters. It is more parsimonious to interpret the absence of dorsally displaced epipleurals in Bathypterois, Bathymicrops, and Parasudis as rever- sals. Our small cleared and stained specimen of Ipnops lacks ossified epipleurals anteriorly, and we were not able to conclusively identify ligamentous epipleurals anteriorly in that specimen. (550) = All epipleurals beneath the horizontal septum (55j) = One or more epipleurals displaced dorsally into horizontal septum 56. Abrupt Transition of Epipleurals in and beneath the Horizontal Septum?Two states of the derived condi- tion of dorsally displaced anterior epipleurals occur in aulopiforms. In one, the transition between epipleurals in and beneath the horizontal septum is abrupt, such that the last posterolaterally directed epipleural in the horizontal septum is followed imme- diately by a ventrolaterally directed epipleural that is completely below the horizontal septum. This occurs only in Pseudotrichonotus, Synodus, and Trachinocepha- lus. In other aulopiforms, including Harpadon and Saurida, the transition is gradual, occurring over a series of vertebrae. We initially coded this character as having three states (no epipleurals in the horizontal septum, an abrupt transition of epipleurals in and beneath the horizontal septum, and a gradual transition of those epipleurals) to determine the primitive aulopiform state. Parsimony indicates that a gradual transition is primitive for aulopiforms and a synapomorphy of the order. However, it does not seem valid to consider both the presence of one or more dorsally displaced epipleurals (55x) and a gradual transition of epipleu- rals in and beneath the horizontal septum as synapo- morphies of aulopiforms because the latter is a state of the former. Accordingly, for this character, we group the most common outgroup condition (no dor- sally displaced epipleurals) and the primitive ingroup condition (gradual transition of epipleurals) as a single state. This allows us to recognize the abrupt transition of epipleurals in some synodontoids as a derived fea- 14. Interrelationships of Aulopiformes 383 ture without creating two synapomorphies of aulopi- forms where only one is warranted. (560) = No epipleurals displaced dorsally into the horizontal septum or the transition between epipleurals in and beneath the horizontal septum is gradual. (56^ = Abrupt transition between epipleurals in and beneath the horizontal septum 57. One or More Epipleurals Forked Distally?In the region where the epipleurals leave the horizontal sep- tum in notosudids, around V19 (Ahliesaurus) or V20- V24 (Scopelosaurus), the epipleurals are bifurcate dis- tally. In other aulopiforms and outgroups, epipleurals are not forked distally at the transition in and beneath the horizontal septum (or none are in the septum). (570) = Epipleurals not forked distally (571) = Epipleurals forked distally at transition of epipleurals in and beneath the horizontal septum 58. Epipleural on First and Second Vertebrae Fused to Centrum?In Omosudis, Alepisaurus, and Paralepis, the epipleurals on VI and V2 are fused to the centrum. Those epipleurals are free in other paralepidids, au- lopiforms, and outgroups. Fusion of the epipleurals in Paralepis is independent of that in the alepisaurid lineage. (580) = Epipleurals on VI and V2 autogenous (58j) = Epipleurals on VI and V2 fused to centrum 59. Epipleurals Not Attached to Axial Skeleton?Most epipleurals are not attached to the axial skeleton in Omosudis, Paralepis, and Arctozenus, but most or all are attached in Alepisaurus, other aulopiforms, and outgroups. As with the epineurals (see character 63 below), Paralepis and Arctozenus have the anterior epipleurals forked anteriorly, and the branch that atta- ches the bone to the axial skeleton disappears posteri- orly leaving a large series of unattached epipleurals. (590) = Most or all epipleurals attached to axial skeleton (59x) = Most epipleurals not attached to axial skeleton (592) = Most epipleurals are free dorsal branches 60. Reduced Number of Epipleurals?Most aulopi- forms have a long series of epipleurals that begin on VI or V2. Most outgroups also have a well-developed series of epipleurals, although they begin more poste- riorly than in aulopiforms (see character 54). In the paralepidids Lestrolepis, Macroparalepis, and Sudis, the epipleural series is confined to the first five or fewer vertebrae. Epipleurals are not evident in our small specimens of Uncisudis, Stemonosudis, and Lestidiops. (600) = Long series of epipleurals (60%) = Epipleural series not extending posteriorly beyond V5 61. Origin of Epineurals?In Scopelosaurus and Ahlie- saurus, anterior epineurals originate on the neural arch. The origin of subsequent epineurals descends to the centrum or parapophysis, and then it reascends in posterior epineurals to the neural arch. A similar configuration of ventrally displaced epineurals occurs in evermannellids, scopelarchids, and Bathysauroides. In those taxa, the origin of epineurals always returns to the neural arch posteriorly, and usually less than half of the epineurals originate on the centrum. In Bathysaurus, the first five epineurals originate on the neural arch, and the origin of the rest descends to the centrum. In Gigantura, all epineurals originate on the centrum. In other aulopiforms, Diplophos, myctophi- forms, and Metavelifer, all epineurals originate on the neural arch. In Polymixia, epineurals on V3-10 origi- nate on the centrum, those more anterior and poste- rior originate on the neural arch or spine. The origin of some of the central epineurals on the centrum (with reascension posteriorly) and the origin of most or all epineurals on the centrum (without reascension pos- teriorly) are derived conditions within the Aulopi- formes. (610) = All epineurals originate on neural arch (61x) = Some epineurals originate on the centrum or parapophysis; these flanked anteriorly and posteriorly by epineurals originating on the neu- ral arch (612) = Most or all epineurals originate on centrum; epineurals not reascending to neural arch poste- riorly 62. First One to Three Epineurals with Distal End Dis- placed Ventrally?In some aulopiforms, the first one to three epineurals are turned downward such that they extend lower than their successors. The distal end of the epineurals on V1-V3 is so modified in Aulopus, Pseudotrichonotus (V1-V2), Synodus (V1-V2), Trachinocephalus (VI), Chlorophthalmus (V1-V2), Ipnops (VI), and Bathysauroides (V1-V2). Patterson and John- son (1995) suggested that Chlorophthalmus may be the sister group of synodontoids based on this feature, but our analysis indicates that having the distal end of the anteriormost one or more epineurals turned downward evolved independently in Chlorophthal- mus, Bathysauroides, and the ancestral synodontoid. The condition is reversed in Harpadon and Saurida. (620) = Distal end of epineurals not displaced ven- trally 384 CAROLE C. BALDWIN AND G. DAVID JOHNSON (62j) = Distal end of first one to three epineurals displaced ventrally 63. Some Epineurals and Epipleurals Forked Proxi- mally?Beginning on about V12 or V15, the epineurals in Chlorophthalmus and Parasudis are forked proxi- mally. Posteriorly, the dorsomedial branch, which at- taches the epineural to the axial skeleton, disappears, leaving a short series of unattached epineurals. A sim- ilar condition occurs primitively in myctophiforms. Patterson and Johnson (1995) did not identify forked epineurals in Paralepis, but our examination of addi- tional specimens of that genus indicates that the ante- riormost five or six epineurals are forked proximally, the dorsal branch disappearing posteriorly, leaving a long series of unattached epineurals (see character 65). A nearly identical pattern characterizes Arctozenus. A unique branching of the epineurals characterizes Gi- gantura (Patterson and Johnson, 1995). Proximal branching of epipleurals occurs in the same pattern as that of the epineurals among aulopi- forms, and we group the branching of the two series of bones as a single character. (630) = No epineurals (or epipleurals) forked proxi- mally (63j) = Epineurals (and epipleurals) from about V12-V15 to near end of series forked proximally (632) = Epineurals (and epipleurals) on about VI-V5 forked proximally (633) = "Gigantura" pattern of branching 64. Epineurals Fused to Neural Arch?Epineurals are fused to the neural arch on V1-V10 in Harpadon and Saurida. Epineurals typically are not fused to the axial skeleton in aulopiforms and outgroups, although they are fused to the neural arch on VI-V5 in Diplophos and Alepisaurus and on VI in Paralepis and Macroparalepis; most are fused to the centrum in Rosenblattichthys and Bathysaurus. Fusion of epineurals to the axial skeleton has thus evolved several times within aulopiforms, and our analysis suggests that this condition is phylo- genetically significant only as a synapomorphy of Har- padon and Saurida. (640) = Epineurals not fused to axial skeleton (64j) = Epineural fused to neural arch on VI (642) = Epineurals fused to neural arch on V1-V5 (643) = Epineurals fused to neural arch on V1-V10 (644) = Most epineurals fused to centrum 65. Epineurals Attached to Axial Skeleton?-In Alepi- saurus and Omosudis, most epineurals are not attached to the axial skeleton, and in Anotopterus, all are unat- tached. Paralepis and Arctozenus have the anterior epi- neurals forked and attached to the axial skeleton by the dorsal branch of the fork; on about V5 or V6, only the ventral branch remains, and the epineurals are thus unattached posteriorly. In other aulopiforms and outgroups, all or most epineurals are attached. (650) = Most or all epineurals attached to axial skeleton (65j) = Most epineurals unattached (652) = All epineurals unattached (653) = Unattached epineurals represent only free ventral branches of forked epineurals 66. Epicentrals?Paralepidids, Bathysaurus, and Gi- gantura lack epicentrals. Omosudis and Alepisaurus have them ossified and beginning on V3. Parasudis has all epicentrals ossified and beginning on VI. All other aulopiforms and outgroups have ligamentous epicentrals, except the anterior epicentrals are cartilag- inous in evermannellids (see next character), and the ligamentous epicentrals of Polymixia contain a cartilag- inous rod distally (Patterson and Johnson, 1995). It is equally parsimonious to consider ligamentous epicen- trals, ossified epicentrals, or no epicentrals as the an- cestral condition for the clade comprising alepisaurids and paralepidids, but it seems unlikely that ligamen- tous epicentrals transformed into ossified ones and then were lost or that ligamentous epicentrals were lost and then regained as ossified epicentrals. There is evidence from Parasudis that ligamentous epicentrals ossify and from giganturoids that ligamentous epicen- trals are lost (the ancestral chlorophthalmoid and gi- ganturoid intermuscular systems are characterized by ligamentous epicentrals), and thus we believe it most likely that the ossification of ligamentous epicentrals is a synapomorphy of alepisaurids, and loss of liga- mentous epicentrals is a derived feature of paralepi- dids. To reflect this, we partially ordered this character such that a single step is required to lose or ossify ligamentous epicentrals, but two steps are required to lose ossified epicentrals or gain ossified epicentrals when none existed ancestrally. Considering this char- acter entirely unordered does not change the phylog- eny but eliminates a synapomorphy of alepisaurids and one of paralepidids. (660) = Epicentrals ligamentous (66x) = Epicentrals ossified (662) = Epicentrals absent (663) = Epicentrals cartilaginous anteriorly, ligamen- tous posteriorly 67. Anterior Epicentrals Closely Applied to Distal End of Epipleurals?Evermannellids are unique among au- lopiforms in having the anterior epicentrals present as small rods of cartilage closely applied to the distal ends of the epipleurals. This is unusual because epi- 14. Interrelationships of Aulopiformes 385 centrals are almost always attached to the centrum or parapophyses. A similar condition occurs in scopelar- chids except that the anterior epicentrals are in lig- ament. (670) = All epicentrals attached to centrum or para- pophyses (67j) = Anterior epicentrals attached to distal end of epipleurals F. Postcranial Axial Skeleton 68. Number of Supraneurals?Presence of three su- praneurals preceding the dorsal fin is a synapomor- phy of eurypterygians (Johnson and Patterson, this volume), but many aulopiforms have two or fewer, and numerous reconstructions of the reductions are possible. We interpret a single supraneural as a syna- pomorphy of chlorophthalmoids with reversals in Chlorophthalmus, Bathysauropsis, and Bathytyphlops. Presence of two supraneurals is a synapomorphy of alepisauroids, with further reduction to one (Omo- sudis) or none (Alepisaurus) in alepisaurids (or in their common ancestor) and reversal to three in the ances- tral scopelarchid. (680) = Three or more supraneurals (68j) = Two supraneurals (682) = One supraneural (683) = No supraneurals 69. Number of Caudal Vertebrae?Aulopiforms, stomiiforms, and ctenosquamates primitively have about half (40-60%) of the vertebrae as caudal verte- brae. A reduction in the number of caudal vertebrae occurs independently in the synodontid-harpadontid clade (17-19%) and in giganturoids (11-24%). Scopel- archids and evermannelHds have 62-70% caudal ver- tebrae, a condition that we interpret as synapomor- phic for those families. A large number of caudal vertebrae occur independently in Chlorophthalmus (62%), Bathymicrops (68%), and Arctozenus (70%). It seems reasonable that both the very low and very high numbers of caudal vertebrae were derived from the primitive aulopiform condition of about 50% (coded as 69j), and thus we consider the three states to form an ordered transformation series (690 <-? 69x ** 692). (690) = < 25% caudal vertebrae (69J = 40-60% caudal vertebrae (692) = > 60% caudal vertebrae 70. Accessory Neural Arch?An accessory neural arch on VI is present in Diplophos, Aulopus, and synod- ontids. It is absent in all ctenosquamates, Pseudotricho- notus, chlorophthalmoids, alepisauroids, and gigant- uroids. Polarity of this character for aulopiforms is equivocal, but in our analysis, an accessory neural arch is a synapomorphy of synodontoids. (700) = Accessory neural arch absent (70j) = Accessory neural arch present 71. First Neural Arch with Brush-like Growth?There is a unique brush-like posterodorsal outgrowth of bone on the first neural arch of Synodus and Trachino- cephalus (Patterson and Johnson, 1995). (710) = No brush-like growth on first neural arch (71j) = Brush-like growth on first neural arch 72. Number of Open Neural Arches?In chlorophthal- moids (except Ipnops), alepisauroids, and Bathysaur- oides, the neural arch on VI and sometimes V2-V4 is open dorsally. In ctenosquamates, all neural arches are closed dorsally (see also Stiassny, this volume), whereas many are open in synodontoids, Bathysaurus, Gigantura, and Diplophos. The latter is the primitive aulopiform condition, and thus a reduced number of open neural arches is a synapomorphy of the Chloro- phthalmoidei + Alepisauroidei + Giganturoidei. Hav- ing many open neural arches is a reversal uniting giganturids and bathysaurids. (720) = Many neural arches open dorsally (72%) = Neural arches open on VI and sometimes V2-V4 (722) = All neural arches closed dorsally 73. Origin of First Rib?The origin of the first rib var- ies among aulopiforms from VI to V5. The first rib origi- nates on V3 primitively in aulopiforms, but its origin changes within all aulopiform suborders. Nearly 75 re- constructions of this character are possible in aulopi- forms, the only hypothesis of relationship common to all of them being that a more posterior origin (V4) of the first rib is a synapomorphy of Pseudotrichonotidae + Synodontidae, with the origin shifting to V5 in the ancestor of Synodus and Trachinocephalus and to V2 in Harpadon. Our analysis also suggests that the origin of the first rib moved anteriorly from V3 to V2 in the ances- tral ipnopid and from V3 to VI in the ancestor of the alepisaurid + paralepidid clade. (730) = First rib originates on V3 (73j) = First rib originates on V4 (732) = First rib originates on V5 (733) = First rib originates on V2 (734) = First rib originates on VI (735) = Ribs absent 74. Ossification of Ribs?In synodontoids, alepi- saurids, and paralepidids, all ribs ossify in membrane 386 CAROLE C. BALDWIN AND G. DAVID JOHNSON bone. In most scopelarchids, ribs are ligamentous, but in Scopelarchoides signifer, most ribs ossify in membrane bone, and only the last two are ligamentous. In all other aulopiforms except Bathymicrops and Gigantura, which lack ribs, at least some ribs ossify in membrane bone. In the outgroups, all ribs ossify in cartilage. Having any or all ribs ossify in membrane bone is derived for aulopiforms, but the distribution of the two states is such that the ancestral aulopiform condi- tion could be either. However, ossification of only some ribs in membrane bone is primitive for the clade comprising chlorophthalmoids, alepisauroids, and gi- ganturoids, and thus having all ribs ossify in mem- brane bone in paralepidids and alepisaurids is derived for that group. (740) = All ribs ossify in cartilage (74j) = Some ribs ossify in membrane bone (742) = All ribs ossify in membrane bone (743) = Ribs absent (744) = Some or all ribs ligamentous 75. Origin of Baudelot's Ligament?Baudelot's liga- ment originates on more than one vertebra in most paralepidids (VI and V2) and alepisaurids (V2-V4). In all other aulopiforms and outgroups, Baudelot's ligament originates on VI (VI and the occiput in Meta- velifer). (750) = Baudelot's ligament originates on VI (751) = Baudelot's ligament originates on more than one vertebra (752) = Baudelot's ligament originates on VI and the occiput 76. Ossification of Baudelot's Ligament?Baudelot's ligament is ossified in Harpadon and Saurida, a derived condition that occurs independently in Bathymicrops. Baudelot's ligament is lacking in Gigantura. (760) = Baudelot's ligament is ligamentous (76j) = Baudelot's ligament is ossified (762) = Baudelot's ligament is absent G. Caudal Skeleton 77. Modified Proximal Segmentation of Caudal-fin Rays?Johnson et al. (1996, Figs. 20, 23, and 26) de- scribed a peculiar proximal segmentation of most prin- cipal caudal rays in synodontoids in which a small proximal section is separated from the remainder of the ray by a distinctive joint. The ends of the hemi- trichs that meet at this joint are round, whereas those meeting at joints of the normal segmentation of caudal rays are laterally compressed and curved. (770) = Proximal portion of principal caudal-fin rays not modified (77-,) = Proximal portion of most principal caudal rays with modified segment 78. Segmentation Begins on Distal Half of Each Caudal Ray?In most aulopiforms and outgroups, segmenta- tion of caudal rays begins on the proximal half of each ray, sometimes very close to the attachment of the rays to the caudal skeleton. In Ipnops and Bathymicrops, segmentation of caudal rays begins much farther pos- teriorly, on the distal half of each ray. Our analysis suggests evolution of posteriorly displaced segmenta- tion in the ancestor of Ipnops, Bathymicrops, and Bathy- typhlops, with reversal in the last. Gigantura lacks seg- mentation of caudal-fin rays. (780) = Segmentation begins on proximal half of each caudal ray (78a) = Segmentation begins on distal half of each caudal ray (782) = Caudal rays not segmented 79. Median Caudal Cartilages?A pair of autogenous median caudal cartilages ("CMCs" of Fujita, 1990) is present primitively in aulopiforms and outgroups ex- cept acanthomorphs which have none. CMCs are also absent in synodontoids, Bathymicrops, Bathytyphlops, and Ipnops. The dorsal CMC is absent in Neoscopelus, notosudids, Bathypterois, Lestrolepis, and Stemonosudis and reduced in size in Uncisudis, Lestidium, and Lesiidi- ops. Gigantura has a single median CMC. (790) = Two CMCs, about equal in size (79j) - Two CMCs, the dorsal one minute (792) = One CMC (793) = No CMCs 80. Urodermal?Fujita (1990) noted that a small os- sified urodermal occurs near the proximal end of a caudal-fin ray of the dorsal caudal-fin lobe in some myctophids, Neoscopelus, one species of Aulopus, one Chlorophthalmus, and Bathysaurus. A urodermal is lack- ing in all other aulopiforms and other outgroups ex- amined except Bathysauroides. (800) = No urodermal (80x) = Small urodermal in upper caudal lobe 81. Expanded Neural and Haemal Spines on Posterior Vertebrae?Synodontoids (except Harpadon) have broad laminar expansions on the last three to six pre- ural vertebrae (Johnson et al., 1996, figs. 16, 20, and 21). Neural and haemal spines on PU2 and PU3 are expanded in Bathysauroides and on PU2 in Gigantura. (810) = Posterior neural and haemal spine not ex- panded 14. Interrelationships of Aulopiformes 387 (81 j) = Neural and haemal spines of PU2 expanded (812) = Neural and haemal spines of PU2 and PU3 (to PU6 in some) expanded 82. Number of Hypurals?Presence of six hypurals is primitive for aulopiforms and outgroups. Loss of the sixth hypural occurs in the ancestor of the synod- ontoid clade Pseudotrichonotidae + Synodontidae (Johnson et ah, 1996) and independently in the ances- tral alepisaurid. Five hypurals also characterize Arcto- zenus, but the reduction is the result of fusion of the first and second hypurals (or failure of the two bones to differentiate). Other reductions in number of hy- purals occur in Anotopterus (four or five; one and two sometimes fused, sixth lost) and Bathymicrops (two plates in the young specimen we examined, one com- prising the parhypural and first two hypurals fused distally, and the other hypurals 3-5 fused distally? the distal portions of the plates are cartilaginous, and further differentiation of hypurals may accompany their ossification). (820) = Six hypurals (82a) = Five hypurals; the sixth lost or fused (822) = Five hypurals; the first and second not dif- ferentiated (823) = Four hypurals; the first and second not dif- ferentiated, the sixth lost or fused (824) ?= Two hypurals 83. Number of Epurals?The presence of three epur- als is primitive for aulopiforms and the four aulopi- form suborders, but the number is reduced within each. In synodontoids, a single epural is a synapomor- phy of pseudotrichonotids and synodontids, with re- versal to two in the ancestral harpadontid. Within the Chlorophthalmoidei, two epurals is a derived feature oilpnops, Bathytyphlops, and Bathymicrops, with further reduction to one in the last. Among alepisauroids, evermannellids share a single epural, a reduction in- dependently derived in Anotopterus and Gigantura. Adults of several other aulopiforms, including Para- sudis, Omosudis, Alepisaurus, and some paralepidids, also have only two epurals, but one of them is split, suggesting that it may represent partial fusion of two epural bones. Where available, ontogenetic evidence supports this hypothesis, and we do not recognize this condition as distinct from that of three epurals here (but see character 118 below). Accordingly, al- though R. K. Johnson (1982) interpreted the reduction of epurals to one or two as a synapomorphy of ever- mannellids, omosudids, and alepisaurids, we dis- agree. (830) = Adults with two or three epurals; if two, one split (83J = Adults with two epurals, neither split (832) = Adults with one epural H. Median Fins 84. Fusion of Adjacent Pterygiophores (Figs. 16A, and 16B)?In Omosudis (Fig. 16A), the posterior portion of the proximal-middle element of the penultimate anal-fin pterygiophore is fused to the anterior aspect of the same element of the ultimate pterygiophore. The nine posteriormost pterygiophores are fused in this manner in Alepisaurus (Fig. 16B). The only other aulopiform examined with fused pterygiophores is the paralepidid, Uncisudis, which has most of the dor- sal-fin pterygiophores fused. Among the outgroups, pterygiophores are fused only in Metavelifer. The three aulopiforms in which we observed fused pterygio- phores are young specimens, and the fused cartilagi- nous pterygiophores may separate upon ossification. Nevertheless, the cartilaginous pterygiophores of no other young aulopiform specimens examined are fused. (840) = No fusion of pterygiophores of dorsal or anal fin (84J = Adjacent posterior anal-fin pterygiophores fused (842) = Adjacent dorsal-fin pterygiophores fused 85. Pterygiophores of Dorsal Fin Triangular Proximally (Fig. 16C)?The proximal end of each dorsal-fin ptery- giophore in all evermannellid genera is roughly tri- angular, the result of an expansion of the small flanges that flank the central axis. No other aulopiforms or outgroups have the proximal ends of the dorsal-fin pterygiophores triangular. (850) = Pterygiophores of dorsal fin not triangular proximally (85j) = Pterygiophores of dorsal fin triangular proxi- mally 86. Pterygiophores of Anal Fin Triangular Proxi- mally?Evermannella and Odontostomops have anterior pterygiophores of the anal fin that are triangular proxi- mally. The anal-fin pterygiophores are not modified in Coccorella or in other aulopiforms and outgroups except Scopelarchoides, in which the posterior pterygio- phores of the anal fin are broadened proximally. (860) = Pterygiophores of anal fin not triangular proximally (86x) = Anterior pterygiophores of anal fin triangu- lar proximally (862) = Posterior pterygiophores of anal fin triangu- lar proximally 388 CAROLE C. BALDWIN AND G. DAVID JOHNSON A Penultimate Pterygiophore .Ultimate Pterygiophore B FIGURE 16 Proximal-middle pterygiophores of three alepisauroids: (A) and (B), lateral view of posterior anal-fin pterygiophores from left side of Omosudis lowei, USNM 219982 and Alepi- saurus sp., MCZ 60345, respectively; (C) lateral view of dorsal-fin pterygiophores from left side of Evermannella indica, USNM 235141. I. Pelvic and Pectoral Girdles and Fins 87. Medial Processes of Pelvic Girdle Joined Medially by Cartilage (Figs. 4 and 5)?As already noted, aulopi- forms have very elongate medial pelvic processes that are joined by cartilage. The medial processes are typi- cally much smaller and do not articulate with one another in stomiiforms and myctophiforms and, al- though they overlap in some acanthomorphs (Stia- ssny and Moore, 1992), they are never fused medially as in aulopiforms. (870) = Medial processes not joined medially (87j) = Medial processes joined medially by car- tilage 88. Posterior Processes of Pelvic Girdle Elongate and Widely Separated (Fig. 5A)?Posterior pelvic processes are lacking in most aulopiforms (character 89), and they are small, usually slender processes of various shape and orientation in the outgroups. In all synod- ontoids, the posterior processes are very well devel- oped, widely separated, elongate structures that give the pelvic girdle a "bowed" appearance. (880) Posterior pelvic processes small (or absent) (88:) Posterior pelvic processes elongate, widely sep- arated 89. Posterior Processes of Pelvic Girdle Absent (Figs. 4B, 4D, and 5B-5D)?Posterior pelvic processes are 14. Interrelationships of Aulopiformes 389 primitively present in aulopiforms. In chlorophthal- mids, Bathysauropsis, notosudids, alepisauroids, and giganturoids, the posterior pelvic processes are ab- sent. In Bathypterois, the cartilage joining the medial processes divides posteriorly and forms two slender, widely separated cartilages (Fig. 4C). A similar condi- tion occurs in Ipnops and Bathytyphlops, but the poste- rior cartilaginous processes are very short. In some alepisauroids, the cartilage between the medial pro- cesses continues posteriorly as a single, median carti- lage, and the posterior tip of this cartilage is some- times bifurcate. We interpret this condition as a terminal bifurcation of a median cartilage, in contrast to the formation of short cartilaginous posterior pro- cesses in some ipnopids. It is sometimes difficult to distinguish the two conditions, especially in, e.g., Ba- thytyphlops where the cartilaginous posterior pro- cesses are not as widely separated as in Bathypterois and Ipnops. It is most parsimonious to hypothesize loss of the posterior processes in the ancestor of chlorophthal- moids + alepisauroids + giganturoids with evolution of cartilaginous posterior processes in the ancestral ipnopid. The small specimens of Bathymicrops that we examined appear to lack posterior processes, but in- vestigation of larger material is needed. In young Para- lepis, the lateral edges of the median cartilaginous plates ossify first, creating the impression of well- separated, ossified posterior processes, but these are not homologous with the posterior processes of syn- odontoids. (890) = Ossified posterior processes of pelvic girdle present (89J = Posterior processes are cartilaginous (892) = Posterior processes of pelvic girdle absent 90. Lateral Pelvic Processes (Figs. 4B-4D)?Where the central process bends laterally and terminates, it is capped by a very large cartilaginous process in chlorophthalmoids. InChlorophthalmus, Parasudis, and Scopelosaurus, the process is partially or entirely ossi- fied, but it is cartilaginous in our small specimens of other chlorophthalmoids. All aulopiforms and out- groups examined have a lateral pelvic-fin process, but it is typically only a small nubbin of cartilage capping the tip of the central process. In young specimens of some alepisauroids, a large cartilage also caps the central process, but in adults, only a small lateral carti- lage is present, along with an autogenous cartilage that apparently is pinched off of the large cartilage (character 91). The retention of a large cartilaginous or ossified lateral pelvic process is a synapomorphy of chlorophthalmoids. A similar cartilage is present in Scopelarchus analis, an acquisition independent of that in chlorophthalmoids. (900) = Lateral pelvic processes small (90x) = Lateral pelvic processes large, sometimes os- sifying in adults 91. Autogenous Pelvic Cartilages (Fig. 5D)?Para- lepidids (except Anotopterus), Alepisaurus, and ever- mannellids have a well-developed cartilage that ex- tends dorsally into the body musculature from the region where the lateral pelvic-fin rays articulate with the girdle. In some young specimens, this cartilage is attached by a small rod of cartilage to the cartilage capping the central process, suggesting that it origi- nates as part of the lateral cartilage. A similar cartilage is present in myctophids (Fig. 4A) but lacking in other aulopiforms and outgroups. (910) = Autogenous pelvic cartilages absent (91j) = Autogenous pelvic cartilages present 92. Ventrally Directed Posterior Cartilage of the Pelvic Fin (Fig. 5B)?In scopelarchids, the cartilage joining the medial processes continues posteriorly beyond the posterior tips of the medial processes as a broad carti- laginous plate. It narrows posteriorly then abruptly curves ventrally, terminating as a small, ventrally di- rected process that is bound by connective tissue to the abdominal cavity wall (R. K. Johnson, 1974a). In other alepisauroids, the median cartilage may extend posteriorly beyond the medial processes, but it never deviates from the horizontal. (920) = Cartilage between medial processes, if pres- ent, not terminating in ventrally directed process (922) = Cartilage between medial processes terminat- ing in ventrally directed process 93. Posterior Pelvic Cartilage Elongate (Fig. 5Q-?In evermannellids, the cartilage joining the medial pelvic processes also extends posteriorly as a broad plate, but it is uniquely elongate in this family, extending well beyond the posterior tips of the medial processes, reaching up to two-thirds the length of the bony girdle (R. K. Johnson, 1982). (930) = Cartilage extending posteriorly from be- tween medial processes, if present, not elongate (93J = Cartilage extending posteriorly from be- tween medial processes elongate 94. Position of Pectoral and Pelvic Fins?In alepisaur- oids, the pectoral fins are positioned low on the body (closer to the ventral midline than to the lateral mid- line), and the pelvics are abdominal. These are primi- tive teleostean and neoteleostean features, but they are derived within aulopiforms, which primitively 390 CAROLE C. BALDWIN AND G. DAVID JOHNSON have high-set pectorals and subthoracic pelvics as in synodontoids, chlorophthalmoids, Bathysaurus, and Bathysauroid.es. As noted by Rosen (1973), amore dorsal placement of the pectoral fins and an anterior shift in the pelvic fins appear to be synapomorphies of aulopi- forms plus ctenosquamates, i.e., the Eurypterygii. (940) = Pectoral fins set high on body, pelvics sub- thoracic (94a) = Pectoral fins set low on body, pelvics ab- dominal 95. Relative Position of Abdominal Pelvic Fins?- Primitively in alepisauroids, the abdominal pelvic fins are inserted beneath or behind a vertical through the origin of the dorsal fin. In Sudis, Macroparalepis, Unci- sudis, Lestidiops, Stemonosudis, and Lestrolepis the dor- sal fin originates more posteriorly than in most other alepisauroids (except in Anotopterus in which it is lack- ing), and the abdominal pelvic fins insert anterior to a vertical through the origin of the dorsal. Pelvic fins are absent in juvenile and adult Gigantura; in larvae, they are abdominal and insert beneath the origin of the dorsal fin. (950) = Pelvic fins subthoracic or, if abdominal, in- serting beneath or behind a vertical through the origin of the dorsal fin (95x) = Pelvic fins abdominal, inserting anterior to vertical through dorsal fin 96. Number of Postcleithra?Gottfried (1989) consid- ered the presence of two postcleithra (the second and third of primitive teleosts) as a synapomorphy of cten- osquamates and noted that although the number of postcleithra varies among aulopiforms, the presence of three in basal taxa such as Aulopus indicates that three is primitive for aulopiforms. However, most synodontoids, chlorophthalmoids, and alepisauroids have two or fewer postcleithra, and our analysis sug- gests the primitive number for the order is two. Loss of the dorsal postcleithrum may be a synapomorphy of eurypterygians, not ctenosquamates, as proposed by Stiassny (this volume). Further study of the homol- ogy of postcleithral elements among aulopiforms and ctenosquamates is needed, but Gottfried (1989) noted the two postcleithra of Synodus and Trachinocephalus appear to be the same two (the second and third) that characterize ctenosquamates. Of phylogenetic significance within aulopiforms is the presence of three postcleithra in evermannellids (a synapomorphy of the three included genera), Bathy- sauroides, and Bathysaurus. Gigantura lacks postclei- thra. The number of postcleithra is reduced in most ipnopids (one in Bathypterois and none in Bathymicrops and Ipnops), but the primitive state for the family is am- biguous. (960) = Two postcleithra (96x) = One postcleithrum (962) = Postcleithra absent (963) = Three postcleithra 97. Cleithrum with Strut Extending to Dorsal Postclei- thrum (Fig. 17)?In certain paralepidids, there is a distinctive projection extending from the cleithrum to the dorsal postcleithrum. It is very narrow where it arises from the cleithrum and then broadens posteri- orly at or near its contact with the postcleithrum, and it is often closely applied to the lateral surface of the scapula. This strut occurs among aulopiforms only in Anotopterus, Macroparalepis, Uncisudis, Lestidium, Stem- onosudis, and Lestrolepis and is absent in the out- groups, but some aulopiforms have a small, blunt posterior cleithral projection in the same region. (970) = Cleithrum with small rounded posterior pro- jection or projection absent (972) = Cleithrum with strut extending posteriorly to postcleithrum 98. Orientation of Pectoral-Fin Base?The pectoral- fin base is oriented more horizontally than vertically in alepisauroids, Diplophos, and Metavelifer, and more vertically in other aulopiforms, myctophiforms, and Polymixia. The latter is primitive for aulopiforms. Parr (1928) noted that scopelarchids and evermannellids differ markedly in insertion and development of pec- toral fins, but our observations suggest that although the insertion of the pectorals in scopelarchids (and paralepidids) is not as low on the body as in everman- nellids and alepisaurids, in all of those taxa the base of the fin is more horizontal than in cladistically primitive aulopiforms. In preserved specimens, this reorienta- tion of the pectoral-fin base is easily identified because rather than lying flat against the body the fin projects ventrolaterally. In the reoriented position, the fin movement is more up and down than front and back as in other aulopiforms. The very high-set pectoral fins of Gigantura, which also have a nearly horizontal base, are autapomorphic. (980) = Pectoral-fin base more vertical than hori- zontal (98j) = Pectoral-fin base more horizontal than verti- cal, inserted on the ventrolateral surface of the body (982) = Pectoral-fin base horizontal, inserted on dor- solateral surface of body 99. Greatly Elongated Supracleithrum?Bathytyphlops and Bathymicrops have a very long supracleithrum, 14. Interrelationships of Aulopiformes 391 Cleithral Strut FIGURE 17 Lateral view of pectoral girdle from left side of Stemonosudis rothschildi, AMS I. 22826001. equal to or longer than the cleithrum (Sulak, 1977, fig. 10). In other aulopiforms and outgroups, the su- pracleithrum is shorter than the cleithrum. Merrett and Nielsen (1987) noted that the supracleithrum in Discoverichthys praecox is about equal in length to the cleithrum, suggesting a possible relationship with Ba- thymicrops and Bathytyphlops. (990) = Supracleithrum shorter than cleithrum (99i) = Supracleithrum equal to or longer than clei- thrum 200. Ventral Limb of Posttemporal Not Ossified?R. K. Johnson (1982) noted that scopelarchids and everman- nellids are unique among iniomes in having an un- forked posttemporal. In those families, an ossified dorsal limb articulates with the epiotic, but there is no ossified ventral limb. Instead, a ligament connects the main body of the posttemporal to the intercalar. A forked posttemporal in which both the dorsal and ventral limbs are ossified is present in other aulopi- forms and outgroups. (1000) = Posttemporal forked, both branches os- sified (lOOj) = Posttemporal unforked, the ventral branch ligamentous /. External Morphology 101. Margin of Anal Fin Indented?A derived feature of the Alepisauroidei is the shape of the anal fin, the margin of which is deeply indented near the anterior end. In other aulopiforms and outgroups, the margin of the anal fin may be straight, slightly convex, or concave, but it is usually not deeply indented. Some Polymixia (e.g., P. nobilis) have the anal fin indented similar to that of alepisauroids. Absence of an in- dented anal fin in Anotopterus is a reversal. (1010) = Margin of anal fin not indented (lOlj) = Margin of anal fin indented 202. Scales?Ossified scales on the body and lateral line are primitive for aulopiforms. R. K. Johnson (1982) noted that ossified scales or scale-like structures are absent in three families, Evermannellidae, Omo- sudidae, and Alepisauridae. However, our investiga- tion indicates that ossified lateral-line structures are 392 CAROLE C. BALDWIN AND G. DAVID JOHNSON present in all evermannellids. Body scales are absent in evermannellids, as they are in alepisaurids and all paralepidids examined except Paralepis and Arcto- zenus. Only alepisaurids and giganturids lack ossified body and lateral line scales. (1020) = Body and lateral-line scales present and os- sified (102%) = Body scales absent, lateral-line scales or structures at least partially ossified (1022) ? Body and lateral-line scales or structures absent 103. Fleshy Mid-lateral Keel?R. K. Johnson (1982) considered the presence of a fleshy, mid-lateral keel on each side of the posterior section of the body as a derived feature of alepisaurids (the keel is restricted to the caudal peduncle in Omosudis and covers the posterior one-third to one-half of the body in Alepi- saurus). We agree and note that a fleshy midlateral keel does not occur elsewhere among aulopiforms or outgroups except in Anotopterus, which has a pair of fleshy keels on each side of the caudal peduncle (Ro- fen, 1966c, fig. 182). (1030) = Fleshy mid-lateral keel absent (103J = Single fleshy mid-lateral keel on posterior portion of body (1032) = Pair of fleshy mid-lateral keels on caudal peduncle 204. Body Transparent, Glassy in Life?Rofen (1966b, p. 210) noted that Sudis and all other paralepidids except Paralepis and Arctozenus are "transparent or nearly so, glassy in life, the surface of the skin irides- cent in a kaleidoscope of colors." Our review of the literature indicates that other paralepidids, aulopi- forms, and outgroups may be iridescent, but if so they are silvery and not transparent. We have not examined any living or fresh specimens of Paralepidi- dae, but we tentatively consider the glassy appearance described by Rofen (1966a) as a synapomorphy of Sudis, Macroparalepis, Uncisudis, Lestidium, Lestidiops, Stemonosudis, and Lestrolepis. (104o) = Appearance in life not transparent or glassy (104J = Appearance in life transparent, glassy 105. Scale Pockets in Continuous Flap of Skin?Hartel and Stiassny (1986) hypothesized a sister-group rela- tionship between Parasudis and Chlorophthalmus, citing as evidence the presence of scale pockets in a continu- ous flap of skin. The skin flap is pigmented distally, and thus the overall appearance of pigmentation in those genera is a zig-zag or herringbone pattern (Hartel and Stiassny, 1986; Mead, 1966d). Other au- lopiforms, including other chlorophthalmoids, do not have scales implanted in pockets along a continuous flap of pigmented skin. (1050) = Scale pockets not in continuous flap of skin (105x) ? Scale pockets in a continuous flap of mar- ginally pigmented skin 206. Elliptical or Keyhole Aphakic Space?Mead (1966d) noted that Chlorophthalmus and Parasudis have a keyhole-shaped pupil, created by a conspicuous aphakic (i.e., lensless) space anteriorly. Marshall (1966) and Bertelsen et al. (1976) described a similar, but elliptical, lensless space in notosudids, and we have observed the same condition in Bathysauropsis malayanus, B. gracilis, and Bathysauroides. An aphakic space is lacking in other aulopiforms and outgroups. If the two forms of aphakic space are considered as separate states, the character is ambiguous, and nei- ther state is phylogenetically informative. If we accept the two conditions as primary homologues, the pres- ence of an aphakic space is a synapomorphy of chloro- phthalmoids. We code this character as "missing" in ipnopids, which have greatly reduced or modified eyes. The aphakic space of Bathysauroides is best inter- preted as independently derived; a modification of the iris of that species (incomplete or divided anteriorly at least in subadults) may be further evidence that the eye morphology of Bathysauroides is unique. (1Q60) = No aphakic space (106J = Elliptical or keyhole shaped aphakic space 207. Eye Morphology?A laterally directed round eye characterizes synodontoids, chlorophthalmids, alepisaurids, paralepidids, and Odontostomops. Within the Chlorophthalmoidei, there is a trend toward re- duction in eye size, from slightly flattened or elliptical in Bathysauropsis and notosudids, to minute in most ipnopids. Ipnops lacks recognizable eyes but has broad, lensless light-sensitive organs on the surface of the head (see Mead, 1966c). It is most parsimonious to hypothesize a reduction in eye size in the ancestor of Bathysauropsis, notosudids, and ipnopids with fur- ther reduction in the last. An elliptical eye also charac- terizes Bathysauroides and Bathysaurus. Giganturids are unique among aulopiforms in having anteriorly di- rected telescopic eyes. Scopelarchids and most evermannellids have dor- sally directed semitubular or tubular eyes. The later- ally directed round eyes in Odontostomops, which is the sister group of Evermannella, are best interpreted as a reversal. Lending support to the interpretation of tubular eyes as a synapomorphy of the Evermannel- lidae and Scopelarchidae is the fact that larvae of both families have dorsoventrally elongate eyes (R. K. 14. Interrelationships of Aulopiformes 393 Johnson, 1974a, 1982), implying a similar ontogeny of the adult condition (Character 114). (1070) = Eyes laterally directed, round (107%) = Eyes slightly flattened to elliptical (1072) = Eyes minute or absent (1073) = Eyes dorsally directed, semitubular or tu- bular (1074) = Eyes anteriorly directed, telescopic (1075) = Eyes are broad, lensless plates on dorsal surface of head 108. Gular Fold?Mead (1966b) noted that Bathypt- erois has a thick gular fold that covers the ventral surface of the branchiostegal membranes where they overlap anteriorly. Hartel and Stiassny (1986) noted that a well-developed gular fold is characteristic of all ipnopids as well as Bathysauropsis, and they consid- ered this feature as further evidence that Bathysaurop- sis is an ipnopid. We examined the gular region of all aulopiforms and found that the thickness of the gular fold varies with size. Nevertheless, the gular fold of ipnopids is different from the typical aulopiform condition in that the posterior edge of the fold is crescent-shaped and is not tightly bound to the branchiostegal membranes except along the lateral edges. In most other aulopi- forms, the posterior margin of the gular fold is tent- shaped and tightly bound to the branchiostegal mem- branes. Thus in ipnopids, a probe inserted beneath the fold can be extended to near the symphysis of the dentary bones, whereas in other aulopiforms and most outgroups, extension of a probe beneath the fold anteriorly is impossible because of the attachment of the fold to the branchiostegal membranes. Notosudids, but not Bathysauropsis, also have a crescent-shaped gular fold that is loosely bound to the branchiostegal membranes, a derived feature that we consider further evidence of a sister-group rela- tionship between notosudids and ipnopids. A cres- cent-shaped gular fold is independently derived in Polymixia. (1080) = Gular fold tent-shaped (108;) = Gular fold crescent-shaped 109. Adipose Fin?Presence of a dorsal adipose fin is primitive for aulopiforms, although several out- groups (Diplophos and acanthomorphs) lack an adi- pose fin. Among aulopiforms, an adipose fin is lacking in Pseudotrichonotus, Bathysaurus, and the ipnopids Bathymicrops, Bathytyphlops, and Ipnops. (1090) = Adipose fin present (109J = Adipose fin absent K. Internal Soft Anatomy 110. Mode of Reproduction?R. K. Johnson (1982) hypothesized that synchronous hermaphroditism evolved three times among iniome fishes?once in the ancestor of his chlorophthalmid + ipnopid + noto- sudid + scopelarchid lineage, once in bathysaurids, and again in the ancestor of his alepisauroid clade. Our phylogeny suggests that all of those taxa form a monophyletic group, and thus we hypothesize a sin- gle origin of hermaphroditism, in the ancestor of our chlorophthalmoid + alepisauroid + giganturoid lin- eage. Synodontoids, myctophiforms, and Polymixia have separate sexes (see R. K. Johnson, 1982, for refer- ences) and, although the mode of reproduction in many stomiiforms is unknown, gonochorism also ap- pears to be the primitive aulopiform strategy. (1100) = Separate sexes (llOj) = Synchronous hermaphrodites 111. Thin-Walled, Heavily Pigmented Stomach?R. K. Johnson (1982) considered the presence of a highly distensible black stomach as a derived feature of alepi- saurids. He noted that other iniomes examined by him have a heavily muscularized, unpigmented stomach. (1110) = Stomach not highly distensible, with thick unpigmented walls (lllj) = Stomach highly distensible, with thin heav- ily pigmented walls 112. Swimbladder?Aulopiforms lack a swimblad- der, but ctenosquamates primitively have one (absent in some myctophiforms) as do most stomiiforms (e.g., gonostomatids, sternoptychids, photichthyids, some astronesthids, and stomiids) (Marshall, 1954, 1960; Marshall and Staiger, 1975; R. K. Johnson, 1982). (1120) = Swimbladder present (112x) = Swimbladder absent L. Larval Morphology 113. Enlarged Pectoral Fins?Okiyama (1984b) noted that ipnopid larvae share the derived condition of greatly enlarged, fanlike pectoral fins. Larvae of Sudis hyalina also have elaborate pectorals (Okiyama, 1984a, fig. 113F), and larvae of Bathysaurus have all fins except the caudal greatly enlarged (Okiyama, 1984a, fig. 111C). The pectoral fins of larval Rosenblattichthys are well developed relative to other scopelarchids (see R. K. Johnson, 1984a, fig. 127A,B) but not nearly as much as in ipnopids. Okiyama (1984b, p. 256) indi- cated in his character matrix that alepisaurids have elongate pectoral fins, but the illustrations of A. brevir- ostris and A. ferox (Okiyama, 1984a, fig. 112A,B) do 394 CAROLE C. BALDWIN AND G. DAVID JOHNSON not reflect this condition. Pectoral fins are enlarged in certain myctophiforms (e.g., some Lampanyctus, Moser et al., 1984, Fig. 124F) but not in other out- groups. (1130) = Pectoral fins not enlarged in larvae (113i) ? Pectoral fins enlarged in larvae 224. Elongate Eyes?The eyes are dorsoventrally elongate in larval scopelarchids and evermannellids. R. K. Johnson (1984b) noted that the eyes are not elongate in larvae of Odontostomops, and thus ever- mannellids and scopelarchids may have indepen- dently acquired them. We have not examined larval Odontostomops, but in illustrations of O. normalops (R. K. Johnson, 1982, Figs. 5D and 6D) the eye appears to be slightly wider than in other everman- nellids, but it is dorsoventrally elongate rather than round. Notosudid larvae also have narrow eyes, but they differ from evermannellid eyes in being elongate in the anteroposterior plane (Bertelsen et al., 1976; Oki- yama, 1984a, fig. 111A). Some myctophids have dor- soventrally elongate eyes, but round eyes are primi- tive for aulopiforms. (114g) = Eyes in larvae round (114i) = Eyes in larvae elongate; the horizontal axis longer than the vertical (1142) = Eyes in larvae elongate; the vertical axis longer than the horizontal 225. Head Spination?Head spines are uncommon in larvae of non-acanthomorph teleosts, but serrate cranial ridges and preopercular spines are present in a strikingly similar configuration in Alepisaurus ferox and Omosudis (Okiyama, 1984a, figs. 112B, 112E, and 112F). Larvae of A. brevirostris apparently lack head spines (Rofen, 1966b, fig. 171; Okiyama, 1984a, fig. 112A), and thus ornamentation in the two genera could be nonhomologous. However, the presence of two nearly identical patterns of head spines among a group of teleosts that are not known for elaborate head ornamentation leads us to believe that the conditions in A. ferox and Omosudis are ho- mologous. The paralepidid Sudis also has head ornamentation, in the form of serrate cranial ridges and a large, stron- gly serrate spine at the angle of the preopercle. Other paralepidids lack head spines, and it is thus most parsimonious to hypothesize independent acquisition of head ornamentation in Sudis and the Alepisauridae. (H5o) (HSi) Head spines lacking in larvae Head spines present in larvae 226. Peritoneal Pigment?As noted, R. K. Johnson (1982) suggested that peritoneal pigment in larvae may be diagnostic of Rosen's (1973) Aulopiformes, a notion supported in our analysis, despite the absence of peritoneal pigment in larvae of some chlorophthal- moids (notosudids, Ipnops, Bathymicrops, and some Bathypterois). Okiyama (1984b) and R. K. Johnson (1982) described several states of this character: a sin- gle, unpaired peritoneal pigment "section"; multiple, unpaired pigment sections; a single unpaired section changing ontogenetically to several unpaired sections; multiple paired pigment spots; and absence of perito- neal pigment. Johnson et al. (1996) considered the presence of paired peritoneal pigment spots in larvae and juveniles a synapomorphy of Pseudotrichonotus and synodontids. These spots are retained in the ab- dominal wall of adults as tiny dense discs of pigment. Our investigation suggests that the presence of one or more unpaired peritoneal pigment sections is prim- itive for aulopiforms, and thus we concur with John- son et al. (1996) that the presence of paired peritoneal pigment sections in some larval synodontoids is de- rived. (1160) = Peritoneal pigment absent in larvae (116^ = Single or multiple unpaired peritoneal pig- ment sections in larvae (1162) = Multiple paired peritoneal pigment sections in larvae 227. Ontogenetic Reduction of Large Maxilla (Fig. 18)?Adults of Gigantura have only a small maxillary remnant posteriorly, but in larval giganturids ("Ro- saura") the maxilla is a very large, leaf-shaped bone that tapers abruptly anteriorly near its articulation with the premaxilla (Fig. 18B). Rosen (1971) discussed the relationships of Regan's (1903) Macristiidae, a "myctophoid" family described on the basis of a single specimen of Macristium chavesi that is now lost. He described a new Macristium-]ike larval fish (the "Chain" larva) and concluded that it is probably the young stage of Bathysaurus, a notion corroborated by R. K. Johnson (1974b). In his paper, Rosen (1971) illus- trated a lateral view of the skull bones of the "Chain larva" (Fig. 18A). Adult Bathysaurus have only a small, anterior remnant of the maxilla (e.g., Sulak, 1977, fig. 5A), but Rosen's illustration shows a very large max- illa in the larva that bears a striking resemblance to that of larval Gigantura. It is large and leaf-shaped and tapers abruptly anteriorly (Fig. 18). Dramatic ontogenetic reduction of a large maxilla is thus shared by Bathysaurus and Gigantura, and we have not observed it elsewhere in the Aulopiformes, including synodontids in which the maxilla is reduced in adults (see e.g., Okiyama, 1984a, figs. 111D-111G). Larval Bathysauroides are undescribed, but adults have a well-developed maxilla; we thus predict that the 14. Interrelationships of Aulopiformes .fr 395 brstg- 5 MM H -I FIGURE 18 Larvae of (A) Bathysaurus (from Rosen, 1971, fig. 5, depicting the syncranium of the "Chain" larva) and (B) Gigantura (from Tucker, 1954, fig. 1). maxilla in larval Bathysauroid.es is not enlarged or re- duced ontogenetically. (1170) = Maxilla not enlarged in larva, not greatly re- duced ontogenetically (117J = Maxilla enlarged in larva, greatly reduced ontogenetically 118. Ontogenetic Fusion of Epurals?Adult Parasudis have two epurals, but the anterior is split distally. Larval Parasudis have three epurals, suggesting the adult condition is the result of partial ontogenetic fu- sion of the first and second epurals. Partial fusion of two epurals also apparently occurs in Omosudis, Alepisaurus, Lestrolepis, Lestidiops, and Stemonosudis, which have two epurals in adults, one of which is split proximally. As in larval Parasudis, larval Stemonosudis have three cartilaginous epurals. We have not exam- ined this feature in larvae of other paralepidids and alepisaurids listed above, but it is reasonable to assume that the divided epurals in adults of those taxa are also the result of ontogenetic fusion. Our analysis indicates that such ontogenetic fusion occurred three times within aulopiforms: in Parasudis, in the ancestral alepi- saurid, and in the ancestor of the paralepidid clade comprising Lestidiops, Stemonosudis, and Lestrolepis. (1180) = No ontogenetic fusion of epurals (118i) = Partial ontogenetic fusion of two epurals 396 CAROLE C. BALDWIN AND G. DAVID JOHNSON VI. Discussion Monophyly of Rosen's (1973) Aulopiformes is supported by seven derived features. Four characters were previously recognized as synapomorphies of the order: (1J an enlarged EB2 uncinate process (Rosen, 1973); (2j) absence of a cartilaginous condyle on PB3 for articulation of EB2 (Johnson, 1992); (54x) anterior extension of the epipleural series to at least V2 (Patterson and Johnson, 1995); and (116J perito- neal pigment in larvae (R. K. Johnson, 1982). Two previously described characters were not recognized as aulopiform synapomorphies: (55x) displacement of one or more of the anterior epipleurals dorsally into the horizontal septum (Patterson and Johnson, 1995) and (112x) absence of a swimbladder (e.g., R. K. Johnson, 1982). We identified a seventh diag- nostic feature of aulopiforms, (87J fusion of the medial processes of the pelvic girdle. Additionally, although not recognized formally in our analysis, a benthic existence may be a synapomorphy of aulopiforms. Because stomiiforms and primitive ctenosquamates are pelagic (polymixiids and Meta- velifer are benthopelagic), and adults of most aulopi- forms are benthic, a transition from a pelagic to a benthic environment may have characterized the ancestral aulopiform. Several aulopiforms have rein- vaded the pelagic realm. Aulopiform genera comprise four major clades that we designate the suborders Synodontoidei, Chloro- phthalmoidei, Alepisauroidei, and Giganturoidei. Be- low we summarize the evidence supporting the mo- nophyly of those clades and relationships among them. Within suborders, we emphasize characters supporting newly proposed clades as well as those previously undescribed or recognized as synapomor- phies at different taxonomic levels. Limits and relationships of the Synodontoidei of Johnson et al. (1996) are well supported in this study, with each clade being diagnosed by five or more un- ambiguous derived features (Fig. 6). Aulopus is cladis- tically the most primitive synodontoid, a hypothesis that conflicts with previous proposals (e.g., Rosen, 1985; Hartel and Stiassny, 1986) in which aulopids and sometimes chlorophthalmids were considered more closely related to ctenosquamates than to other aulopi- forms. Synodontoids share eight derived features (Fig. 6), including two not recognized by Johnson et al. (1996): (172*) gap between BB4 and CB5s and (88,) elongate, widely separated posterior pelvic processes. Most of the homoplasy within the group occurs in the highly modified Pseudotrichonotus and the secondarily free-swimming Harpadon. Sulak (1977) considered Bathysaurus as a subfamily of his expanded Synodontidae, but our data reject that notion. Bathysaurus lacks all synapomorphies of synodontoids and the clade comprising Pseudotricho- notus + synodontids and has only 2 of the 10 derived features uniting synodontids (Fig. 6): (5X) gill rakers reduced to toothplates and (690) reduced number of caudal vertebrae. Rosen (1973) argued that synodontids and harpa- dontids are closely related to alepisauroids and in- cluded the superfamily Synodontoidea in his subor- der Alepisauroidei. He appears to have based this on three characters, a single upper pharyngeal toothplate (UP4 or UPS), enlarged orobranchial teeth, and gill rakers present as toothplates. Johnson (1992) noted that all alepisauroids except Anotopterus have both UP4 and UPS. UP4 is absent (3J only in Pseudotricho- notus and synodontids (Fig. 3B), a derived feature of that clade. Enlarged orobranchial teeth also fails as a synapomorphy of synodontids and alepisauroids because the enlarged teeth are premaxillary in syno- dontids and their relatives, whereas in alepisauroids, premaxillary teeth are often minute, and the enlarged teeth are on the palatine (362). Rosen's third character, (5i) gill rakers present as toothplates, is apparently independently derived in synodontoids and alepi- sauroids. The remaining aulopiforms?chlorophthalmoids, alepisauroids, and giganturoids?form a novel clade diagnosed on the basis of four derived features: (30j*) anterior ceratohyal bearing four or fewer branchioste- gals, (722) neural arches open dorsally only on the anteriormost four or fewer vertebrae, (892) ossified posterior pelvic processes absent, and (HOJ sexual reproduction by synchronous hermaphroditism. Most of these fishes inhabit depths of 1000 to 6000 m, and the evolution of synchronous hermaphroditism may have contributed to their successful radiation into the deep. Synodontoids have separate sexes and are pri- marily shallow-water fishes. The Chlorophthalmoidei include the Chlorophthal- midae, Bathysauropsis (c.f. B. gracilis and B. malayanus), Notosudidae, and Ipnopidae. Monophyly of chlor- ophthalmoids is supported by the following: (6J prox- imal end of PB2 expanded laterally; (45a) medial edge of premaxilla with a dorsomedially directed process; (682*) one supraneural; (90%) central process of pelvic girdle capped laterally by a very large winglike pro- cess, ossified in some taxa; and (106a) pupil elliptical or keyhole-shaped, with a prominent aphakic space anteriorly (except in ipnopids where eyes are minute or greatly modified). The Chlorophthalmidae (Chlorophthalmus and Para- sudis) share three previously described derived fea- 14. Interrelationships of Aulopiformes 397 hires (442, 63], and 105%) relating to squamation, inter- musculars, and jaw morphology (Hartel and Stiassny, 1986; Stiassny, 1986; Patterson and Johnson, 1995). Ipnopids have a small, obliquely aligned basihyal, and its presence in Bathysauropsis gracilis led Hartel and Stiassny (1986, fig. 7) to reassign Bathysauropsis to the Ipnopidae (from the Chlorophthalmidae). Our phy- logeny indicates that Bathysauropsis is the sister group of ipnopids + notosudids, and thus we interpret (34*) an obliquely aligned basihyal as a synapomorphy of Bathysauropsis, notosudids, and ipnopids, with rever- sal in notosudids. The Bathysauropsis clade also shares (23j) gill rakers extending onto lateral surfaces of deep basibranchials, (300*) five or more branchiostegals on anterior ceratohyal, and (107j) reduced or modified eyes relative to the very large, round eyes of chloroph- thalmids. A sister-group relationship between notosudids and ipnopids has not been proposed previously. Ber- telsen et al. (1976) suggested that notosudids are most closely related to chlorophthalmids, R. K. Johnson (1982) placed notosudids as the sister group of his scopelarchid + chlorophthalmid + ipnopid clade, and Patterson and Johnson (1995) considered notosudids as the sister group of the Scopelarchidae + Everman- nellidae. R. K. Johnson (1982) based his hypothesis on two derived features, absence of a swimbladder and presence of synchronous hermaphroditism, but we consider those features as synapomorphies of au- lopiforms and the chlorophthalmoid + alepisauroid + giganturoid clade, respectively. Patterson and John- son (1995) cited the origin of epineurals on the cen- trum or parapophysis on about vertebrae 5-15 as evi- dence for their placement of notosudids, but our analysis suggests independent evolution of ventrally displaced epineurals in notosudids and alepisauroids. Our hypothesis of a sister-group relationship between notosudids and ipnopids is supported by (62) an un- usual modification of PB2 in which the proximal end has an extra uncinate process, (792) absence of at least one CMC, and (108%) a thick, crescent-shaped gular fold. The notosudid genera Scopelosaurus and Ahliesaurus share 10 derived features (Fig. 6), including (20]) elon- gate BB1, (33]) anteriormost branchiostegal on ventral hypohyal, (38j) quadrate with two cartilaginous heads, and (57]) epipleurals forked distally at transi- tion of epipleurals in and beneath horizontal septum. Although we did not examine the monotypic Lucio- sudis, information from Bertelsen et al. (1976) suggests that L. normani has at least two synapomorphies of Scopelosaurus and Ahliesaurus, (46]) seven infraorbitals and (114]) horizontally elongate eyes in larvae. We conclude that the Notosudidae are monophyletic, but further study is needed to elucidate relationships among the three genera. Ipnopids (Bathypterois, Bathytnicrops, Bathytyphlops, and Ipnops) share nine derived features (Fig. 6), in- cluding (113]) an enlarged pectoral fin in larvae, a condition that occurs elsewhere among aulopiforms and the outgroups only in Sudis, Rosenblattichthys, and Bathysaurus. Ipnopids also share the following: (7j) UP2 usually absent, (42]) metapterygoid free from hy- omandibular, (52]) frontal expanded laterally over or- bit, (53j) sphenotic with an anteriorly directed process extending beneath frontal, (733) ribs, when present, beginning on V2, and (89J posterior processes of pel- vic girdle cartilaginous; most have (1072) minute eyes. Some of these features are reversed in Bathymicrops, which lacks ribs, has UP2, and apparently lacks poste- rior pelvic processes. Nevertheless, our analysis places Bathymicrops as the sister group of Bathy- typhlops, as proposed by Sulak (1977). The two share (43]) a horizontally oriented hyomandibular and oper- cle, (99]) a long supracleithrum, (21]) an elongate BB2, and (24]) ossification of the ligament between HB1 and the hyoid. Bathypterois, formerly placed in a separate family (Bathypteroidae; see, e.g., Mead (1966b)), is the sister group of the other ipnopid genera, which are united on the basis of several, mostly reductive, derived fea- tures: (463) five (or fewer) infraorbitals, (793) loss of CMCs, (832) two (or one) epurals, and (109]) absence of an adipose fin. They also share (692) a high percent- age of caudal vertebrae. Ipnops and Bathymicrops have (19]) ossified BB4 and (78]) segmentation of caudal rays beginning on distal half of each ray, additional features treated as synapomorphies of Ipnops + Ba- thymicrops + Bathytyphlops in our analysis, with rever- sal in Bathytyphlops. We did not examine the single known specimen of the ipnopid Discoverichthys praecox, but we used data from Merrett and Nielsen (1987) to explore its relation- ships. Although the configuration of the gill arches, pelvic girdle, and intermusculars are unknown, Discoverichthys lacks a swimbladder and is hermaphro- ditic, suggesting that it belongs in the chlorophthal- moid + alepisauroid + giganturoid clade of aulopiforms. Because the premaxilla is the dominant tooth-bearing bone of the upper jaw, and the gillrakers are lathlike, Discoverichthys is best placed in the chlorophthalmoid lineage. It has the well-developed gular fold of noto- sudids, Bathysauropsis, and ipnopids, the small oblique basihyal of Bathysauropsis and ipnopids, the minute eye of most ipnopids, and, like Bathymicrops, Bathytyphlops, and Ipnops, it lacks an adipose fin. Dis- coverichthys does not have the opercle and hyomandi- bular reoriented as in Bathymicrops and Bathytyphlops, 398 CAROLE C. BALDWIN AND G. DAVID JOHNSON nor does it share with those genera a greatly elongated supracleithrum. We tentatively conclude that Dis- coverichthys is most closely related to the clade com- prising Bathymicrops, Bathytyphlops, and Ipnops, but it does not appear to belong to the Bathymicrops + Bathy- typhlops group. Alepisauroids and giganturoids form another new clade in our tree and share several derived features, most notably the following: (5j) gill rakers present as toothplates, (362) palatine the dominant tooth-bearing bone of the "upper jaw", (542) epipleurals extending to VI, and (61}*) origin of some (or all) epineurals on centrum. Adults of Gigantum lack a dermopalatine and most elements of the branchial skeleton but share with alepisauroids, Bathysaurus, and Bathysauroides the anterior extension of epipleurals to VI (see discus- sion of Gigantura below). Our Alepisauroidei comprise the Alepisauridae (in- cluding Omosudis), Paralepididae (including Anotopt- erus), Evermannellidae, and Scopelarchidae. Rosen's (1973) alepisauroids were characterized by gill-arch morphology, especially attenuation of epibranchial and pharyngobranchial elements, absence of UP2, UPS, and a toothplate on EB3 (ET3), and large pharyn- gobranchial teeth. UP4 and UPS are present in most alepisauroids, and large pharyngobranchial teeth also characterize giganturoids. Aulopiforms vary consider- ably in length of epibranchial and pharyngobranchial elements and the presence of ET3, and neither con- vincingly diagnoses alepisauroids. However, alepi- sauroids do have distinctive gill arches, characterized in part by (7J absence of UP2. Other diagnostic features of alepisauroid gill arches include: (llj) teeth on UP3 (when present) restricted to lateral edge, (152) teeth on CBS restricted to medial edge, and (220) gillrakers (pres- ent as toothplates) not extending onto HB3. Alepisaur- oids also share the following: (462*) eight infraorbitals, (68J*) two supraneurals, (91x*) autogenous lateral pel- vic cartilages, (94J abdominal pelvic fins, (98J a nearly horizontal (or more horizontal than vertical) pectoral- fin base, and (lOlj) an indented anal fin. Furthermore, the pelagic lifestyle of alepisauroids may represent a single evolutionary transition from the benthic exis- tence of primitive aulopiforms. We agree with R. K. Johnson (1982) that Omosudis and Alepisaurus are sister taxa. They share 12 unambig- uous derived characters, including features of the gill arches, intermuscular system, caudal skeleton, exter- nal morphology, internal soft anatomy, and head spi- nation in larvae (Fig. 6), the following several of which are previously unrecognized alepisaurid synapomor- phies: (10J PB3 extending anteriorly beyond EB1 and PB2, (18j) BB3 extending beneath BB4, (58j) epipleur- als on VI and V2 fused to centrum, (65^ most epineur- als unattached, and (84x) adjacent posterior anal-fin pterygiophores fused. The close relationship between Omosudis and Alepisaurus is best represented by refer- ring Omosudis to the Alepisauridae. Patterson and Johnson (1995) hypothesized a sister- group relationship between the Omosudis + Alepi- saurus clade and paralepidids. They based this on three derived features: (742) all ribs ossified in mem- brane bone, (76j) Baudelot's ligament originating on more than one vertebra, and epineurals on the first five or fewer vertebrae fused to the neural arch. Exam- ination of additional taxa indicates that epineurals are free from the axial skeleton except in the two genera, Paralepis and Macroparalepis, examined by Patterson and Johnson (1995). An additional but ambiguous synapomorphy of alepisaurids and paralepidids is (734) ribs originating on VI, Further study of this group is clearly needed. We concur with R. K. Johnson (1982) that paralepid- ids and Anotopterus form a monophyletic lineage. In addition to his character, (48%) a fenestrate premaxilla, they share (202) an elongate BB1, (47j) a prolonged snout, (49j) an anterior extension of the palatine to meet the premaxilla, (50a) a long horizontally oriented lacrimal on the elongate snout, and (662) absence of epicentrals. Relationships among the speciose para- lepidids are poorly understood, and we have contrib- uted little toward their resolution. Our preliminary data do not corroborate all aspects of the classifications of Rofen (1966a) and Post (1987), wherein Sudis is given subfamilial or familial status, respectively, and the remaining genera are divided between two tribes or subfamilies. Post (1987) included Arctozenus, Magni- sudis, Notolepis, and Paralepis in his subfamily Para- lepidinae based on apparently primitive aulopiform features (e.g., cycloid body scales, no luminous or- gans, and no ventral adipose fin). We examined two genera of Post's Paralepidinae, Paralepis and Arcto- zenus, and found that they share three intermuscular characters (592, 632, and 652) as well as (22a) gill rakers (present as toothplates) on HB3 and (32a) branchio- stegals on anterior ceratohyal in 3+1 pattern. They lack the diagnostic features of the lineage comprising Anotopterus and all other paralepidid genera, includ- ing Sudis: (112) UP3 absent, (97a) cleithral strut present; and (102j) body scales absent but ossified lateral-line scales present. A toothplate fused to PB3 is a conserva- tive feature among euteleosts, and its absence is strong evidence of the phylogenetic integrity of this paralepidid group. Placement of Anotopterus as the sister group of one paralepidid clade requires its inclu- sion in the Paralepididae. Sudis shares with Lestidiops, Lestidium, Lestrolepis, Macroparalepis, Stemonosudis, and Uncisudis (60t) a re- 14. Interrelationships of Aulopiformes 399 duced number of epipleurals and (104^ a transparent, "glassy" body. A close association between the main branch of EB2 and its uncinate process (9j) and (32J a 3 + 1 pattern of branchiostegals on the anterior ceratohyal unite all of those genera, excluding Sudis, as a monophyleric assemblage. Uncisudis, Lestidium, Lestidiops, Stemonosudis, and Lestrolepis have (79j) the dorsal CMC reduced to a tiny nubbin (or absent). Lestidiops, Stemonosudis, and Lestrolepis exhibit (118J partial ontogenetic fusion of two epurals. Finally, Les- trolepis and Stemonosudis share (792) absence of the dorsal CMC. No further resolution of relationships among paralepidid genera is evident from our data, and further study is needed. We agree with R. K. Johnson (1982) that Coccorella, Evermannella, and Odontostomops constitute a mono- phyleric Evermannellidae but diagnose the family based on 10 additional derived features (Fig. 6). Most striking among these are (342) basihyal oriented at about a 90? angle to first basibranchial, (663) anterior epicentrals cartilaginous, (85a) pterygiophores of dor- sal fin triangular proximally, and (93j) a long tail of cartilage extending posteriorly from the pelvic girdle. Our data do not corroborate R. K. Johnson's (1982) hypothesis of a sister-group relationship between Coc- corella and Evermannella. Rather, three derived fea- tures indicate that Evermannella and Odontostomops are sister taxa: (27%) third hypobranchials fused ventrally, (31j) posteriormost two branchiostegals close, and (86j) proximal ends of anal-fin pterygiophores ex- panded. The Scopelarchidae are monophyletic, as proposed by R. K. Johnson (1974a, 1982), the four genera (Ben- thalbella, Scopelarchus, Scopelarchoides, and Rosenblat- tichthys) sharing reversals of several derived alepisaur- oid conditions (460*, 680*, and 910) as well as three novel derived features: (35a) large, posteriorly curved basihyal teeth; (744) some or all ribs in ligament; and (92]) a median cartilage extending posteriorly from the pelvic girdle that bends down to terminate as a small, ventrally directed process. Our data do not elucidate relationships within the Scopelarchidae. Although scopelarchids were traditionally placed near evermannellids (e.g., Marshall, 1955; Gosline et at, 1966), R. K. Johnson (1982) suggested that resem- blances between the two families may be superficial. Five unambiguous synapomorphies support a sister- group relationship between the Scopelarchidae and Evermannellidae: (67J attachment of anterior epicen- trals to distal ends of epipleurals, (692) high percent- age (>60%) of caudal vertebrae, (100a) unossified ven- tral posttemporal limb, (1142) dorsoventrally elongate eyes in larvae, and (1073) dorsally directed, semitubu- lar or tubular eyes in adults. Eyes are lateral and not tubular in Odontostomops, and R. K. Johnson (1982) hypothesized that tubular eyes are a synapomorphy of Coccorella and Evermannella. Our hypothesis of a sister-group relationship between Evermannella and Odontostomops indicates that the absence of tubular eyes in Odontostomops is best interpreted as a reversal of the primitive evermannellid + scopelarchid con- dition. R. K. Johnson (1982) hypothesized that everman- nellids, not paralepidids, are the sister group of the alepisaurid clade and that scopelarchids are part of a clade comprising notosudids, chlorophthalmids, and ipnopids. His arrangement of evermannellids and alepisaurids is five steps longer than ours, and inclu- sion of scopelarchids in our Chlorophthalmoidei re- quires at least 18 additional steps. Patterson and John- son's (1995) placement of the Evermannellidae + Scopelarchidae clade as the sister group of notosud- ids, which was based on a single feature of the inter- musculars, is 19 steps longer than our hypothesis. Our giganturoids include Bathysauroides, Bathy- saurus, and Gigantura, but historically relationships of these fishes have been perceived differently: Bathy- sauroides (along with Bathysauropsis gracilis and B. ma- layanus) was considered a chlorophthalmid (Sulak, 1977) or ipnopid (Hartel and Stiassny, 1986); Bathy- saurus was considered a synodontid by Sulak (1977) and a close relative of aulopids and chlorophthalmids by Rosen (1973); and Gigantura, which has only some- times been included in the aulopiforms (see discus- sion below), was considered closely related to synod- ontids by Rosen (1973). Support for the Giganturoidei is not strong because most derived features shared by Bathysauroides and Bathysaurus are absent in the highly modified Gigantura, but our analysis suggests they are united on the basis of five derived features: (13j) elongate FBI (FBI absent in adult Gigantura); (690) reduced number (<25%) of caudal vertebrae; (80]*) small urodermal in upper caudal lobe (absent in Gigantura); (963) three postcleithra (none in Gigantura); and (107]) elliptical eyes (eyes greatly modified in Gi- gantura). Gigantura has usually been placed in a separate order (e.g., Regan, 1925; Berg, 1940; Walters, 1961). Regan (1925) suggested that giganturids might be re- lated to synodontids, and Rosen (1973) concluded that giganturids are alepisauroid aulopiforms, most closely related to synodontids and harpadontids. Ro- sen's hypothesis was not based on explicit evidence, and, as he noted, the gill arches of adult Gigantura are much reduced and do not exhibit the distinctive EB2 uncinate process diagnostic of aulopiforms. The gill arches of larval Gigantura, however, are more com- plete, and our examination of them indicates the pres- 400 CAROLE C. BALDWIN AND G. DAVID JOHNSON PB3 PB2' EB4 FIGURE 19 Ventral view of dorsal gill arches from right side of larval Gigantura chuni, MCZ 60324. ence of the characteristic EB2 uncinate process (Fig. 19). Furthermore, Patterson and Johnson (1995) noted that intermuscular data, particularly (54x 2) the ante- rior extension of epipleurals, support inclusion of gi- ganturids in the Aulopiformes. Gigantura also has three additional aulopiform synapomorphies: (55J first epipleural in horizontal septum (Patterson and Johnson, 1995), (112J swimbladder absent, and (116x) peritoneal pigment in larvae. We believe the evidence convincingly places the bizarre giganturids in the Au- lopiformes. Giganturids are aligned with chlorophthalmoids, alepisauroids, and other giganturoids based on (110J reproduction by synchronous hermaphroditism (Johnson and Bertelsen, 1991), and they share with alepisauroids and other giganturoids (542) anterior ex- tension of epipleurals to VI and (12J large pharyngo- branchial teeth. Patterson and Johnson (1995) suggested a sister- group relationship between Bathysaurus and Gigantura based on two derived features: (690) reduction in num- ber of caudal vertebrae and (61%) origin of most or all epineurals on centra rather than neural arches. A reduced number of caudal vertebrae is a synapomor- phy of giganturoids, and the latter character is ambig- uous (it could be a synapomorphy of giganturoids with reversal in Bathysauroides gigas), but our analysis supports Patterson and Johnson's (1995) interpreta- tion. Bathysaurus and Gigantura also share (662) epicen- tral series absent (this occurs elsewhere among aulopi- forms only in paralepidids); (720) most neural arches open dorsally (a reversal of the primitive chloroph- thalmoid + alepisauroid + giganturoid condition); and (117x) maxilla reduced ontogenetically from a very large broad bone in larvae to a small anterior (Bathy- saurus) or posterior (Gigantura) remnant in adults. In summary, Aulopiformes are monophyletic and comprise four monophyletic suborders. Our suborder Synodontoidei is the same as that of Johnson et al. (1996). Our suborder Chlorophthalmoidei is similar to R. K. Johnson's (1982) chlorophthalmoid clade except that we exclude the Scopelarchidae. Our suborder Alepisauroidei comprises the same recent genera as Rosen's (1973) superfamily Alepisauroidea, and our suborder Giganturoidei combines the new genus Ba- thysauroides with the giganturid-bathysaurid lineage proposed by Patterson and Johnson (1995). Among the most significant aspects of our phylogeny are the following: Aulopus is a synodontoid and thus not closely related to ctenosquamates. Synodontoids are not alepisauroids but the primitive sister group of all other aulopiforms. Bathysaurus is not a synodontid but a giganturoid. Bathysauropsis is polyphyletic, B. gracilis and B. malayanus being more closely related to notosudids and ipnopids than to B. gigas. Bathysaurops gigas Kamohara ( = Bathysauropsis gigas) is the type species of a new genus, Bathysauroides, which is re- lated to bathysaurids and giganturids. Omosudis is re- assigned to the Alepisauridae, and Anotopterus is reassigned to the Paralepididae. Scopelarchids are alepisauroids and the sister group of evermannellids. And finally, Gigantura is an aulopiform and may be the sister group of Bathysaurus. Further study is needed to elucidate relationships within the Paralepididae and Scopelarchidae and to test all poorly supported rela- tionships hypothesized herein. We have examined certain aspects of aulopiform morphology in detail, but there is much yet to be studied; we view this work as a foundation for further study of this diverse order of fishes. VII. CLASSIFICATION As diagnosed here, the extant aulopiforms com- prise 43 genera. Bathysauropsis and Bathysauroides have no familial assignment in our phylogeny, but we assign the remaining 41 genera to 12 families. A new classification of aulopiform genera reflecting phy- logenetic relationships as perceived herein follows (suborders are listed in phyletic sequence): Order Aulopiformes Suborder Synodontoidei Family Aulopidae (Aulopus) Family Pseudotrichonotidae (Pseudotrichonotus) Family Synodontidae (Harpadon, Saurida, Syn- odus, Trachinocephalus) 14. Interrelationships of Aulopiformes 401 Suborder Chlorophthalmoidei Family Chlorophthalmidae (Chlorophthalmus, Parasudis) Bathysauropsis (B. gracilis, B. malayanus) Family Notosudidae (Ahliesaurus, Luciosudis, Scopelosaurus) Family Ipnopidae (Bathymicrops, Bathypterois, Bathytyphlops, Discoverichthys, Ipnops) Suborder Alepisauroidei Family Alepisauridae (Alepisaurus, Omosudis) Family Paralepididae (Anotopterus, Arctozenus, Dolichosudis, Lestidiops, Lestidiunt, Lestrolepis, Macroparalepis, Magnisudis, Notolepis, Para- lepis, Stemonosudis, Sudis, Uncisudis) Family Evermannellidae (Coccorella, Everman- nella, Odontostomops) Family Scopelarchidae (Benthalbella, Rosenblat- tichthys, Scopelarchoides, Scopelarchus) Suborder Giganturoidei Bathysauroides gigas (new genus) Family Bathysauridae (Bathysaurus) Family Giganturidae (Gigantura) VIII. Summary Relationships among aulopiform genera are inves- tigated based on cladistic analysis of 118 morphologi- cal characters. Monophyly of Rosen's (1973) Aulopi- formes, which he diagnosed on the basis of unique modifications in the dorsal gill arches, is corroborated by features of the intermuscular system, internal soft anatomy, and larval pigmentation as well as new evidence from the morphology of the pelvic girdle. Our analysis suggests four aulopiform clades, listed below in phyletic sequence: (1) Synodontoidei (Aulopidae, Pseudotrichonotidae, and Synodonti- dae?including Harpadon and Sauridd); (2) Chloro- phthalmoidei (Chlorophthalmidae, Bathysauropsis, Notosudidae, and Ipnopidae); (3) Alepisauroidei (Alepisauridae?including Omosudis, Evermannelli- dae, Scopelarchidae, and Paralepididae?including Anotopterus); and (4) Giganturoidei (Bathysauridae, Giganturidae, and Bathysauroides, a new genus erected for Bathysauropsis gigas [Kamohara]). Acknowledgments In addition to acknowledging Colin's remarkable contributions to science, we remember here his personal side, which has enriched both our lives. One of us (GDJ) has had the good fortune to work in close collaboration with Colin over the past several years, during which time we have become close friends. We have shared discover- ies about everything from fishes to music, lots of laughs, and an occasional pint or two. Cheers, Colin?here's to many more years of the same. For reviewing the manuscript, we thank A. G. Harold, R. K. Johnson, C. Patterson, R. H. Rosenblatt, and M. L. I. Stiassny. For loans or gifts of specimens, we thank B. Chernoff, M. F. Gomon, K. E. Hartel, T. Iwamoto, R. K. Johnson, J. A. Musick, M. Okiyama, T. Orrell, C. Patterson, J. R. Paxton, R. H. Rosenblatt, M. L. J. Stiassny, and K. J. Sulak. For use of and assistance with a Macintosh computer and help with preparation of the final character matrix, we are grateful to M. Lang. References Berg, L. S. (1940). Classification of fishes both recent and fossil. Trav. Inst. Zool. Acad. Sci. URSS 5(2), 346-517. Bertelsen, E., Krefft, G., and Marshall, N. B. (1976). The fishes of the family Notosudidae. Dana Rep. 86, 1-114. Bertmar, G. (1959). On the ontogeny of the chondral skull in Chara- cidae, with a discussion of the chondrocranial base and the visceral chondrocranium in fishes. Ada Zool. (Stockholm) 42, 151-162. Farris, J. S. (1970). Methods of computing Wagner trees. Syst. Zool. 19, 83-92. Farris, J. S. (1988). "Hennig86, Version 1.5," Program and docu- mentation. Farris, Port Jefferson Station, N.Y. Fink, W. L., and Weitzman, S. H. (1982). Relationships of the stomiiform fishes (Teleostei), with a description of Diplophos. Bull. Mus. Comp. Zool. 150(2), 31-93. Fujita, K. (1990). "The Caudal Skeleton of Teleostean Fishes." Tokai University Press, Tokyo. Gosline, W. A., Marshall, N. B? and Mead, G. W. (1966). Order Iniomi. Characters and synopsis of families. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 1-18. Yale University, New Haven, CT. Gottfried, M. D. (1989). Homology and terminology of higher teleost postdeithral elements. Trans. San Diego Soc. Nat. Hist. 21(18), 283-290. Hartel, K. E., and Stiassny, M. L. J. (1986). The identification of larval Parasudis (Teleostei, Chlorophthalmidae); with notes on the anatomy and relationships of aulopiform fishes. Breviora 487, 1-23. Johnson, G. D. (1992). Monophyly of the euteleostean clades? Neoteleostei, Eurypterygii, and Ctenosquamata. Copeia, 1992, 8-25. Johnson, G. D., and Patterson, C. (1993). Percomorph phylogeny: A survey of acanthomorphs and a new proposal. Bull. Mar. Sci. 52(1), 554-626. Johnson, G. D., Baldwin, C. C, Okiyama, M., and Tominaga, Y. (1996). Osteology and relationships of Pseudotrichonotus altivelis (Teleostei: Aulopiformes: Pseudotrichonotidae). Ichthyol. Res. 43, 17-45. Johnson, R. K. (1974a). A revision of the alepisauroid family Scopel- archidae (Pisces, Myctophiformes). Fieldiana, Zool. 66, 1-249. Johnson, R. K. (1974b). A Macristium larva from the Gulf of Mexico with additional evidence for the synonymy of Macristium with Bathysaurus (Myctophiformes: Bathysauridae). Copeia, 1974, 973-977. Johnson, R. K. (1982). Fishes of the families Evermannellidae and Scopelarchidae: Systematics, morphology, interrelationships, and zoogeography. Fieldiana, Zool. [N.S.] 12, 1-252. Johnson, R. K. (1984a). Scopelarchidae: Development and relation- ships. In "Ontogeny and Systematics of Fishes" (H. G. Moser, W. J. Richards, D. M. Cohen, M. P. Fahay, A. W. Kendall, Jr., and S. L. Richardson, eds.), Spec. Publ. No. 1, pp. 245-250. 402 CAROLE C. BALDWIN AND G. DAVID JOHNSON American Society of Ichthyologists and Herpetologists, Law- rence, KS. Johnson, R. K. (1984b). Evermannellidae: development and rela- tionships. In "Ontogeny and Systematics of Fishes" (H. G. Moser, W. J. Richards, D. M. Cohen, M. P. Fahay, A. W. Ken- dall, Jr., and S. L. Richardson, eds.), Spec. Publ. No. 1, pp. 250-254. American Society of Ichthyologists and Herpetologists, Lawrence, KS. Johnson, R. K. and Bertelsen, E. (1991). The fishes of the family Giganturidae: Systematics, development, distribution and as- pects of biology. Dana Rep. 91, 1-45. Kamohara, T. (1952). Revised descriptions of the offshore bottom- fishes of Prov. Tosa, Shikoku, Japan. Rep. Kochi Univ., Nat. Sci. No. 3, 1-122. Leviton, A. E., Gibbs, R. H., Jr., Heal, E., and Dawson, C. E. (1985). Standards in herpetology and ichthyology: Part I. Stan- dard symbolic codes for institutional resource collections in her- petology and ichthyology. Copeia, pp. 802-832. Maddison, W. P., andMaddison, D. R. (1992). "MacClade: Analysis of Phytogeny and Character Evolution. Version 3.04." Sinauer Assoc, Sunderland, MA. Marshall, N. B. (1954). "Aspects of Deep Sea Biology." Philosophi- cal Library, New York. Marshall, N. B. (1955). Alepisauroid fishes. 'Discovery' Rep. 27, 303-336. Marshall, N. B. (1960). Swimbladder structure of deep-sea fishes in relation to their systematics and biology. 'Discovery' Rep. 31, 1-122. Marshall, N. B. (1966). Family Scopelosauridae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 194-203. Yale University, New Haven, CT. Marshall, N. B., and Staiger, J.C. (1975). Aspects of the struc- ture, relationships, and biology of the deep-sea fish Ipnops murrayi Gunther (Family Bathypteroidae). Bull. Mar. Sci. 25, 101-111. McAllister, D. E. (1968). Evolution of branchiostegals and classifi- cation of teleostome fishes. Bull.?Natl. Mus. Can. 221, 1-237. Mead, G. W. (1966a). Family Aulopidae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 19-29. Yale University, New Haven, CT Mead, G. W. (1966b). Family Bathypteroidae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 114-146. Yale University, New Haven, CT. Mead, G. W. (1966c). Family Ipnopidae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 147-161. Yale University, New Haven, CT. Mead, G. W. (1966d). Family Chlorophthalmidae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 162-189. Yale University, New Haven, CT. Merrett, N. R., and Nielsen, J. G. (1987). A new genus and species of the family Ipnopidae (Pisces, Teleostei) from the eastern North Atlantic, with notes on its ecology. /. Fish Biol. 31, 451-464. Moser, H. G., Ahlstrom, E. H., and Paxton, J. R. (1984). Myctophi- dae: Development. In "Ontogeny and Systematics of Fishes" (H. G. Moser, W. J. Richards, D. M. Cohen, M. P. Fahay, A. W. Kendall, Jr., and S. L. Richardson, eds.), Spec. Publ. No. 1, pp. 218-239. American Society of Ichthyologists and Herpetologists, Lawrence, KS. Nelson, G. J. (1967). Epibranchial organs in lower teleostean fishes. /. Zool. 153, 71-89. Nixon, K. C. (1992). "Clados Version 1.2." Cornell University, Ithaca, NY. Okiyama, M. (1984a). Myctophiformes: Development. In "Ontog- eny and Systematics of Fishes" (H. G. Moser, W. J. Richards, D. M. Cohen, M. P. Fahay, A. W. Kendall, Jr., andS. L. Richard- son, eds.), Spec. Publ. No. 1, pp. 206-218. American Society of Ichthyologists and Herpetologists, Lawrence, KS. Okiyama, M. (1984b). Myctophiforms: relationships. In "Ontogeny and Systematics of Fishes" (H. G. Moser, W. J. Richards, D. M. Cohen, M. P. Fahay, A. W. Kendall, Jr., and S. L. Richardson, eds.), Spec. Publ. No. 1, pp. 254-259. American Society of Ich- thyologists and Herpetologists, Lawrence, KS. Olney, J. E., Johnson, G. D., and Baldwin, C.C. (1993). Phytogeny of lampridiform fishes. Bull. Mar. Sci. 52(1), 137-169. Parin, N. V., and Kotlyar, A. N. (1989). A new aulopodid species, Hime microps, from the eastern South Pacific, with comments on geographic variations of H. japonica. Jpn. }. Ichthyol. 35,407-413. Parr, A. E. (1928). Deepsea fishes of the Order Iniomi from the waters around the Bahama and Bermuda Islands. Bull. Bingham Oceanogr. Coll. 3(3), 1-193. Patterson, C, and Johnson, G. D. (1995). The intermuscular bones and ligaments of teleostean fishes. SmitJison. Contrib. Zool. 559, 1-83. Paxton, J. R. (1972). Osteology and relationships of the lanternfishes (Family Myctophidae). Bull. Nat. Hist. Mus. L.A. Cty., Sci. 13, 1-81. Post, A. 1987. Results of the research cruises of FRV "Walther Herwig" to South America. LXVII. Revision of the subfamily Paralepidinae (Pisces, Aulopiformes, Alepisauroidei, Paralepi- didae). I. Taxonomy, morphology and geographical distribu- tion. Arch. Fischereiwiss. 38, 75-131. Regan, C. T. (1903). On a collection of fishes from the Azores. Ann. Mag. Nat. Hist. [7] 12, 344-348. Regan, C. T. (1911). The anatomy and classification of the teleostean fishes of the Order Iniomi. Ann. Mag. Nat. Hist. [8] 7, 120-133. Regan, C. T. (1925). The fishes of the genus Gigantura, A. Brauer; based on specimens collected in the Atlantic by the "Dana" expeditions, 1920-22. Ann. Mag. Nat. Hist. [9] 15, 53-59. Rofen, R. R. (1966a). Family Paralepididae. In "Fishes of the West- ern North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 205-461. Yale University, New Haven, CT Rofen, R. R. (1966b). Family Omosudidae. In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 464-481. Yale University, New Haven, CT Rofen, R. R. (1966c). Family Anotopteridae. In "Fishes of the West- ern North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 5, pp. 498-510. Yale University, New Haven, CT. Rosen, D. E. (1971). The macristiidae, a ctenothrissiform family based on juvenile and larval scopelomorph fishes. Am. Mus. Novit. 2452, 1-22. Rosen, D. E. (1973). Interrelationships of higher euteleosteans. In "Interrelationships of Fishes" (P. H. Greenwood, R. S. Miles, and C. Patterson, eds.), pp. 397-513. Academic Press, London. Rosen, D. E. (1985). An essay on euteleostean classification. Am. Mus. Novit. 2827, 1-57. Stiassny, M. L. J. (1986). The limits and relationships of the acantho- morph teleosts. /. Zool. (B) 1, 411-460. Stiassny, M. L. J., and Moore, J. A. (1992). A review of the pelvic girdle of acanthomorph fishes, with comments on hypotheses of acanthomorph intrarelationships. Zool. }. Linn. Soc. 104, 209-242. Bulak, K. J. (1977). The systematics and biology of Bathypterois (Pisces: Chlorophthalmidae) with a revised classification of ben- thic myctophiform fishes. Galathea Rep. 14, 49-108 Swofford, D. L. (1991). "PAUP: Phylogenetic Analysis Using Parsi- mony, Version 3.0s." Computer program distributed by the Illinois Natural History Survey, Champaign. Swofford, D. L., andMaddison, W. P. (1987). Reconstructing ances- tral character states under Wagner parsimony. Math. Biosci. 87, 199-229. 14. Interrelationships of Aulopiformes 403 Tucker, D. W. (1954). Report on the fishes collected by S. Y. "Ro- saura" in the North and Central Atlantic, 1937-38. Part IV. Families Carcharhinidae, Torpedinidae, Rosauridae (nov.), Sal- monidae, Afepocephalidae, Searsidae, Clupeidae. Bull. Br. Mus. (Nat. Hist.) 2(6), 163-214. Walters, V. (1961). A contribution to the biology of the Giganturi- dae, with description of a new genus and species. Bull. Mus. Comp. Zool. 125, 297-319. Woods, L. P., and Sonoda, P. M. (1973). Order Berycomorphi (Beryciformes). In "Fishes of the Western North Atlantic," Sears Found. Mar. Res., Mem. No. 1, Part 6, pp. 263-396. Yale Univer- sity, New Haven, CT. Appendix 2 Material Examined Our analysis included examination of representa- tives of more than 40 neoteleostean genera listed be- low using institutional abbreviations specified by Lev- iton et al. (1985). Whole and cleared and stained specimens or parts of specimens (e.g., gill arches and paired fins) dissected from very large specimens were examined for most taxa. Cleared and stained lots are indicated by "cs." Appendix 1 Abbreviations Used in Text Figures AC Anterior Ceratohyal APC Autogenous Pelvic Cartilage BBn nth Basibranchial Br Branchiostegal CBn nth Ceratobranchial Cl Cleithrum Co Coracoid CPP Central Pelvic Process DH Dorsal Hypohyal Ecp Ectopterygoid EBn nth Epibranchial Enp Endopterygoid HBn nth Hypobranchial Hy Hyomandibular Ih Interhyal LPP Lateral Pelvic Process Me Mesethmoid Mep Metapterygoid MPP Medial Pelvic Process Mx Maxilla MxS Maxillary Saddle P Palatine Para Parasphenoid PBn nth Pharyngobranchial Pc Postcleithrum PC Posterior Ceratohyal Pmx Premaxilla PPC Posterior Pelvic Cartilage PPP Posterior Pelvic Process PR Pectoral-fin Radial Q Quadrate Sea Scapula Sc Supracleithrum Sy Symplectic UP Uncinate Process UPn nth Upper Pharyngeal Toothplate V Vomer VH Ventral Hypohyal Aulopiformes?Ahliesaurus berryi: USNM 240503, 240505 (cs). Alepisaurus brevirostris USNM 200817 (gill arches, pelvic fin cs), 201275. Alepisaurus sp.: MCZ 60345 (cs). Anotopterus pharao: CAS 164180 (cs); SIO 5553 (cs); USNM 140825 (cs), 201286, 221035, 221035 (cs), 206844; SIO 62-775 (cs). Arctozenus rissoi USNM 302410 (1 cs), 283485 (cs). Aulopus filamentosus: USNM 292105 (cs), 301018. Aulopus japonicus: AMNH 28635SW (cs); FMNH 71831 (cs). Aulopus sp.: AMNH 28635 (cs). Bathymicrops regis: BMNH 1989.7.25.56.61 (cs). Bathypterois longipes USNM 35635. Bathypterois pectinatus: FMNH 88982 (cs). Bathypterois sp. MCZ 40567 (cs). Bathypterois viridensis USNM 117215. Bathy- sauropsis gracilis AMS IA6934 (cs): NMV A6932. Bathy- sauropsis malayanus USNM 098888 (holotype of Bathy- saurops malayanus). Bathysaurus ferox AMS 1.29591001; MCZ 62409 (cs); USNM 316825. Bathysaurus mollis: VIMS 6107 (cs). Bathysauroides gigas: AMS I, 22822001 (cs); NMV A5770, A4438, A4440 (cs). Bathytyphlops marionae USNM 336666 (cs), 336713 (formerly VIMS 06104); 341861 (gill arches cs). Benthalbella dentata: SIO 63-379 (cs). Benthalbellaelongata USNM207279. Benthal- bella macropinna USC E1671. Chlorophthalmus agassizi: AMNH 40829SW (cs); USNM 159385 (cs), 302386. Chlorophthalmus atlanticus USNM 339774 (1 cs). Cocco- rella atlantica: USNM 235170, 235189 (cs), 235199 (cs). Evermannella balbo USNM 301265. Evermannella indica: U.H. 71-3-9 (cs); USNM 235141. Gigantura chuni AMNH 55345SW (cs); MCZ 60324 (cs). Gigantura indica MCZ 54133 (cs): SIO 76-9; USNM 215407. Harpadon nehereus: AMNH 17563 (cs); FMNH 179018 (cs); USNM 308838. Harpadon squamosus: FMNH 80823 (cs). Ipnops agassizi: USNM 54618 (gill arches cs). Ipnops meadi SIO 61-175 (cs). Ipnops murrayi USNM 101371, 336711 (for- merly VIMS 6736), 336712 (formerly VIMS 6737). Les- tidiops affinis MCZ 60632 (cs). Lestidiops sp.: USNM 307290 (cs). Lestidium atlanticum: USNM 201183 (cs), uncat. (cs). Lestidium sp.: USNM 341877 (1 cs). Lestro- lepis intermedia USNM 290253 (2 cs). Lestrolepis sp. USNM 307290 (1 cs). Macroparalepis affine: USNM 302410 (cs); 201184 (cs). Macroparalepis sp.: FMNH 404 CAROLE C. BALDWIN AND G. DAVID JOHNSON 49988 (cs); USNM 201186 (cs). Odontostomops normal- ops: USNM 235029 (cs), 274377 (1 cs). Omosudis lowei USNM 219982 (cs), 206838, 287310. Paralepis breviros- tris: USNM 196109 (cs). Paralepis coregonoid.es: USNM 196098,290253 (cs). Parasudis truculentus: FMNH 67150 (cs); MCZ 62398 (cs); USNM 159096 (1 cs), 159407 (cs), 159850 (cs). Pseudotrichonotus altivelis: USNM 280366 (cs); ZUMT55678 (cs), 59882 (cs). Rosenblattichthys hub- bsi MCZ 52821 (cs). Saurida brasiliensis: USNM 185852 (cs); 187994 (cs). Saurida gracilis: USNM 256409 (cs). Saurida normani: USNM 341878 (cs). Saurida parri: USNM 193763 (cs), 340398 Saurida undosquamous: USNM 325180 (cs). Scopelarchus analis: MCZ 62599 (cs); USNM 234988 (cs). Scopelarchoides nicholsi: USNM 201154 (cs), 207295. Scopelarchoides signifer: USNM 274385 (cs). Scopelosaurus argenteus MCZ 63321 (cs), 62105 (cs), 62405 (cs). Scopelosaurus fedorovi SIO 60-251 (cs). Scopelosaurus hoedti: USNM 264256 (2 cs). Syno- dontidae: USNM 309851 (cs). Stemonosudis rothschildi AMS I. 22826001 (cs). Stemonosudis sp. USNM 330273 (cs). Sudis atrox MCZ 60336 (cs); USNM 330285 (cs). Sudis hyalina USNM 340399 Synodus jenkensi: USNM 321745 (1 cs). Synodus synodus: USNM 318960 (1 cs). Synodus variegatus: USNM 140825 (cs); 315318 (cs). Trachinocephalus myops: FMNH 45392 (cs); MCZ 62106 (cs); USNM 305292, 185861 (cs); 339775 (larva, cs); 339776 (cs). Uncisudis advena MCZ 68531 (cs). Stomi- iformes?Diplophos taenia: MCZ 55469 (cs); USNM 206614 (cs), 274404. Myctophiformes?Lampanyctus cuprarius USNM 300490 (cs). Myctophum obtusirostre: AMNH 29140SW (cs). Neoscopelus macrolepidotus: USNM 188056 (cs); 317160 (cs). Neoscopelus sp. USNM 159417 (cs). Notoscopelus resplendens: AMNH 25928SW (cs). Lampridiformes?Metavelifer: BPBM 23953 (cs). Polymixiiformes?Polymixia lowei USNM 137750, 185204 (cs), 308378 (cs).