Measurements and Terminology

Measurements conducted on specimens were taken in a straight line, point-to-point, modified from Bigelow and Schroeder (1953), Rosa (1985), and Miyake (1988), and are as follows: TL, total length (snout tip to posterior tip of tail); DL, disc length (from anterior disc margin at snout region to posterior disc margin at disc axil); DW, disc width (distance between lateral disc margins at greatest width, i.e., at level of scapulocoracoid; width may be underestimated in some specimens because of broken distal tips of pectoral radials); TAL, tail length (from puboischiadic bar to posterior tip of tail); TAW, tail width (greatest width of tail as exposed, just posterior to pelvic fins); POL, preoral snout length (snout length from mouth opening to anterior margin of disc); MSC, distance between mouth and scapulocoracoid (from mouth opening to anterior margin of scapulocoracoid); SCW, scapular width (greatest width of scapulocoracoid, not including pectoral basal cartilages); PGW, pelvic girdle width (greatest width of pelvic girdle); STL, sting length (from sting base to posterior sting tip); PGT, distance between pelvics and tip of tail; CL, clasper length (from posterior margin of pelvic girdle and posterior tip of clasper); NCL, neurocranial length (from anterior contour of neurocranium at nasal capsule confluence to occipital region); NCW, neurocranial width (greatest width at level of nasal capsules). The following meristic counts were taken: PRO, propterygial radials; MES, mesopterygial radials; MET, metapterygial radials; TPR, total pectoral radials (= PRO + MES + MET); PVR, pelvic radials; VSP, vertebral centra from scapulacoracoid to pelvic girdle; VPS, vertebral centra from pelvic girdle to caudal sting; VST, vertebral centra posterior to sting; TV, total vertebral centra (= VSP + VPS + VST). The number of toothrows could not be inferred with accuracy, as teeth are not preserved in discrete rows but are scattered in the vicinity of the jaws and neurocranium. Many measurements traditionally taken on extant specimens could not be made on the fossils due to preservational imperfections (e.g., measurements involving the cloaca). The measurements above were taken to provide a general notion of size and proportions of the fossils, but because their accuracy depends on the preservation of the material, which varies significantly, they should not be accepted as strictly as if taken from extant specimens; the counts provided should be interpreted in a similar fashion. Measurements and counts are summarized in tables 2 and 3 for the new stingray taxon and table 4 for †Heliobatis.

Anatomical terminology is according to Daniel (1934), Miyake (1988), and Miyake and McEachran (1991) unless otherwise noted. We follow Rosa (1985) in using “sting” (as opposed to “spine”) to refer to the elongated and serrated dermal derivative on the dorsal surface of the tail (usually more than one sting is present, and sometimes more than two). “Sting” and “serrated caudal sting” are used interchangeably. The term “stingray” is used here as equivalent to “myliobatiform rays”. There are many different groups of myliobatiform rays, and most of these have a corresponding trivial name (e.g., eagle, cownose, whiptail, manta rays). “Stingray” is an all-encompassing term used for the entire order (Myliobatiformes) and does not include here platyrhinids (thornback rays) and Zanobatus, considered to be successive sister-groups to stingrays by McEachran et al. (1996). In the descriptive sections below the composition of the different myliobatiform families is based on Compagno (1973, 1977, 1999) and on the phylogenetic analysis presented after the descriptive accounts. The terms “benthic stingrays”, “non-myliobatid stingrays”, or “nonpelagic stingrays” apply to Recent and/ or fossil stingrays that have a mostly benthic niche, excluding the morphologically specialized pelagic stingrays (which are here lumped into a single family, Myliobatidae; the exception is the monotypic Pteroplatytrygon, which is a “dasyatid” but is also pelagic). Benthic stingrays include Hexatrygon, gymnurids, urolophids, Plesiobatis, urotrygonids, potamotrygonids, almost all dasyatids, and the fossil taxa from the Green River and Monte Bolca Formations (except †Promyliobatis, a myliobatid from Monte Bolca). Consequently, we do not refer to a monophyletic group when making reference to “benthic stingrays”, because they are not monophyletic without myliobatids (see our phylogenetic analysis below); these terms are used only for convenience. Synonymies are in abbreviated form, with the full citation included only in the references. The synonymy for †Heliobatis radians is not exhaustive.

The procedures employed in the phylogenetic analyses are described in Phylogenetic Procedures and Coding, after the morphological characterization of †Heliobatis.

Comparative Stingray Material

Recent stingray materials examined for this study are listed below, and include cleared-and-double-stained specimens (C&S), dry skeletons (SD), skeletons preserved in alcohol (SW), X-ray radiographs (XR), and dissected specimens (D). Dry and “wet” skeletons were prepared by subjecting dissected specimens to dermestid beetles (prepared by the staff of the Department of Ichthyology, AMNH). Cleared-and-stained specimens were prepared according to Dingerkus and Uhler (1977). Additional specimens of many stingray taxa were examined for external morphology, but only a few of these are included in the list below (belonging to species in which available specimens are more rare). Anatomical information extracted from the literature was confirmed by examining specimens whenever possible. Stingray classification in this list follows Compagno (1999).

Abbreviations

CLASS CHONDRICHTHYES HUXLEY, 1880

SUBCLASS ELASMOBRANCHII BONAPARTE, 1832

DIVISION SQUALEA sensu SHIRAI, 1992

SUPERORDER HYPNOSQUALEA CARVALHO AND MAISEY, 1996

SERIES BATOIDEA sensu COMPAGNO, 1973

ORDER MYLIOBATIFORMES sensu COMPAGNO, 1973

SUBORDER MYLIOBATOIDEI insertae sedis

†Asterotrygon, new genus

†Asterotrygon maloneyi, new species Figures 2–13, 18–26a, c, d, 27; tables 2, 3

Neurocranium

The neurocranium (n) is preserved at least partially in almost all specimens, but FMNH PF 12989 (paratype; fig. 4) was particularly informative. Specimens FMNH PF 15166 (holotype; fig. 2), FMNH PF 15180 (fig. 12), AMNH P 11557 (paratype; fig. 3), FMNH PF 14069 (paratype; fig. 5), FMNH PF 12914 (fig. 6), FMNH PF 14567 (fig. 7), FMNH PF 14098 (fig. 8), USNM 2028 (fig. 10B), and NHM P 61244 (fig. 11A) also provided structural details. In dorsal view, the neurocranium of †Asterotrygon has an outline generally indistinct from extant nonmyliobatid stingrays such as potamotrygonids, Trygonoptera, urotrygonids, and most dasyatids. The neurocranium is widest anteriorly at the level of the nasal capsules, where it is almost in contact with the propterygium on both sides. A keyhole-shaped dorsal fontanelle is present and can be seen in most specimens, even those that are exposed ventrally (e.g., fig. 4). Neurocranial features that can be observed are those that are prominent dorsoventrally. Prismatic calcification covers almost all of the neurocranium, including the nasal capsule area, and some specimens are heavily calcified (fig. 7). The intense shagreen of small to medium-sized denticles, covering much of the neurocranium, contributes to the difficulty of observing certain features, such as the shape of the postorbital processes in the male paratype (fig. 4).

A rostral cartilage (or rostrum; ro) is not present in our adult fossil specimens, and †Asterotrygon lacks any calcified neurocranial structure anterior to the nasal capsules, as do adult stingrays in general. This contrasts with the condition observed in embryonic or juvenile specimens of various extant stingray taxa (Urobatis, Urotrygon, Urolophus, Trygonoptera, Potamotrygon, Plesiotrygon), in which a small, detached subtriangular rostral extremity (not the rostral appendix of Miyake et al., 1992b) is present between the pectoral fin extremes (Holmgren, 1940; Rosa, 1985; Miyake, 1988; Miyake et al., 1992b; McEachran et al., 1996). The subtriangular rostral extremity of stingrays forms in connection with the anterior medial outgrouth of the trabecula; it is carried forward by the expanding neurocranium in conjunction with pectoral fin extension, but appears to be absent in pelagic stingrays (Mobula kuhlii [AMNH 15319]; however, more ontogenetic data are needed for pelagic stingrays). The rostral extremity is also absent in most small specimens of †Asterotrygon (e.g., fig. 11D), which is not surprising given that in extant stingrays it is a hyaline, uncalcified structure; its absence in †Asterotrygon may therefore be preservational. This conclusion is supported by specimen NHM P 61244 (40 mm in DW and DL, 86 mm in TL; fig. 11A), in which a small, wedgelike structure (re?) occurs between both anterior pectoral fin extremes; this structure is partially covered by denticles and may represent an impression of the rostrum. In juvenile stingray specimens in which it is present, the rostral extremity is usually not posteriorly elongated. This may be only an artifact of preparation, however, as the rostral extremity of Urolophus cruciatus (AMNH 98247) extends farther posteriorly, comprising about one-third of the distance between the anterior disc margin and the anterior neurocranial contour. Whether the rostral extremity continues all the way to the anterior neurocranial margin in this specimen, perhaps as a very slender and uncalcified rod, is not clear. In addition, stingrays are reported to have a vestigial medial outgrowth of the trabecula at the junction of the nasal capsules (Holmgren, 1940; Miyake et al., 1992b), which eventually disappears in fully formed individuals (there is no indication of this structure in Urolophus cruciatus [AMNH 98247]). This outgrowth, supposedly the rudimentary rostral base, is lacking in Gymnura (fig. 31), Taeniura lymma (fig. 32), Urotrygon chilensis (fig. 33), and potamotrygonids (fig. 34), but may be vestigially present in Urobatis halleri (fig. 38A, ro?); its arrangement and distribution among stingrays is yet to be determined. Among †Asterotrygon specimens, only FMNH PF 15180 appears to have a small, triangular outgrowth at the anterior margin of the neurocranium between the nasal capsules (fig. 20). However, this structure may be part of the nasal capsules, the rostral base per se, or the rostral extremity that has not been dislocated anteriorly with the anterior extension of the pectoral fin.

The nasal capsules (nc) in †Asterotrygon are wider than long and are oval in general outline. There is a slight notch or concavity anteriorly in between both nasal capsules, where they become confluent. This is more apparent in certain specimens (e.g., fig. 18), including the holotype, where the anterior cranial contour is well preserved, but is only faintly visible in paratype AMNH P 11557. It is also not possible to precisely determine the extent of the dorsal internasal septum in most specimens, but FMNH PF 12989 clearly has a slender prismatically calcified area between the closely set olfactory canals (oc), somewhat similar to Urotrygon venezuelae (Miyake, 1988: 163, fig. 35a). The dorsal internasal septum, considered to be the anterior segment of the trabecular cartilage (El-Toubi and Hamdy, 1959; Miyake et al., 1992b), separates the nasal sacs and capsules; its extent is determined by the trabecular contribution, nasal aperture size and position relative to each other, the arrangement of the olfactory canals, and by the forward and lateral expansions of the central nucleus and lateral pallium of the telencephalon (Northcutt, 1978). The olfactory canals in FMNH PF 12989 are visible through the precerebral fontanelle (seen in dorsoventral view), and the dorsal internasal septum appears to be relatively slender (fig. 18). In the holotype, however, the internasal septum is wider, with a greater distance between the olfactory canals. The width of the dorsal internasal septum varies among stingray taxa, being relatively wide in myliobatids, Hexatrygon, Plesiobatis, Trygonoptera, Urobatis, and some species of Himantura and Dasyatis (according to material examined). Species of potamotrygonids also vary in this character, and it is difficult at present to precisely quantify the width of the dorsal internasal septum in stingrays (more comparisons are presently being undertaken). The precerebral fossa (a perforation of the anterior cranial or ethmoidal wall) was presumably truncated, as in modern stingrays and batoids in which the rostral cartilages are reduced (Miyake, 1988). As it has not been excavated and is partially crushed, the depth of the precerebral fossa is not known in our fossils. Nasal cartilages (ncr) are not preserved in †Asterotrygon, even in FMNH PF 15180, a specimen with preserved olfactory canals (fig. 20).

The neurocranium is markedly curved anterolaterally at the level of the nasal capsules, and appears to abut the internal surface of the propterygium. A preorbital process (prp) was present on both sides of the neurocranium, as in extant stingrays. In Recent benthic stingrays the preorbital process is triangular, directed posterolaterally and situated behind the nasal capsules on the dorsal surface of the neurocranium. In our fossils, however, this region is obscured by massive jaw cartilages and perhaps also by the ventrally situated antorbital cartilages. The antorbital cartilages (aoc) are almost ventral topological counterparts of the dorsal preorbital processes in many extant and fossil stingrays. They are situated ventrolaterally on the nasal capsule just anterior to the level of the preorbital processes, and are generally more massive than the preorbital processes. Because fossil stingrays are dorsoventrally preserved, structures that are topologically on the same dorsoventral plane will usually be obliterated. The antorbital cartilages are subtriangular structures, usually curved and tapering caudally from the ventral posterolateral surface of the nasal capsule with which they articulate. They are anteroposteriorly flattened in stingrays. In FMNH PF 12989 the antorbitals are about one-fifth as long as the neurocranium (fig. 18), reaching to midneurocranial length, and are therefore proportionally larger than in many extant urolophid and potamotrygonid species. There is some variation in antorbital cartilage shape, as in FMNH PF 14567; the left antorbital, which remains articulated, is triangular but straighter than in other fossils (fig. 19; this structure is identified as the antorbital, as it appears to be articulated and is too large to be the angular cartilage, judging from the little space available for the hyomandibular-Meckelian ligament in this specimen). In many specimens the antorbitals remain articulated in their original position, but in others they are displaced to fill the space between the orbits and internal margins of the propterygia. The left antorbital cartilage is partially preserved in paratype AMNH P 11557, and is relatively smaller than in FMNH PF 12989, appearing to taper less while being more blunt at its posterior tip. The preorbital canal (poc), through which pass branches of the trigeminal (V) and other nerves from the orbit to the ethmoid region, is not visible in specimens of †Asterotrygon, even though it is enlarged in Recent stingrays anterior to the preorbital process on the dorsal surface of the nasal capsule (figs. 31–34).

A single elongated dorsal fontanelle, representing the unchondrified anterior roof of the cranial cavity (de Beer, 1937), is present in stingrays and occupies over two-thirds of neurocranial length. However, some modern stingray taxa exhibit a median constriction, representing a remnant of the epiphysial bar (epb), partially separating an anterior precerebral fontanelle (pcf) from a posterior frontoparietal fontanelle (fpf) (Miyake, 1988). The epiphysial bar is more intact in juvenile stingray specimens, but is especially well developed in Potamotrygon, where it can be almost complete (numerous specimens; fig. 34A); it appears as an anteromedian constriction in the dorsal fontanelle in Taeniura lymma (separating the pcf and fpf in fig. 32A; also Garman, 1913: pl. 71, fig. 6). Two dorsally exposed (FMNH PF 14098 and USNM 2028) and one ventrally exposed specimen (FMNH PF 12989) support the presence of a median constriction in †Asterotrygon, and the dorsal fontanelle is faintly divided into anterior (precerebral fontanelle) and posterior (frontoparietal fontanelle) regions, as in species of Urolophus, potamotrygonids, Pastinachus sephen, Pteroplatytrygon violacea, Plesiobatis daviesi, and Taeniura lymma. Most ventrally exposed †Asterotrygon specimens (the majority of specimens, including the holotype) display a single unconstricted dorsal fontanelle (precerebral fontanelle + frontoparietal fontanelle), but this may be an artifact of preservation. The fontanelle is relatively wide in the holotype, where its internal borders are well delineated posteriorly. The posterior portion of the dorsal fontanelle displays no constriction in the fossils, and therefore the tectum orbitale is not visible. However, some stingrays may retain a posterior constriction as well, faintly dividing the dorsal fontanelle farther posteriorly (Meng, 1984). The frontoparietal fontanelle tapers posteriorly, ending in an oval curve, which is clearly visible in many †Asterotrygon specimens, including the holotype.

The orbital region of †Asterotrygon is longer than wide and occupies less than two-thirds of neurocranial length, being widest anteriorly at the level of the preorbital processes. The orbits are therefore relatively elongate, but this does not mean that the eye diameter was also large. The eyes in †Asterotrygon may have been large and bulging, as in some Recent stingray genera (Potamotrygon, Dasyatis, Taeniura, Urobatis, Urolophus), but orbit length does not necessarily correlate with eye diameter (Plesiotrygon has orbital proportions similar to Potamotrygon, even though it has significantly smaller eyes in comparison). The orbits are slightly concave in dorsal perspective. It is difficult to determine the width and extent of the supraorbital shelf in our two-dimensional fossils, but it appears to have been confluent anterioly with the preorbital process, which is largely obscured. The orbital region of †Asterotrygon was arranged as in most benthic stingrays, and not shortened as in gymnurids (fig. 31) and myliobatids.

Whether the supraorbital shelf was continuous with the postorbital process (pop), or if a notch separated both structures, is of some systematic significance. Species of Urolophus, Plesiobatis daviesi, Pteroplatytrygon, and pelagic stingrays (with the exception of Myliobatis and possibly Aetomylaeus) have the supraorbital crest continuous with the postorbital process, while dasyatids (except Pteroplatytrygon), potamotrygonids, Urobatis and Urotrygon, Trygonoptera, and Hexatrygon present a groove or notch between both structures. The postorbital groove allows for the passage of the infraorbital lateral line canal. Accordingly, the anterior margin of the postorbital process is at an oblique angle to the supraorbital crest in the taxa presenting a groove. In Urolophus, Pteroplatytrygon, Plesiobatis, and most pelagic stingrays, however, the canal passes through a dorsoventral opening in the postorbital process (Miyake, 1988), which also has a more transversely directed anterior profile. In †Asterotrygon, enlarged nuchal, orbital, and/or spiracular dermal denticles, along with the prominent hyomandibulae, obscure the postorbital process (e.g., in the holotype and in FMNH PF 12989; fig. 18), which is best preserved in FMNH PF 14567. The postorbital process of this specimen is slightly directed anteriorly, forming an angle in relation to the long axis of the neurocranium. In FMNH PF 15180 the anterior margin of the left postorbital process is preserved, which is obliquely oriented, reaching to about one-third of the length of the hyomandibula (situated directly behind it). A small triangular protuberance (the equivalent of the supraorbital process, sp; fig. 17A) is apparent anterior to the postorbital process of FMNH PF 14567 (fig. 19), separated from it by a slight indentation of the supraorbital crest. (Holmgren [1942: 179] referred to this protuberance as the “supraorbital process”, a term we adopt, and used it to strengthen his presently unsupported view of the close relationship between certain petalichthyid placoderms, such as †Macropetalichthys, and rays; cf. Jarvik, 1980: 376–378). The supraorbital process is visible on both sides of the neurocranium, and †Asterotrygon therefore presents the condition seen in dasyatids, potamotrygonids, Trygonoptera, and urotrygonids among nonmyliobatid stingrays. The postorbital process is broad and shelflike in †Asterotrygon, as in all stingrays. Foramina through the supraorbital crest for rami of the superficial ophthalmic nerve are not discernible in any specimen of †Asterotrygon.

The parietal fossa (paf) in stingrays may contain both the endo- and perilymphatic formina, or only a single pair of endolymphatic openings, and it comprises the posterior one-fifth of the dorsal surface of the neurocranium. In †Asterotrygon the fossa is filled with matrix, and is similar to Recent stingrays in proportions, but whether both pairs of foramina are confluent or separate cannot be determined. A crest outlines the fossa along its margins. The hyomandibular facet is located laterally on the ventral aspect of the neurocranium, more or less between the frontoparietal fontanelle and parietal fossa, and posterior to the postorbital processes. In Recent stingrays there is also an articular facet for the anteroposteriorly compressed dorsal pseudohyoid element posterior to the hyomandibular articulation, but this cannot be observed in the fossils. The otic capsules at the posterior corners of the neurocranium are tumid, but along with the occipital condyles they are largely obscured or crushed in †Asterotrygon by elements of the pseudohyoid arch, gill arches, and possibly synarcual cartilage, and little anatomical detail remains. As in all stingrays, a jugal arch is absent from the posterolateral corners of the neurocranium.

Jaws, Hyomandibulae and Visceral Arches

The jaws are well preserved in the female paratypes (fig. 21), but especially in the juvenile specimen FMNH PF 15180 (all of which are ventrally exposed; fig. 20). The jaws are dislocated, obscured, or even missing in other specimens. The jaws of †Asterotrygon are robust, massive structures as in most Recent stingrays (not slender as in Paratrygon and gymnurids), extending laterally to occupy almost the entire space between both propterygia. Both sets of jaws are strongly reinforced with prismatic calcification, and both upper and lower jaws are composed of mirror-image antimeres that meet mesially. The upper jaws (palatoquadrates; pq) are slightly thicker than the lower jaws (mandibular or Meckel's cartilages; Mc) in AMNH P 11557, but not so in FMNH PF 14069 or FMNH PF 15180, and this may be due to differences in the orientation of the jaws when preserved. The mandibular cartilages are more arched than the palatoquadrates and bear a widely rounded posterior margin (figs. 20, 21). Both antimeres of the mandibular cartilages and palatoquadrates are not fused symphysially (as in certain myliobatid stingrays), and are separated by a small space that originally contained strong horizontal ligaments attached to both antimeres. The arched appearance of the jaws varies among individuals and is also probably strongly correlated with jaw position when mineralized. Both mandibular and palatoquadrate cartilages taper slightly toward the midline, and both have pronounced outer corners that are curved in most specimens. The upper surface of the mandibular cartilage is relatively straight in AMNH P 11557 and FMNH PF 14069, but in FMNH PF 15180 there is a circular opening between the lower surface of the palatoquadrate and the upper aspect of the lower jaw, as in most Recent nonmyliobatid stingrays. The palatoquadrate resembles the condition in potamotrygonids, Dasyatis, Himantura, Taeniura, and urolophids in being relatively straight on its dorsal flange. The mandibular cartilage contains a strong concavity at its laterointernal corner in which it articulates with the palatoquadrate in extant nonmyliobatid stingrays. There is also a pronounced curvature lateroexternally for the stout hyomandibular ligament. Both of these curved surfaces appear to be present in our fossils, along with the large fossa for the adductor-mandibulae musculature, but because of overlying structures and partial crushing of the jaws, the degree of development of these surfaces and the exact nature of the palatoquadrate-mandibular cartiage articulation cannot be further discerned, even though they appear to resemble that of extant stingrays (Dasyatis, Potamotrygon).

The ventrolateral processes of the mandibular cartilage are absent in all specimens of †Asterotrygon except FMNH PF 15180. The ventrolateral process is a triangular, posterior projection from both outer corners of the mandibular cartilage, and it is present in Taeniura, potamotrygonids, and many other nonmyliobatid stingrays. Its precise distribution and importance as a systematic character are still to be determined. The ventrolateral process of FMNH PF 15180 is more apparent on the right side, projecting past the palatoquadrate. This feature, along with any superficial feature of the external aspect of the lower jaw, was probably obliterated in other specimens. The winglike process (a stout, lateral or posterolateral projection of the lower jaw) appears to be restricted to pelagic stingrays (Nishida, 1990) and is clearly absent in †Asterotrygon.

The hyomandibulae of stingrays are elongate cartilages articulating with the braincase at its posterolateral surfaces, from which they extend anteroventrally to the level of the outer jaw corners. In †Asterotrygon, the hyomandibula (hyo) is strongly calcified and therefore relatively well preserved in most specimens. It projects not only anteroventrally but also slightly laterally, reaching the internal wall of the propterygium just posterior to the lower jaw corner, and almost reaching the jaw joint. The hyomandibula is stout when observed from a lateral perspective in extant stingrays, but it is more slender when observed in dorsoventral view. In †Asterotrygon, however, the hyomandibulae are relatively stout, especially in the holotype (where they are not very elongated), even though they are dorsoventrally preserved (the hyomandibulae are also fairly wide in the paratypes and in FMNH PF 12914 and PF 14567; fig. 19). The hyomandibula is generally more slender in smaller specimens (NHM 61244, FMNH PF 14069), but FMNH PF 15180 is an exception (fig. 20), as it presents relatively wide hyomandibulae that do not taper from their articulation toward the jaws (this may be due to its orientation when preserved). In most specimens, the hyomandibula is widest close to its attachment to the ventral otic region of the neurocranium (from a dorsoventral perspective), as in most extant benthic stingrays. The internal surface of the hyomandibula is slightly concave, especially close to the jaw corners. The proximal portion of the hyomandibula is also partially obscured by enlarged nuchal denticles in some specimens (e.g., fig. 18). The hyomandibula is devoid of rays, as in extant stingrays. Pre- and postspiracular cartilages and “hyomandibular accessory cartilages” (Nishida, 1990) are not preserved or cannot be identified in †Asterotrygon.

In stingrays, the articulation between the hyomandibular cartilage and the lower jaw can be either through a strong, stout hyomandibular-Meckelian ligament (Taeniura, Dasyatis, Urolophus, Trygonoptera, Himantura), through the ligament reinforced by separate cartilages formed within it (angular cartilages of Potamotrygon and Plesiotrygon; Garman, 1913), or through a more direct connection, without a stout ligament between the hyomandibulae and Meckelian cartilages (Hexatrygon). A very small and inconspicuous ligament is present in pelagic stingrays and gymnurids, and this is considered to be a more direct connection as in Hexatrygon. In †Asterotrygon, there is a small, slightly transverse element situated between the hyomandibula and lower jaw (more conspicuous in FMNH PF 15166, AMNH P 11557, FMNH PF 15180, FMNH PF 12989, NHM 61244). This element does not represent the recurved, medially directed distal tip of the hyomandibula. The distal (anterior) section of the hyomandibula can be oriented in this fashion even though the stout hyomandibular-Meckelian ligament is present (e.g., Urobatis halleri [fig. 38A], Urotrygon chilensis [fig. 33], Himantura krempfi [AMNH 41567]). This inflection of the hyomandibula, however, is usually small and does not extend to contact the lower jaw. The transverse element of †Asterotrygon contacts the lower jaw and is heavily covered by prismatic calcification (especially in FMNH PF 15180; however, the individual prisms are mostly missing or scattered, and usually only their impressions remain). This prismatic calcification supports the notion that a separate cartilage, similar to the angular cartilages (ac) of potamotrygonids, was present in the ligament uniting the hyomandibula and lower jaw in †Asterotrygon (best observed in FMNH PF 15180 and in NHM 61244, both juvenile specimens). Holmgren (1943: 74) considered the angular cartilages (his “mandibular rays”) to be of mandibular arch origin.

Even though the hyomandibular-Meckelian ligament may become strengthened, slightly chondrified, and even reinforced with scattered internal calcification, actual prismatic calcification is absent from the ligament in specimens presenting these conditions (e.g., Taeniura lymma [fig. 32], Trygonoptera testacea [fig. 40B]). This condition may also vary intraspecifically (such as in Trygonoptera testacea) and does not appear in most radiographs. The angular cartilages of Potamotrygon and Plesiotrygon, however, are prismatically calcified and clearly visible in radiographs (fig. 40). These cartilages are stout, oval to rectangular in shape, and embedded within the strong hyomandibular ligament. As such, they are unique among Recent stingray taxa to potamotrygonids except Paratrygon according to our specimens (see also Garman, 1913; cf. Lovejoy, 1996). Himantura schmardae (NHM 1908.5.28:2–3) has very small calcified elements that are barely visible in radiographs (Lovejoy, 1996); they may be easily overlooked in dissections. The distal (anterior) extremity of the hyomandibula of potamotrygonids is not inflected medially to a large extent, where the hyomandibular-Meckelian ligament is present. Although somewhat tentative because of preservational issues, we conclude that the condition in †Asterotrygon resembles the discrete angular cartilages of Potamotrygon and Plesiotrygon, because these appear to be prismatically calcified, discrete cartilages. In turn, this suggests that the hyomandibular-Meckelian ligament was present in †Asterotrygon as well, even though impressions of the ligament are not clearly preserved in the fossils (which is to be expected). In FMNH PF 15180, the putative angulars appear to be composed of a single, relatively elongate element on each side, slightly directed posteriorly. A small separate cartilage located along the internal margin of the hyomandibula, and posterior to the attachment of the hyomandibular ligament, is present in numerous species of Urolophus (observed in many specimens) and pelagic stingrays (Garman, 1913). This cartilage (termed here the secondary hyomandibular cartilage), however, is considered functionally isolated and not homologous to the angular cartilages of potamotrygonids (Lovejoy, 1996; McEachran et al., 1996), but this remains to be elucidated. No such element is visible in specimens of †Asterotrygon.

Stingrays typically possess five gill arches (although there are six in Hexatrygon) in addition to the pseudohyoid arch located between the hyomandibular and first gill arch. Elements of the ventral gill arch skeleton have been studied in more detail (Miyake, 1988; Miyake and McEachran, 1991), and stingrays characteristically possess an enlarged medial plate (comprising posteriorly the basibranchial copula, considered to be the result of ontogenetic fusion of hypobranchial and basibranchial components; bb), ankylosis or slight overlapping of the proximal segments of the last two ceratobranchials (unconfirmed in Hexatrygon), anteriorly and obliquely directed short, separate hypobranchial 1 elements (hb1), and a small transverse and separate basihyal cartilage (bh). Additionally, the ceratobranchial and pseudohyoid cartilages are fused in stingrays (the number of fused ceratobranchial elements varies among taxa). Most pelagic stingrays lack the first hypobranchial along with the basihyal (Nishida, 1990), but whether this is primitive or secondary depends on the myliobatiform phylogeny adopted (derived in our phylogenetic study below; cf. appendix 2). Primitively in stingrays, the last pair of ceratobranchial and epibranchial elements articulate with the anteromedial aspect of the coracoid bar, and the basihyal and the medial plate are separated by a large gap (Nelson, 1969).

The pseudohyoid and gill arches are poorly preserved in †Asterotrygon. The most informative specimen is FMNH PF 15180 (figs. 12, 20), which has certain elements of the ventral gill skeleton intact. Five pairs of gill arches are present in addition to the pseudohyoid arch (also observed in FMNH PF 14567 and FMNH PF 12914). The basihyal cartilage (bh) in †Asterotrygon is a very small and separate element, slightly wider than long, and similar to the basihyal of Urobatis jamaicensis (AMNH 30385), U. halleri (fig. 38B), and Dasyatis americana (AMNH 30607), but stouter than the basihyal of gymnurids, and not fragmented as in potamotrygonids. It is situated just posterior to the lower jaw symphysis (fig. 20). Unfortunately, the pseudohyoid arch is mostly obscured in FMNH PF 15180, but the right ventral pseudohyoid (vph) is present as a very slender, prismatically calcified element that still articulates with its associated pseudohyoid rays (which are posteriorly directed). Lateral to the basihyal are the first hypobranchials (hb1), which in FMNH PF 15180 are exceptionally well preserved (fig. 20). The first hypobranchials are shaped as in potamotrygonids (especially Potamotrygon magdalenae [AMNH 55620]; see fig. 34B), urolophids, urotrygonids (fig. 33B), and other benthic stingrays, that is, tapering posteriorly only slightly from their articulation with the basihyal, but not as tapered as in Gymnura (fig. 35B). Both hypobranchial 1 elements are almost as long as one of the mandibular cartilages from symphysis to jaw corner. The hypobranchials are obliquely oriented and contact posteriorly the ventral pseudohyoid, lateral to the anterior extension of the basibranchial copula. The first hypobranchial elements in †Asterotrygon were not as broad anteriorly as in Taeniura, or even as much as in some species of Dasyatis. Posterior to the hypobranchials, and separated from them by a small gap, is the basibranchial copula (bb; medial plate). The copula bears a poorly preserved anterior projection, very similar to the condition seen in most benthic stingrays (Urobatis, Urolophus, Urotrygon, Taeniura, potamotrygonids), but absent in Plesiobatis and pelagic stingrays (Miyake and McEachran, 1991). This projection in †Asterotrygon appears to be more anteriorly extended than in Gymnura as well. The full posterior extent of the basibranchial copula is difficult to determine, but in stingrays it usually extends as a very slender projection to the level of the scapulocoracoid. In †Asterotrygon, the posterior extension of the copula is difficult to observe, but is present in FMNH PF 15180 as a sightly elevated, slender ridge of prismatic cartilage, similar to the basibranchial copula present in Trygonoptera, Urobatis, and potamotrygonids, and not as short as in gymnurids. The extent to which ceratobranchials are fused to each other (if at all) and to the ventral pseudohyoid bar cannot be determined in the fossils.

The articulation between the last pair of ceratobranchials and the scapulocoracoid can be seen in FMNH PF 12989 and FMNH PF 15180, where the ceratobranchials appear relatively straight and narrow. We consider these articulating elements to be ceratobranchials and not epibranchials, as both specimens are ventrally preserved and ceratobranchials are typically more slender and straight compared to the slightly curved epibranchials of nonmyliobatid stingrays. The preserved posterior portion of both ceratobranchial 5 elements are prismatically calcified and articulate with the coracoid lateral to its connection with the synarcual cartilage, at about one-third of synarcual length. The gill basket occupied almost the entire gill cavity between the pectoral propterygia in †Asterotrygon.

In specimens FMNH PF 15166 (fig. 2), FMNH PF 15180 (figs. 12, 20), FMNH PF 12989 (fig. 4), FMNH PF 14069 (fig. 5), FMNH PF 12914 (fig. 6) and FMNH PF 14567 (fig. 7) numerous disarticulated gill rays (gr) are scattered in the gill chamber, but some are preserved in their original positions in FMNH PF 15180 (fig. 20) and in FMNH PF 15166. In Recent stingrays, gill rays supporting the gill lamellae are present on the dorsal and ventral pseudohyoid arches and on cerato- and epibranchials 1–4. In our fossils it is generally not possible to associate the slender and fragmented gill rays with any particular arch, even though they occur throughout the entire gill chamber (but in FMNH PF 15180 the ventral pseudohyoid is associated with corresponding gill rays; fig. 20). The gill chamber is compacted anteroposteriorly in FMNH PF 12989 and in FMNH PF 14069. A large gap is visible between the basihyal and basibranchial anterior extension in FMNH PF 14069, even though the basihyal is not well preserved and its precise shape cannot be discerned (our identification of a patch of prismatic calcification just posterior to the mandibular cartilages as the basihyal in this specimen is tentative). There are no signs of extrabranchial cartilages in †Asterotrygon (these are formed from the fused distal tips of gill rays). The gap that separates the mandibular cartilages from the anterior limit of the basibranchial (which itself is usually not visible) contains disarticulated elements of the visceral arches and synarcual cartilage in some of the fossils.

Synarcual and Vertebral Skeleton

Stingrays have two synarcual cartilages, an anterior cervicothoracic synarcual (ct syn) and a posterior thoracolumbar synarcual (tl syn) (Compagno, 1973, 1977). The thoracolumbar synarcual is unique to stingrays. The anterior synarcual articulates with the occipital region of the neurocranium and extends posteriorly to meet the scapular processes of the shoulder girdle. It is prismatically calcified and consists of a dorsal and ventral component, both formed ontogenetically by fusion of different vertebral (basidorsal and basiventral) elements (Miyake, 1988). The anterior synarcual is relatively well developed and prominent in stingrays. In †Asterotrygon the cervicothoracic synarcual is preserved in many specimens (e.g., figs. 6, 7, 22). The medial crest (mdc) of the synarcual is a dorsal ridge that runs anteroposteriorly along almost its entire length and is visible in dorsally exposed specimens, especially FMNH PF 14567 (fig. 22B) and FMNH PF 12914, where the ridge is markedly elevated. One ventrally exposed specimen (FMNH PF 12989) clearly displays a string of vertebrae running the entire length of the preserved synarcual (13 partially fused vertebrae along the synarcual midline are present). In most stingrays, vertebral centra occupy only the posterior one-half to one-third of synarcual length. Lateral stays (ls; supposedly formed by fused pleural ribs and basiventrals) are not visible in any specimen, even though they are prominent components of cervicothoracic synarcuals of stingrays. Dorsally, the posterior segment of the first synarcual is composed of the secondarily fused suprascapulae that articulate with the ascending lateral scapular processes of the shoulder girdle (FMNH PF 12989, FMNH PF 14567). The posterior, suprascapular portion of the synarcual appears rectangular in outline in almost every specimen, even those that are ventrally preserved. The suprascapular portion of the synarcual in †Asterotrygon is robust compared to extant nonmyliobatid stingrays, where it appears to be more slender (Urotrygon, Urobatis, Potamotrygon). The overall width of the synarcual anterior to the suprascapulae is similar to extant benthic taxa as seen in NHM P 61244, where the synarcual width is just less than neurocranial width at the occipital region. Spinal nerve foramina cannot be identified and counted in our fossils.

The posterior or thoracolumbar synarcual is less pronounced than the cervicothoracic synarcual, and is a relatively simple structure. This synarcual tapers posteriorly, ending at about midway between the scapulocoracoid and pelvic girdle, and has unfused individual vertebral centra along its entire length. The second synarcual in †Asterotrygon is very similar to extant nonmyliobatid stingrays. In FMNH PF 12989, where it is well preserved, the thoracolumbar synarcual is about five times the width of individual centra at its origin, where it attains its greatest lateral dimensions, and is prismatically calcified (fig. 22). Observation of the second synarcual requires the fossil specimen to have a dorsoventrally preserved tail region posterior to the shoulder girdle (e.g., FMNH PF 12989); in other words, the vertebral column cannot be dislocated and rest on its side, as is commonly the case. FMNH PF 14069 has the column preserved laterally, and only the very anterior portion of the second synarcual is present.

Vertebral centra do not articulate with pleural ribs in †Asterotrygon, and ribs associated with basiventrals are absent, as in all stingrays. Numbers of vertebral centra posterior to the first synarcual are given in tables 2 and 3. It is more difficult to determine transitions between mono- and diplospondyly in our fossils than in extant stingrays, because myomeres are not preserved and individual centra may be displaced. In stingrays, this transition usually occurs at the level of the pelvic girdle or just posterior to it, and that appears to be the case in NHM P 61244. In FMNH PF 14069, where the vertebral centra are exceptionally well preserved, the area of transition is not entirely clear but appears to be posterior to the pelvic girdle, close to the posterior edge of the pelvic fin. Specimens FMNH PF 15166 (holotype; fig. 2), AMNH P 11557 (paratype; fig. 3), FMNH PF 14069 (paratype; fig. 5), FMNH PF 12914 (fig. 6), and FMNH PF 14567 (fig. 7) have enlarged, laterally compressed neural spines (ns) or arches from just posterior to the scapulocoracoid to close to the caudal stings. The neural arches are obliquely oriented posteriorly in relation to the centra and are closely situated to each other. Ventral arches are also present in extant stingrays, but are much smaller in comparison to the neural arches at least until the level of caudal stings, and this was probably the case in †Asterotrygon as well. Individual basidorsals or interdorsals cannot be discerned in the fossils. Individual vertebrae are spool-shaped in lateral view and more elongated in the caudal region, even though they are much smaller compared to centra close to the second synarcual. Faint rings of calcification are present peripherally within the centra and can be observed in specimens in which the centra are preserved, exposing their anterior or posterior surfaces (FMNH PF 14098, USNM 2028).

†Asterotrygon, unlike almost all extant stingrays that lack a caudal fin, has discrete vertebral centra continuing to the posterior tip of the tail, which can be observed in juveniles, small specimens, and adults. The unsegmented cartilaginous rod (cr), a notochordal extension that continues caudally beyond the caudal stings (instead of separate vertebral centra), is therefore absent in our fossils (cf. figs. 24, 26 to figs. 37C, D, and 40E). This rod is present in most stingray genera and is absent in taxa that have a caudal fin (urotrygonids and urolophids), Plesiobatis (Nishida, 1990), Hexatrygon (Heemstra and Smith, 1980), as well as in all other nonmyliobatiform batoids. However, this condition may vary slightly among species, as Dasyatis annotatua has individual vertebrae extending well beyond the caudal sting (CSIRO T 694, CSIRO T 697) and Gymnura has discrete vertebrae extending to the tip of the tail (we have confirmed this in four species of Gymnura so far), contrary to the findings of Lovejoy (1996) and McEachran et al. (1996).

Appendicular Skeleton

The entire shoulder girdle (scapulocoracoid) consists of a single, somewhat flattened U-shaped structure in †Asterotrygon. The scapulocoracoid (sc) is composed of a transverse ventral coracoid bar (co) and lateral, dorsally projecting scapular processes (scp) that articulate medially with the suprascapulae (ssc) which are fused to the posterior segment of the cervicothoracic synarcual. The coracoid bar is best preserved in ventrally exposed specimens (FMNH PF 15180 [fig. 20], FMNH PF 14069), but the entire shoulder girdle is easily recognized as a prominent transverse structure at mid disc in all specimens. The anteroposterior length of the coracoid bar varies slightly among specimens, but how much of this variation is due to its orientation during preservation is unknown. In FMNH PF 14069 the coracoid resembles that of Recent Potamotrygon and other benthic genera in being slender transversally, but unlike Potamotrygon the anterior border of the coracoid is relatively straight across and not slightly concave. The scapular processes are best preserved in FMNH PF 14567 (fig. 22B) and FMNH PF 12914 (fig. 6), where they appear as large subtriangular pieces articulating with the synarcual by means of prominent articular surfaces. This articulation in stingrays is of the ball-and-socket type (Compagno, 1973, 1977), and we can infer on the basis of its position and appearance in dorsoventral view that this was the case in †Asterotrygon as well. The scapular processes are obliquely positioned in relation to the synarcual cartilage, as in modern stingrays (Urotrygon, Urolophus). The foramen present on the dorsal component of the scapular process of some stingray taxa (“af” fenestra of Miyake, 1988; “foramen of the scapular process” of Lovejoy, 1996) cannot be identified in the fossil material.

The scapulocoracoid laterally joins the internal skeleton of the pectoral fins or disc, which is composed of basal cartilages (pro-, meso-, and metapterygium) articulating with numerous elongated and subdivided radial elements that extend to the outer margins of the disc (plesodic condition). The propterygium (pro) extends for slightly over one-half of disc length. In both †Asterotrygon and extant stingrays, the anteriormost portion of the propterygium becomes subdivided into three or more segments anterior to the level of the nasal capsules (even in Mobula [AMNH 15319], which bears an elongated propterygium that is anteriorly expanded to support the cephalic lobes, but in which there is also a small anterior segment). In some extant nonmyliobatid stingrays (Urolophus, Trygonoptera, Dasyatis, Himatura) the propterygium becomes subdivided at about midnasal capsule length (cf. Lovejoy, 1996), but in †Asterotrygon this division appears to be in front of the anterior cranial margin, even though this is difficult to determine in the fossils. The anterior segments of the propterygia bend medially toward the midline, leaving a prominent gap that was occupied by the small rostral cartilage (a rostral piece was presumably present—possible vestiges of it are found in NHM P 61244, see above). The anteriormost segment of the propterygium (or the first radial propterygial element anteriorly) is bifid, articulating with two rows of radials as in extant nonmyliobatid stingrays. The propterygium bears an elongated ridge and sulcus for the insertion of the pectoral fin muscles along almost its entire length.

In †Asterotrygon, the posterior section of the propterygium contacts the scapulocoracoid, as in various extant benthic taxa (Dasyatis, Urolophus, Potamotrygon), by means of an elongated articular surface with two distinct condyles (procondyles). The anterior procondyle is at the anterolateral corner of the shoulder girdle, and the posterior one is close to the mesocondyle (the articular surface of the mesopterygium). The pro- and mesocondyles are clearly separated from each other in stingrays (Miyake, 1988), and this is the case in †Asterotrygon (fig. 22A). The lateral aspect of the scapulocoracoid, however, is not visible in our two-dimensional fossils, but the articulation between the propterygium and scapulocoracoid can be discerned from a dorsoventral perspective, appearing very similar to Recent nonmyliobatid stingrays. The medial or internal surface of the posterior tip of the propterygium is slightly concave where it articulates with the scapular process (FMNH PF 12989), contacting it in at least two different places (presumably at the procondyles) but possibly throughout more of its length, as is common in many extant benthic stingrays. The lateral aspect of the scapular process is relatively straight, and consequently the propterygium is not separated from the mesopterygium by a small process in †Asterotrygon.

The mesopterygium (mes) is more difficult to observe in our fossils due to scattering of its prismatic calcification. It is slightly wedge-shaped and subtriangular in the holotype FMNH PF 15166 (right side), in FMNH PF 14567 (fig. 22B), and in the left side of paratype AMNH P 11557, as in many nonmyliobatid stingrays, but in FMNH PF 12989 (fig. 22A) the mesopterygium is more elongate, projecting anteriorly somewhat obliquely to the propterygium. This discrepancy is probably due to incomplete preservation in the latter specimen. The scapular process is situated at close to the posterior two-thirds of its length. Approximately 13 mesopterygial radials are present in FMNH PF 12989 (counted on both sides; table 2).

The metapterygium (met) is elongated and curved, tapering posteriorly and terminating close to the level of the pelvic girdle. In †Asterotrygon, the metapterygium appears as a single piece, unlike many nonmyliobatid stingrays where it can become subdivided posteriorly into smaller segments that also articulate with radials. The metapterygium also bears a pronounced groove for the insertion of the pectoral fin musculature; it is similar to the propterygium but is not as elongate or as wide.

The radial elements articulating with basal cartilages are identical in structure (numbers of radials for each basal are shown in tables 2 and 3). All radials are segmented from the basal cartilages to the outer disc margin. In FMNH PF 12989, FMNH PF 12914, and FMNH PF 14567, where radial segments are clearly preserved, there are at least 15 radial segments, but perhaps one or two additional segments were present distally. In the holotype (fig. 2), 17 radial segments are present at the level of greatest disc width. The individual segments of the radials are relatively short and reduce slightly in length in the direction of the outer disc margin. In specimens FMNH PF 15166 and FMNH PF 12914, radials of the ninth row (counting from basals to outer margin of disc) are bifid, but in FMNH PF 12989 and FMNH PF 14567 it is the eighth radial segment that bifurcates. Therefore, there are always twice as many radials at the outer disc margin than articulating with the basal cartilages. The derived condition of Gymnura and myliobatids (Nishida, 1990) in having laterally expanded radial elements that articulate with adjacent radials (fig. 39A, B) is not present in †Asterotrygon.

The puboischiadic bar (pib; pelvic girdle), being broad and dorsoventrally flattened, is well preserved in most specimens, but more so in FMNH PF 15166, FMNH PF 15180 (right side only), PF 12989 (fig. 23A), and PF 12990. The girdle is stoutest at its corners, where the basipterygia (bp) originate and extend posteriorly, tapering to a variable extent. Its anterior surface is roughly straight in most specimens, and is much less arched in comparison to the slightly concave posterior surface (FMNH PF 12989). In other specimens (FMNH PF 15166, FMNH PF 14567) the pelvic girdle is slightly arched at both anterior and posterior margins. Minor distortion and preservational differences may account for the observed variation among specimens.

The pelvic girdle resembles that of many stingrays in general aspects, excluding potamotrygonids. It is arched anteriorly but not as much as in Gymnura (fig. 36C, D) or pelagic stingrays (Mobula). It is difficult to determine if a median prepelvic projection is present due to crushing from vertebrae, but if one was present it would have been a short, triangular process similar to extant stingrays other than potamotrygonids (Urotrygon chilensis [fig. 36E], Taeniura lymma [fig. 36A, B], Urobatis jamaicensis [AMNH 30385], Urolophus cruciatus [AMNH 59890], species of Dasyatis). Also, no pronounced lateral prepelvic processes as in Potamotrygon can be seen in †Asterotrygon, but the pelvic girdle projects anterolaterally at its corners in some specimens (e.g., FMNH PF 12914, FMNH PF 12990), similar to the condition in Urotrygon, Plesiobatis, and other stingrays (Nishida, 1990).

There is some evidence of small ischial processes (isp) (Miyake, 1988) on the inner corners of the posterior margin of the pelvic girdle in †Asterotrygon (ischial processes are present in most stingrays, and are particularly well developed in potamotrygonids; fig. 16B). In FMNH PF 12989 (right side; fig. 23A), FMNH PF 12990, FMNH PF 14069, and NHM P 61244, these processes appear as ill-defined protuberances. In FMNH PF 14567 and FMNH PF 14097, however, the processes are more triangular and located where the posterior concavity of the pelvic girdle curves inward, as in Urotrygon and Urobatis. Other than the ischial processes, there are no distinct features of the posterior margin of the girdle, such as median postpelvic processes described and figured by Nishida (1990: fig. 36) for platyrhinids. (Nishida's “post-pelvic processes” [also McEachran et al., 1996] seem to amount to irregularities in the posterior surface of the puboischiadic bar, and not to distinct projections. This is corroborated by the apparent absence of any distinct process from Platyrhina sinensis [AMNH 26413, AMNH 44055] and from Platyrhinoidis triseriata [OSU 40]; Carvalho, in press.) The iliac process (ilp), which also projects posteriorly and medially from the corners of the puboischiadic bar in many stingrays and more basal batoids, may be present in our fossil specimens. However, because this process may overlie or be aligned with the basipterygium to some extent, its identification is difficult in the fossils (FMNH PF 14097). The iliac process is preserved in FMNH PF 12989 (fig. 23A), where it projects posteriorly in an oblique fashion from the pelvic girdle. It is elongated, more or less straight, and tapers posteriorly (extending to at least the sixth radial element in FMNH PF 12989). The iliac process of many extant stingrays may be slightly curved to project medially (and dorsally), and is not as tapered. The iliac process in extant stingrays is not as elongate as in FMNH PF 12989. The iliac process in †Asterotrygon may also have a posterodorsal orientation, but this is not evident in the fossils. Obturator foramina (of), through which pass diazonal nerves, are not well preserved, but can be observed in FMNH PF 15180 (one small foramen on right side; fig. 12).

The pelvic fin is also plesodic; the first pelvic radial element is characteristically enlarged (efr), as in many stingray groups, and articulates with the lateral aspect of the puboischiadic bar. Even though it has been called the pelvic propterygium (Nishida, 1990), this element should not be considered the serial homolog of the pectoral propterygium; it is formed by ontogenetic fusion of radials. The numbers of pelvic radials are given in tables 2 and 3, and they fall within the ranges observed in many extant stingrays, increasing slightly in relation to specimen size (cf. Nishida, 1990: 58, table 3). Each pelvic radial in FMNH PF 12989 has at least three segments (four in FMNH PF 14567), the most basal segment being much more elongate than the more distal segments. The most distal segment is bifurcated.

Claspers are present and well developed in only one specimen (figs. 4, 23A). Prismatic calcification is present over the entire axial cartilage (ax), further corroborating that this specimen is sexually mature. The claspers are relatively long and straight, with preserved axial cartilages extending along most of its length. The entire length of the claspers is roughly just under one-third of tail length as measured from posterior margins of pelvics. No beta cartilages can be seen, but this may be preservational. On the right clasper, running more or less parallel to the axial cartilage, there is a slightly curved piece that may be the beta cartilage (terminology from Miyake, 1988). The axial cartilage bends outward slightly at its posterior tip, and the entire clasper is spoon-shaped. Discrete cartilages of the terminal clasper components are not readily observed, but more than one is present. The largest element apparently is the ventral covering piece (ventral terminal cartilage; tc), but only its impression remains in the fossil. Anterior to the ventral terminal cartilage are impressions of the dorsal and/or ventral marginal cartilages.

The dorsal fin is present in most specimens of †Asterotrygon in which the distal portion of the tail is laterally preserved to some degree (fig. 24). The dorsal fin is more prominent and well preserved in one female paratype (FMNH PF 14069; fig. 24B) and in specimen FMNH PF 15181 (fig. 24F). NHM 61244, a small neonate, also has a weak impression of the dorsal fin present anterior to caudal stings. The holotype has a small dorsal fin preserved, covered by denticles, but no internal structure is apparent (fig. 24E). In specimens FMNH PF 14069 and FMNH PF 15181 the tail is markedly preserved on its side, accounting for the excellent exposure of the dorsal fin (also FMNH PF 14567; fig. 24D). The dorsal fin is located just anterior to the caudal stings, but its posterior lobe slightly overlaps the caudal stings. Presumably, the overlapping is due to the posterior lobe being free from the dorsal surface of the tail, with the fin inflecting a short distance anteriorly before inserting onto the dorsal tail surface. The fin is relatively small and rounded in outline. Small hooklike denticles are present over most of the fin in most specimens (FMNH PF 15166, FMNH PF 12914, AMNH P 11557), but in both FMNH PF 14069 and FMNH PF 15181 denticles are more apparent on the anterior margin or ridge of the fin. The presence of hooklike denticles covering the dorsal fin or at least its dorsal ridge is a unique feature for †Asterotrygon, not occurring in other stingrays where a dorsal fin is present.

The dorsal fin is clearly plesodic in FMNH PF 14069 and FMNH PF 15181, with internal skeletal support primarily composed of slender, elongated radial elements, which radiate from the fin base to the outer fin margin (less evident in FMNH PF 14567). In FMNH PF 15181 the radials bifurcate distally (less so than FMNH PF 14069). Basal components of the dorsal fin endoskeleton also appear to be present in FMNH PF 14069, but these are not present in FMNH PF 15181. These elements are slightly wider than the more distal radials and are organized into a poorly defined row from the fin base to about one-third of its height. The radial elements are very prominent in FMNH PF 14069, extending to the outer fin margin. In this respect the dorsal fin of †Asterotrygon resembles the dorsal fin of Urolophus flavomosaicus (CSIRO 718.22) and Aetoplatea zonura (Nishida, 1990; fig. 41A here), and even the dorsal fins of rajids (Nishida, 1990). The dorsal fins of Recent myliobatids have enlarged basal elements and occupy a greater portion of the dorsal fin base compared to FMNH PF 14069 (Aetomylaeus maculatus [AMNH 32500], Aetobatus narinari [fig. 41E]; Myliobatis, Rhinoptera, Mobula, Manta; Nishida, 1990). The “basal elements” in FMNH PF 14069 are probably not enlarged basal cartilages as in myliobatids, but just small radial segments horizontally arranged (as in Aetoplatea zonura; fig. 41A). The height of the dorsal fin is just greater than tail height at the level of the dorsal fin in FMNH PF 14069 and FMNH PF 14567, but in AMNH P 11557 and FMNH PF 12914 the dorsal fin is not as tall as the tail at its insertion (fig. 24). This may be due to different orientation of the fin and tail in the specimens. Although the dorsal fin is not apparent in all specimens (usually not in those in which the tail is preserved dorsoventrally), a small cluster of denticles is sometimes observed anterior to the caudal stings, indicating that the fin was originally present (e.g., FMNH PF 12989).

Specimen FMNH PF 12990 is unique in having an impression of a putative caudal fin posterior to the caudal sting (fig. 10A). The fin is incompletely preserved but was apparently rounded dorsally and posteriorly, with a more prominent dorsal lobe and a smaller ventral lobe. If this interpretation is correct, then the caudal fin is similar to some species of Urobatis, Urolophus, and Trygonoptera in being taller than long. Alternatively, the caudal fin in FMNH PF 12990 may represent incompletely preserved tail-folds, as only a small portion of the tail posterior to the caudal sting is present. We cannot be sure because the posterior extent of this “fin” is not known. The tail region posterior to the caudal stings in FMNH PF 14567 (figs. 7, 24D) has small dorsal and ventral elements (rudimentary radials according to Nishida, 1990), reminiscent of dorsal and ventral tail-folds of some Potamotrygon species. Rudimentary radials, both dorsal and ventral, are also present in the holotype (fig. 24E), FMNH PF 12989, and to a lesser extent in FMNH PF 14069 (not visible in fig. 24B). In FMNH PF 12989, rudimentary radials occur only on one side of the vertebral column, but it is not possible to discern which side, as the tail is preserved laterally. Low tail-folds were therefore present in †Asterotrygon, and it probably extended to the posterior tip of the tail and originated shortly after the caudal stings. The height of the tail-fold both dorsally and ventrally was slightly greater than the height of the tail where the tail-folds originated.

Dermal Skeleton

ORDER MYLIOBATIFORMES SENSU COMPAGNO, 1973

SUPERFAMILY MYLIOBATOIDEA

†Heliobatis Marsh, 1877

†Heliobatis radians Marsh, 1877 Figures 26B, 28–30, table 4

Phylogenetic Procedures and Coding

Characters that do not vary within the ingroup (synapomorphies of Myliobatiformes) were excluded from the matrix (characters 1– 6 and 8 listed above), as were characters restricted to terminal taxa (autapomorphies). However, autapomorphies were included if a character is present in one terminal and scored as uncertain or unavailable (?) in another, or if the autapomorphy is part of a multistate series. Dasyatis and Urobatis are each coded as a single terminal, contrary to the coding of Lovejoy (1996), but similar to that of McEachran et al. (1996). Lovejoy coded Dasyatis as two separate terminals because the species he examined of this genus differed in one character (character 3 below; however, it is not entirely clear what species of Dasyatis are included in each of Lovejoy's terminals). We have coded Dasyatis as uncertain for this character. Urobatis was coded in Lovejoy (1996) as two terminals, one each for Pacific and Atlantic species, which differed in two characters in his matrix (his characters 19 and 21). Urobatis is coded here as a single terminal because only one of these characters is included in our analyses (character 25 [= character 19 of Lovejoy, 1996], coded here as polymorphic for Urobatis). The other character used by Lovejoy, degree of extension of the lateral stay of the cervicothoracic synarcual cartilage, is not included because the degree of extension of the lateral stay is difficult to properly divide into discrete categories (as also concluded by McEachran et al., 1996). Lovejoy's (1996) matrix coded Plesiotrygon and Potamotrygon identically, as well as for Indo-west Pacific Himantura and “Dasyatis 1”. These taxa are not identically coded in our matrix, however, due to the inclusion of additional characters and differential coding of others. Genera not available for study, or insufficiently described anatomically in the literature, are not included in our analysis (e.g., Pastinachus and Urogymnus; the former is monotypic and the latter has two species). Because specimens of Aetoplatea were unavailable, only Gymnura is coded for the Gymnuridae. Himantura is coded as two terminals, following Lovejoy (1996), because the amphi-American species (H. schmardae and H. pacifica) form a monophyletic group (Carvalho and Lovejoy, in prep.) and are morphologically very distinct from the other species of Himantura, all of which are Indo-west Pacific in distribution. The genera of Potamotrygonidae are coded separately because we provide additional observations for some of their characters, some of which may have an effect on relationships within the family. The outgroups included are composed of two more distantly related batoids (a guitarfish and a skate), but our morphological comparisons included all batoid groups. All multistate characters were run as unordered (characters 2, 3, 20, 22, 25, 28, 30, 35, 37, and 43). The final matrix (table 5) is composed of 44 characters, but about a dozen additional characters were considered for inclusion and eventually discarded because they either varied intraspecifically or proved too difficult or arbitrary to divide into discrete states, and were therefore deemed inappropriate for phylogenetic analysis in the present context. Character coding should be as precise and explicit as possible, involving confirmation in more than one specimen of a given terminal when enough material is available, as accurate scoring is a fundamental prerequisite of phylogenetic analysis (Patterson and Johnson, 1997; Grande and Bemis, 1998), especially in groups that have previously manifested high levels of homoplasy.

Phylogenetic analyses were conducted using Hennig86 (Farris, 1988), and implemented with the aid of the software Tree Gardiner (Ramos, 1997) and WinClada (Nixon, 1999). An exact parsimony strategy (*ie) was employed to search for all equally most parsimonious trees. Nodes were diagnosed with the aid of the Dos Equis (xx) function of Hennig86 and with tree diagnostics featured in Tree Gardiner and WinClada. The matrix was also run in Nona (Goloboff, 1999), which produced the same results as Hennig86 (with fewer minimum-length trees, however). Characters used to diagnose nodes are only those that displayed the same optimization on all of the equally most parsimonious trees (i.e., only unambiguous characters were used to diagnose monophyletic groups; the mapping of conflicting characters onto the strict consensus tree artificially increases its length due to its polytomies, and it is therefore less parsimonious than any of the original trees). The strict consensus tree is depicted in figure 43. Characters that are ambiguous due to multiple optimizations that require the same number of steps are resolved by favoring reversals over independent gains. This procedure was also used for characters scored as unknown (?) in the Green River fossils but that could be pushed farther down the tree (which assumes, therefore, that they were present in the fossils), as this procedure saved a step for each character (characters 12, 22, and 44; all ambiguous characters are shown separately in fig. 44). Accelerating the transformations of these characters ascribes a greater generality to their distribution, which is considered to be a more coherent implementation of parsimony (for a review, see discussion in Schuh, 2000). However, ambiguous characters in which a step is not saved by accelerating their transformations (characters scored with a “?” in at least one terminal), are placed at the node where they are known to occur only with certainty (see fig. 44; these features were not pushed farther down the tree; in other words, no assumptions were made about their presence in taxa scored as uncertain when no steps could be saved; see legend of fig. 44 for more details).

Character Descriptions

The numbers below correspond to character numbers in the matrix (table 5). Terminals in which the character is present are given in brackets after the respective character-state description (except for state 0). Literature citations containing more morphological information are included for characters when appropriate. Table 6 provides the length (number of steps), consistency index (CI), and retention index (RI) for all characters, as well as the number of minimum-length trees in which each character is present with these indices, including the strict consensus tree (see Farris, 1989, for a discussion of these indexes). For most characters these indices are optimal for all minimum-length trees. If more than one solution is possible for a given character in a subset of the minimum-length trees, than variations in length, CI, and RI are given for each subset in table 6.

Results of Phylogenetic Analysis

Phylogenetic analysis of the matrix presented in table 5, summarizing the anatomical information described above, resulted in 35 equally most parsimonious trees (length = 82 steps, CI = 0.68, RI = 0.82), the strict consensus of which is shown in figure 43 (ambiguous characters are shown separately in fig. 44). The strict consensus tree (length = 86 steps, CI = 0.65, RI = 0.79) is not entirely dichotomous, but contains various monophyletic components and only two unresolved nodes (the monophyletic groups discovered, some of which are novel, are described below). The unresolved relationships mostly concern the various dasyatid genera, which are placed in a large polytomy with the node Gymnuridae + Myliobatidae in the strict consensus, and the affinities of †Asterotrygon and Plesiobatis. The 35 minimum-length trees can be reduced to five subgroups of trees once the uncertainty regarding the relationships of dasyatid genera are removed; the five subgroups of trees vary only in the positions of †Asterotrygon and Plesiobatis (summarized in fig. 45). The nonmonophyly of the whiptailed stingrays (Dasyatidae) in the strict consensus is not surprising, given that there are no unequivocal characters known to support it (a monophyletic Dasyatidae does not occur in any of the equally most parsimonious trees either). The presence of tail-folds has traditionally been used to diagnose the family (Bigelow and Schroeder, 1953; Compagno and Roberts, 1982, 1984; Nishida, 1990), but these are absent in amphi-American Himantura and Himantura sensu stricto, and are also present in potamotrygonids and †Heliobatis and are therefore plesiomorphic at this level.

In relation to the Green River stingrays, our analysis indicates that both genera do not form a monophyletic unit. But this is not surprising given that no putative homologies were discovered and included in the matrix. The phylogenetic placement of †Asterotrygon varied in the minimum-length trees; it was resolved as either the next most basal stingray genus after Hexatrygon (in 7 of the 35 minimum-length trees), as the next most basal stingray genus after Hexatrygon but in a polytomy with urolophids (also in 7 of 35 trees), or as the sister-group to the Urolophidae (in 21 of the 35 trees; fig. 45C–E). †Heliobatis is resolved as the most basal genus of the Myliobatoidea, sister-group to the clade Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae)) in all equally most parsimonious trees. Therefore, †Asterotrygon and †Heliobatis (or their direct ancestors) probably independently invaded the freshwater system of Fossil Lake and were phylogenetically divergent well before this invasion occurred (further discussed below under “Biogeographical Implications”).

The relationships of Plesiobatis varied considerably in our minimum-length trees (fig. 45); it was either sister-group to the remaining stingrays except Hexatrygon (in 7 out of 35 trees), sister-group to the remaining stingrays except Hexatrygon but in a polytomy with †Asterotrygon + Urolophidae (again in 7 of 35 trees), or (in 21 of 35 trees) sister-group of the large node Urotrygonidae + Myliobatoidea (where Myliobatoidea = †Heliobatis + (Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae)))). The latter relationship is a departure from previous studies, in which Plesiobatis has figured as either the most basal stingrays genus or the next most basal genus after Hexatrygon (Nishida, 1990; Lovejoy, 1996; McEachran et al., 1996).

Our phylogenetic analysis uncovered two additional stingray synapomorphies that are present in all minimum-length trees: characters 20 (state 1) and 23. The latter feature, anterior expansion of the medial (basibranchial) plate, requires homoplasy as it is reversed in myliobatids, while Plesiobatis has independently lost the basibranchial anterior expansion (a further autapomorphy for this genus). Character 20, regarding the basihyal cartilage, is highly modified in stingray evolution: it is lost in Urotrygon, becomes fragmented for the the large node Urotrygonidae + (†Heliobatis + (Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae)))), but is further lost for myliobatids (note that there is more than one possible optimization of this latter transformation as shown in fig. 44). The loss of the dorsal fin (character 33) is a putative stingray synapomorphy but is not optimized in this way in all resulting trees (it is in some, however; see fig. 44). Myliobatids, Trygonoptera, Aetoplatea, and †Asterotrygon have all regained the dorsal fin independently. In trees that support a monophyletic †Asterotrygon + Urolophidae, it is possible that the dorsal fin evolved only once and was subsequently lost in Urolophus, which would render the mere presence of the dorsal fin plesiomorphic for †Asterotrygon; our use of this character to diagnose this fossil taxon is not compromised in this scenario, however, as the fin is uniquely covered in denticles and has a well-defined plesodic internal structure (whereas the internal structure of the dorsal fin of Trygonoptera, if present, is unknown; it may not be arranged as in †Asterotrygon).

The node comprising all stingrays with the exception of Hexatrygon is unequivocally supported by a single character: presence of the hyomandibular-Meckelian ligament (character 13). Additional features that are placed at this node in the tree in figure 44 are the ventrolaterally projecting nasal capsules (character 12), ankylosis of fourth and fifth ceratobranchial cartilages (character 22; further modified for Myliobatidae), and the nasal curtain reaching posteriorly to level of mouth opening (character 44). These three characters are unknown in the Green River stingrays, precluding their inclusion in the strict consensus of figure 43. There are two additional characters featured at this node, but both are scored as unknown for Hexatrygon: the scapular loop of the trunk lateral-line canal (character 4) and the fusion of the ventral pseudohyoid with the first ceratobranchial (character 21). The former character is unknown in Hexatrygon (and we were unable to examine prepared material), and the latter has been conflictingly described in the literature (see character description above). If both characters prove upon further examination to be absent from Hexatrygon they will represent additional synapomorphies for this node (as in fig. 44), but if present they will be synapomorphic for stingrays.

The node uniting Urotrygonidae and the large group composed of †Heliobatis + Potamotrygonidae + “Dasyatidae” + Gymnuridae + Myliobatidae is supported by the fragmented basihyal cartilage (character 20, state 2), which further reverts to an unfragmented (complete) state in Gymnura, and is independently lost in Urotrygon and pelagic stingrays (figs. 43, 44). This character requires five steps in all equally most parsimonious trees (as well as in the consensus). The component †Heliobatis + Potamotrygonidae + “Dasyatidae” + Myliobatidae is monophyletic, as evidenced by the reduction of the caudal fin to tail-folds (character 35; but homoplasy is required for this character as Paratrygon, amphi-American Himantura, Himantura, and the node Gymnura + Myliobatidae have all lost the tail-folds independently [state 2], and †Asterotrygon has acquired them independently). The clade Potamotrygonidae + “Dasyatidae” + Myliobatidae is supported by the cartilaginous unsegmented rod in the tail (character 34, devoid of homolasy).

The most innovative result of our phylogenetic study is the relationship between dasyatid genera (not as a monophyletic unit, however) and the group Gymnuridae + Myliobatidae. This relationship, not present in the phylogenies of Nishida (1990), Lovejoy (1996), and McEachran et al. (1996), is supported by the dorsal fossa on the scapular process (character 26), which occurs independently in Urotrygonidae and Trygonoptera, and is secondarily lost in Gymnura (note that this character was also included in the analyses of both Lovejoy and McEachran et al.). Even though there is much homoplasy in this character, it supports a more recent relationship between dasyatid genera (not as a clade however) with gymnurids and myliobatids in all minimum-length trees. The subpleural components of the hyomandibular lateral-line canal (character 2) are highly modified in dasyatid genera and in the node Gymnuridae + Myliobatidae, and were scored with different conditions of a multistate character; however, because dasyatid genera do not form a monophyletic unit, placing state 1 of this feature as a synapomorphy of “Dasyatidae” + (Gymnuridae + Myliobatidae) is equivocal (see fig. 44). There are important biogeographic implications derived from this relationship, and these are further discussed below.

Gymnuridae forms a monophyletic group with pelagic stingrays (Myliobatidae), sharing a shortened orbital region (character 10), fragmented mesopterygium (character 28; further lost in higher myliobatids), and lateral extension of radial segments in the pectoral fin (character 29). Loss of the tail-folds (character 35, state 2) is ambiguous at this node (fig. 44). Other characters putatively synapomorphic at this level are the presence of the lateral hook of the subpleural component of lateral-line canal (state 1 of character 2), and the highly arched pelvic girdle (character 32) (see fig. 44). Character 2 has multiple optimizations (either states 1 or 2 can be equally interpreted as a defining feature for this node), while character 32 is absent in Myliobatis, allowing for its independent acquisition in Gymnura and the node Aetobatus + (Rhinoptera + Mobula). The transverse, unfragmented state of the basihyal (character 20, state 1) can also be placed at this node (fig. 44; further modified for myliobatids).

The Australian stingarees, Trygonoptera and Urolophus, are united as a monophyletic Urolophidae based on the enlarged optic nerve foramen (character 8) and the greatly sinuous external margin of the mesopterygium (apparently fused to radials, character 30). Urotrygonidae, containing amphi-American stingrays with a caudal fin (Urobatis and Urotrygon), is monophyletic based on the branched extremities of the subpleural canal (character 1), dorsal fossa on scapular process (character 26, independently acquired elsewhere), and spiracular tentacle (character 42). South American freshwater stingrays (Potamotrygonidae) share the following unambiguous evolutionary novelties: a greatly extended median prepelvic process (character 31), loss of urea retention (40), and reduction of the rectal gland (41) (note that modifications of the suborbital components of the infraorbital lateral-line canal [character 3] may be optimized at this node as well; fig. 44).

Discussion and Comparison to Previous Stingray Phylogenies

Some of our monophyletic components are present in previous phylogenies of stingrays, including Gymnuridae + Myliobatidae (and two additional clades within Myliobatidae: Aetobatus + (Rhinoptera + Mobula)), South American freshwater stingrays (Potamotrygonidae), Potamotrygon + Plesiotrygon, Urotrygonidae (Urotrygon + Urobatis), and a large clade that includes all stingrays except Hexatrygon. The relationships within the large monophyletic group comprising potamotrygonids, “dasyatids,” gymnurids, and myliobatids, which contains most extant species of stingrays, differ substantially from the relationships within this clade advocated by Nishida (1990), Lovejoy (1996), and McEachran et al. (1996) (their phylogenetic schemes are shown in fig. 46). The clade †Heliobatis + (Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae))) and the group comprising dasyatid genera + (Gymnuridae + Myliobatidae) are novel to our analysis, not being present in previous stingray phylogenies. The node Urolophus + Trygonoptera (Urolophidae) is also unique, as Trygonoptera was not included as a separate terminal by Lovejoy (1996) or McEachran et al. (1996), and Nishida (1990) included only Trygonoptera testacea, the type-species of Trygonoptera, in his study (as Urolophus testacea). Our phylogenetic analysis is also the first to include both fossil and extant stingrays, and includes more informative characters (44 with potential for grouping within stingrays; Lovejoy's matrix contains 33 characters that vary within stingrays, whereas McEachran et al.'s phylogeny includes 35 with potential to unite stingray genera; note that McEachran et al.'s analysis was broader in scope, designed to clarify relationships among extant batoids). Molecular approaches to the phylogeny of stingrays are still in their infancy, and some recent results appear highly controversial (such as Gymnura forming a monophyletic group with Potamotrygon; Dunn et al., 2003).

Our strict consensus tree has more components in common with the stingray relationships presented by Lovejoy (1996) and McEachran et al. (1996) than with the phylogeny of Nishida (1990) (cf. figs. 43 and 46). Comparison with Nishida's phylogeny is difficult because his tree was not generated by computerized phylogenetic procedures, and when implemented by Hennig86 the results are entirely at odds with the phylogeny he advocates (Carvalho, 1996a; McEachran et al, 1996). Nonetheless, the tree he favors shares just one node in common with our phylogeny, Gymnuridae + Myliobatidae, which is supported by similar features (see above). None of these authors supports a monophyletic Dasyatidae (neither do we, but see appendix 2). However, our resulting phylogeny also differs in many respects from the cladograms of Lovejoy (1996) and McEachran et al. (1996), which are very similar to each other. Although we included characters discussed by these workers, some of these are coded differently (e.g., characters 13, 34, 35 and 44) and contribute to differences in tree topology. Other characters included in the analyses of Lovejoy (1996) were excluded from consideration. These include the degree of lateral projection of the lateral stays of the first synarcual (Lovejoy's character 21) and the level of first segmentation of the propterygium (his character 25). We found it difficult to quantify both characters, which varied intragenerically to a large degree according to our observations (especially in Dasyatis and Urolophus). Other excluded characters are autapomorphic for individual genera (5 of Lovejoy's 39 characters). We also excluded from consideration the direct insertion of the depressor rostri muscle on the pectoral propterygium (i.e., inserting without an aponeurosis; Nishida, 1990), as it only occurs in gymnurids and Manta, and the latter genus was not included in our matrix.

Relationships among myliobatid genera in our phylogeny are in accordance with the results of Lovejoy (1996) and McEachran et al. (1996), but contrary to the topology advocated by Nishida (1990), who favored a restricted Myliobatidae including only Myliobatis + (Aetobatus + Aetomylaeus), which in turn is sister-group to Rhinoptera + Mobulidae. Most of the characters used by Nishida for this arrangement are optimized differently on our trees (we did not include the “hyomandibular accessory cartilage 1” because it is highly variable within genera), supporting the group Myliobatis + (Aetobatus + (Rhinoptera + Mobula)). Furthermore, the characters that support Aetobatus + Aetomylaeus in Nishida's phylogeny also occur for Rhinoptera + Mobulidae, so that on Nishida's tree more than one optimization exists for these features (they could also be interpreted as synapomorphies for all pelagic stingrays, lost in Myliobatis).

The major difference between our phylogeny and the trees of Lovejoy (1996) and McEachran et al. (1996) is the recognition of a large component uniting all dasyatid genera and Gymnuridae + Myliobatidae. The position of the clade Gymnuridae + Myliobatidae as a monophyletic group nested within successive higher level taxa of benthic stingrays is common to all stingray phylogenies generated to date (cf. appendix 2). Differences between these phylogenies and ours are the result of the inclusion of different taxa and characters, as well as alternative interpretations for certain features. Lovejoy (1996) and McEachran et al. (1996) placed Dasyatis (including Indo-west Pacific Himantura in Lovejoy, 1996) as sister-group to Gymnuridae + Myliobatidae, a grouping not at odds with our phylogeny. These authors, however, also support a group comprising Taeniura, amphi-American Himantura, and the Potamotrygonidae, a relationship contradicted by our results. The characters advanced by these authors supporting this relationship were also included in our analyses (relating to the angular cartilages and patterns of insertion of the spiracularis muscle). The results of Lovejoy (1996) concerning potamotrygonid relationships are not verified here not only because of differential coding of certain features, but also because we implemented our analyses without ordering multistate characters, a prerequisite to achieve the topology advocated by Lovejoy. When his analysis is run without ordering multistate characters, the results obtained are somewhat more varied, and amphi-American Himantura is no longer unequivocally supported as the sister-group to potamotrygonids. We have included the characters supporting Lovejoy's clade Taeniura + (amphi-American Himantura + Potamotrygonidae) in our matrix (see characters 14 and 37 above), but these were not sufficient to support their monophyly in any of our equally most parsimonius trees. This holds true even when character 37 (describing the different insertion patterns of the spiracularis muscle) is run as an ordered multistate feature.

Relationships within Potamotrygonidae, however, are resolved identically as in the analysis of Lovejoy (1996), with Potamotrygon and Plesiotrygon as sister-groups, contrary to the hypothesis of Rosa (1985) and Rosa et al. (1987). We concur with Lovejoy's reasoning for excluding two of the characters used by Rosa to unite Potamotrygon and Paratrygon (modal numbers of pectoral fin radials and pelvic fins covered by pectoral disc). These characters are difficult to code, and more comparative data are needed before numbers of pectoral fin radials can be confidently used. The proximal fusion of the first and second ceratobranchials also used by Rosa (1985) and Rosa et al. (1987) to unite Potamotrygon to Paratrygon occurs in other taxa, such as gymnurids and many myliobatids, and in at least one species of Urotrygon and Urolophus (Miyake and McEachran, 1991). However, this character is a putative synapomorphy of both genera (requiring homoplasy), but it must be included in a more comprehensive species-level phylogeny of stingrays.

Our phylogenetic analysis reinforces that tree topology can be dramatically altered with only small modifications in character coding, or with the addition of characters or terminals, when taxa exhibit such high levels of homoplasy. For instance, scoring character 44 (nasal curtain) identically to the coding given in McEachran et al. (1996; with Plesiobatis displaying the primitive condition of having a shortened nasal curtain) resolves Plesiobatis as sister-group to Urolophus + Trygonoptera and alters resolution elsewhere on our phylogeny (†Asterotrygon is then placed in a polytomy at the node with †Heliobatis, potamotrygonids, dasyatids, gymnurids, and myliobatids). We have coded Plesiobatis differently but accurately in our assessment (with a nasal curtain reaching very close to the mouth; also J.D. McEachran, personal commun.). Another example of this sensitivity emerges if pelagic stingrays are scored as a single terminal (as Myliobatidae), as opposed to the inclusion of various myliobatid genera in the matrix (as in our final matrix of table 5; this experimental analysis is presented in appendix 2). This form of “groundplan” (as opposed to “exemplar”) coding seriously alters the relationships among benthic genera that are more distantly related to pelagic stingrays (i.e., that are not their direct sister-groups), perhaps unexpectedly (for theoretical considerations, see Bininda-Emonds et al., 1998; Wiens, 1998, 2000; Kornet and Turner, 1999; Prendini, 2001; Simmons and Geisler, 2002). Elasmobranchs are notorious for elevated levels of character conflict (see, for example, the phylogenies of Naylor, 1992; Shirai, 1992; McEachran et al., 1996; McEachran and Dunn, 1998; cf. Carvalho, 1996a), and stingrays are no exception. Coding must be done cautiously if taxa are to be combined into a single family-level entry (which should be avoided if possible), either because of inadequate sampling or for the sake of simplifying tree construction, as this single entry will, undoubtedly, contain a high degree of character variation (taxonomic polymorphism in the sense of Nixon and Davis, 1991). This is precisely the case of pelagic stingrays in the analysis of appendix 2 regarding the hyomandibular-Meckelian ligament. When the single terminal Myliobatidae was coded as having the ligament in that analysis (appendix 2), on the basis of this being the inferred ancestral state (“IAS coding” of Rice et al., 1997), resolution was substantially decreased in the strict consensus, and the number of equally most parsimonious trees increased considerably. Similarly, adding to our matrix (table 5) only two characters with potential to unite dasyatid genera would be enough to support a monophyletic Dasyatidae in all minimum-length trees obtained. A monophyletic Dasyatidae, in turn, dramatically affects the relationships of other stingray genera in the analysis, and the outcome would be more similar to the topology presented in appendix 2. To date, however, there have been no unambiguous characters discovered and correctly included in a phylogenetic analysis to support dasyatid monophyly (but see appendix 2).

Removing the Green River fossils from our matrix reduces resolution considerably. Parsimony analysis of the matrix in table 5 without †Asterotrygon and †Heliobatis resulted in 273 trees (length = 80 steps, CI = 0.70, RI = 0.82), and the strict consensus (length = 86 steps, CI = 0.65, RI = 0.78) does not support a monophyletic Urolophidae or the large clade Potamotrygonidae + (“Dasyatidae” + (Gymnuridae and Myliobatidae)), the major contribution to stingray phylogeny presented in this study; dasyatid genera are placed in a polytomy with potamotrygonids. The other components of our phylogeny in figure 43 remain unaltered. Successive approximations weighting of this modified matrix resulted in 123 trees (length = 474 steps, CI = 0. 88, RI = 0.93), and the strict consensus of these (length = 478 steps, other tree indices remain unchanged) adds two clades: (1) Urolophidae + Pteroplatytrygon; (2) Dasyatis, Himantura sensu stricto + (Gymnuridae + Myliobatidae). The former clade (1) is highly counterintuitive; amphi-American Himantura, Taeniura and potamotrygonids remain in a polytomy. The impact of †Heliobatis on the relationships among extant stingrays is greater than that of †Asterotrygon; when only †Asterotrygon was removed, relationships were identical to those presented in figure 43 (derived from the matrix in table 5, with both fossil taxa included). Implementing the analysis solely without †Heliobatis produced the highly modified and more ambiguous results described above. The influence on extant stingray phylogeny of the Green River stingrays is therefore substantial, but differs for both fossil genera. The more dramatic changes caused by the inclusion of †Heliobatis stems from its additional missing entries in the matrix (†Heliobatis has three more unknown entries than †Asterotrygon); both taxa differ otherwise in only one additional feature (dorsal fin, character 33). The different influence of †Asterotrygon and †Heliobatis on stingray phylogeny is therefore determined by this particular combination. This is further evidence that the impact a fossil taxon may have on a phylogeny of living forms depends more on its specific combination of features than on any other factor (Gauthier et al., 1998; Donoghue et al., 1989; Huelsenbeck, 1991; Novacek, 1992; Graybeal, 1998; O'Leary and Geisler, 1999 [cf. Naylor and Adams, 2001]; Kearney, 2002; contra Patterson, 1981; Rosen et al., 1981).

Systematic Issues Within Myliobatiformes

The hyomandibular-Meckelian ligament requires further comparisons within the Myliobatidae. The ligament is present in three of the seven genera of pelagic stingrays (Myliobatis, Aetobatus, and Rhinoptera), but its presence in Myliobatis is based on conflicting evidence (Garman, 1913; Nishida, 1990). The condition in Pteromylaeus is unknown. Lovejoy (1996) excluded this character from his analysis, but he was influenced by the ambiguous account given by Nishida (1990) concerning the condition in Plesiobatis. The hyomandibular-Meckelian ligament was included in the matrix of McEachran et al. (1996), who correctly coded Plesiobatis as having the ligament, the condition present in benthic stingrays except Hexatrygon and gymnurids.

Dissection of various stingrays revealed that the spiracularis muscle is more complex than was indicated in previous interpretations (Lovejoy, 1996; McEachran et al., 1996), including our own observations summarized in the matrix (table 5), which largely follows Lovejoy. We included the character somewhat tentatively to see how it would affect the relationships of Taeniura, amphi-American Himantura, and potamotrygonids, hypothesized by Lovejoy as forming a monophyletic group (it does not support this node in our analysis, even when run as ordered; see above). The spiracularis of these taxa (except Paratrygon) is indeed robust and projects posteroventrally, more so than in most other stingrays examined. However, Pteroplatytrygon and some species of Dasyatis may have the spiracularis inserting posterior to the lower jaws and somewhat continuous with the depressor hyomandibularis muscle as well (also Kesteven, 1942; Miyake, 1988; see character description above). Dissection of additional taxa is still needed. Deactivating this character in our matrix does not alter the strict consensus, but it reduces the number of equally most parsimonious trees to 28 (length = 79 steps, CI = 0.68, RI = 0.82; strict consensus tree: length = 79 steps, CI = 0.65, RI = 0.79).

Previous authors have advocated that Dasyatis and Himantura may not be monophyletic genera, a proposition that we have only partially tested in our analysis (Compagno and Roberts, 1982; Last and Stevens, 1994; Lovejoy, 1996; Rosenberger, 2001b). Himantura, on morphological grounds alone, should be divided into separate genera, one for amphi-American species (H. schmardae and H. pacifica) and one for Himantura sensu stricto for Indo-west Pacific species (Lovejoy, 1996). Both amphi-American species share unique specializations of the dermal skeleton, disc proportions, and shape, and perhaps even of the caudal sting (Carvalho and Lovejoy, in prep.). They were consequently coded as two separate terminals in our matrix (following Lovejoy, 1996), and they do not form a monophyletic Himantura on any of the minimum-length trees obtained. There are characters that may potentially unite species of Himantura sensu stricto (Indo-west Pacific species) as a monophyletic unit, such as arrangement of their nasal capsules (which are anteroposteriorly abbreviated), anterior projection of the scapular process (e.g., Nishida, 1990: fig. 30f), and a narrow, anteriorly rounded pelvic girdle. We have observed these features in various species of Himantura, but they vary among species to a certain degree, and in order to verify this claim they must be included in a more comprehensive species-level matrix.

The situation concerning Dasyatis is more complex. This genus presently includes from 35 to 40 species (Compagno, 1999), many of which are poorly known morphologically. Rosenberger (2001b) attempted the first cladistic test of its monophyly. Her phylogenetic analysis included 13 Dasyatis species (15 if both Pastinachus sephen and Pteroplatytrygon violacea are considered in Dasyatis, as she did), in addition to five species belonging to the genera Gymnura, Taeniura, amphi-American Himantura, Himantura sensu stricto, and Urobatis. Her resulting phylogeny nests H. gerrardi, Gymnura micrura, Pteroplatytrygon, and Pastinachus within Dasyatis. Gymnura micrura clearly does not belong in Dasyatis, as would have been concluded had more species of Gymnura been coded and their many generic synapomorphies included in the matrix. We included Pteroplatytrygon as a separate entry in our analysis because it differed in relation to the postorbital process of the neurocranium (character 9), a feature that did not vary among species of Dasyatis examined (Pteroplatytrygon is also considered valid in recent compilations, e.g., Compagno, 1999; McEachran and Carvalho, 2002). Rosenberger's analysis was partially based on features that are difficult, if not impossible, to divide into discrete states because their variation is either too subtle or continuous (cf. Rae, 1998) (such as her character 1, snout length less than 25% of disc length [state 0], greater than 25% [1]; character 14, midregion of frontoparietal fontanelle greatly constricted and narrow [0], moderately constricted [1], wide [2]; character 15, anterior margin of frontoparietal fontanelle rounded [0], slightly rounded [1], straight [2]; character 24, mesopterygium wide [0], narrow [1]). The frontoparietal fontanelle (her character 14), for example, may be greatly constricted and wide (e.g., in Potamotrygon; fig. 34A), and the difference between a wide and narrow mesopterygium in many species of Dasyatis may be trivial or somewhat arbitrary (the mesopterygium can even vary slightly in length and width on both sides of the same specimen, e.g., Dasyatis pastinaca AMNH 32796). Other characters vary among specimens within certain species according to our data (her character 27, number of metapterygial segments; e.g., in Dasyatis margarita, D. zugei, Urobatis jamaicensis; of course, this does not necessarily imply that they vary among Rosenberger's terminals) or they are simply not very clear (her character 31, “posterior portion of pelvic gridle arch”). Rosenberger's conclusions regarding the validity of Pteroplatytrygon and Pastinachus may eventually prove to be correct, but the former genus did not pair with Dasyatis in any of our minimum-length trees, and there is even a hint of evidence that it may be more closely related to the clade Gymnuridae + Myliobatidae (see comments on swimming patterns below). Rosenberger's analysis is a step in the right direction, but in addition to complementing her matrix with more species, morphological characters that are more reliable at the species level must be included, and these have been highly elusive so far (i.e., characters that are constant within species but that vary among them).

Subdividing the larger benthic genera into geographical components, if these are demonstrated to be monophyletic (as is the case with amphi-American Himantura), may alleviate difficulties in coding generic-level matrices, as an all-inclusive species-level matrix is still too prohibitive at present. This is precisely what has transpired in the systematics of skates, in which very large, traditional taxonomic units (e.g., Raja) have been subdivided into putative monophyletic groups that were previously given subgeneric ranking. These “subgenera” are now recognized as separate genera that are geographically more restricted (McEachran and Miyake, 1990; McEachran and Dunn, 1998). Geographically restricted monophyletic subunits may exist within Dasyatis, but this hypothesis, contradicted by Rosenberger (2001b), remains to be fully explored.

Our observations indicate that Taeniura is probably monophyletic (note that our matrix did not test Taeniura monophyly), as all three species have the first (or anterior) pair of hypobranchials conspicuously cleaver- or club-shaped. Species of other stingray genera may have hypobranchials in which the anterior portion is wider than the posterior half, but in Taeniura the anterior portion is conspicuously broad, with the broad segment extending posteriorly to at least one-half of hypobranchial length (therefore extending posteriorly farther in comparison to other genera in which the first hypobranchial has a broad anterior segment). The anterior hypobranchials also have slightly concave internal (medial) margins in Taeniura. These elements are particularly similar in Taeniura lymma (AMNH 44077) and T. grabata (MNHN 1989–1793). The anterior hypobranchials of T. meyeni are also markedly broad (Nishida, 1990) and resemble the hypobranchials of the other species of Taeniura more than those of other stingrays. More specimens of T. meyeni are needed for further comparisons. Other features that may be indicative of Taeniura monophyly are the shape of the basihyal (uniquely anteroposteriorly extended, about two-thirds as long as wide), the anterior projection of the hyomandibulae (this process is set at an angle to the main ramus of the hyomandibula, accommodating the hyomandibular-Meckelian ligament; marked with an asterisk in fig. 35A), and the arrangement of the ceratobranchials (in which the forward extensions that contact the anterior ceratobranchial element are particularly slender and elongated). These features require confirmation on additional specimens; they may be autapomorphies of Taeniura lymma, synapomorphies of Taeniura, or prove unreliable due to intraspecific variation.

Trygonoptera Müller and Henle, 1841 has been considered a junior synonym of Urolophus Müller and Henle, 1831 by some previous authors (e.g., Whitley, 1940; McEachran, 1982; Nishida, 1990), but others have recognized it as a valid genus (Bigelow and Schroeder, 1953; Last and Gomon, 1987; Last and Stevens, 1994; Compagno, 1999; Last and Compagno, 2000). Although they can be distinguished on the basis of external morphology (primarily by the fleshy nasal lobe in Trygonoptera), skeletal features further separate both genera, particularly those from the neurocranium. Trygonoptera has the infraorbital lateral-line canal passing between the postorbital process and the more anterior lateral projection (supraorbital process), while Urolophus has the canal passing through a large foramen in the postorbital process per se as in most myliobatids (character 9 in our analysis). The dorsal internasal septum is also somewhat distinct in both genera, being relatively wider in Trygonoptera (similar to Plesiobatis; however, the phylogenetic significance of the dorsal internasal septum has yet to be fully explored). Both genera also differ in character 26 (fossa on dorsal scapular region, present in Trygonoptera but absent in Urolophus). Furthermore, in Trygonoptera the supraorbital process is particularly elongated compared to other genera in which it is present (genera with state 0 for this character, listed above). We have observed this feature in all species of Trygonoptera examined (in five of the six species recognized by Last and Stevens, 1994) and accept it as evidence of its monophyly along with the fleshy lateral nasal lobes. Yearsley (1988) arrived at similar conclusions and presented more (and different) skeletal evidence distinguishing both genera.

The degree of development of the tail-folds may still be of some value in delimiting groups within dasyatids, but a comprehensive species-level phylogeny is necessary to evaluate this character more thoroughly, and more information is needed for many stingray species before this can be attempted. Our analysis demonstrates that lack of tail-folds occurs independently for at least four stingray groups (Gymnuridae + Myliobatidae, Paratrygon, amphi-American Himantura, and Himantura), and that possession of only ventral tail-folds appears, when mapped on our trees, independently at least twice (for Taeniura and Plesiotrygon) but probably more times because certain species of Dasyatis also lack dorsal tail-folds (presenting only a dorsal keel in its place; Garman, 1913; Bigelow and Schroder, 1953; Nishida and Nakaya, 1990). As there is much variation in the pattern of tail-folds within Dasyatis (some authors have used Amphotistius Garman, 1913 for species with both dorsal and ventral tail-folds; e.g., Paxton et al., 1989), this character cannot be fully evaluated until a species-level phylogeny is available. In fact, because many characters appear, disappear, and reappear in stingray evolution, amounting to high levels of character variation within genera, a species-level phylogeny is necessary to fully evaluate their potential for grouping (e.g., arrangement of orbital foramina, degree of development of lateral stays of first synarcual, extent of internasal septum, patterns of subdivision of the propterygium).

The phylogenetic analysis presented here allows for some consideration of swimming in stingrays, which is accomplished by the following patterns (or modes, Rosenberger, 2001a): undulatory (wavelike, with anterior to posterior motions of the disc), oscillatory (birdlike, with flapping of the disc), or a combination of these (the “intermediate” condition of Rosenberger, 2001a). Even though the extreme manifestations of “wavelike” and “birdlike” swimming are quite distinct (such as between a skate and a manta ray, respectively), there is a large gray zone of in-between conditions (the undulation/oscillation continuum of Rosenberger, 2001a), rendering the divisions into discrete states somewhat arbitrary. Nonetheless, adding this character to our matrix (i.e., coding swimming patterns) resulted in 107 equally most parsimonious trees (length = 85 steps, CI = 0.68, RI = 0.82) and a strict consensus (length = 89 steps, CI = 0.65, RI = 0.79) identical to the tree obtained without this character (fig. 43, from the matrix in table 5). However, after successive approximations weighting was applied to this dataset (with swimming as a character, resulting in 33 trees: length = 489 steps, CI = 0.89, RI = 0.93), the resulting consensus tree (length = 491 steps, CI = 0.88, RI = 0.93) supported the placement of Pteroplatytrygon as sister-group to the clade containing Gymnuridae and Myliobatidae. We obtained this result with weighting when coding Pteroplatytrygon identical to Gymnura (i.e., with state 1, intermediate) or to myliobatids (state 2, oscillatory). Even though optimization-dependent, this indicates that the oscillatory swimming mode indicative of a pelagic lifestyle may have arisen only once within stingrays (with gymnurids subsequently settling back into a benthic niche), and not twice as advocated by Rosenberger (2001a). Rosenberger's (2001a) study of swimming patterns within stingrays revealed that there may be additional characters for Pteroplatytrygon and myliobatids (and gymnurids), but these were not included in our modified matrix because they either appear to be correlated to the oscillatory swimming mode (and therefore not independent) or we could not yet confirm them in dissections (e.g., greater proportion of red muscle fibers in the disc as opposed to white [red fibers are more important for slow, steady swimming in pelagic forms], angle of articulation between scapular processes and suprascapulae).

Classification of Myliobatiform Genera

Successive approximations weighting (Farris, 1969; Carpenter, 1988, 1994; Fitzhugh, 1989) further resolves one of the polytomies of the strict consensus, resulting in 20 equally most parsimonious trees (strict consensus in fig. 47). The relationships of †Asterotrygon are not altered, but three (Pteroplatytrygon, Dasyatis, Himantura) of the five dasyatid genera form an additional node with the Gymnuridae + Myliobatidae clade, which conforms to the results of Lovejoy (1996) and McEachran et al. (1996). Implementing the weighting scheme of Goloboff (1993; using implied weights, with his program Piwe) produces results identical to those obtained from successive approximations weighting. The following sequenced classification of stingrays is derived from our study of stingray phylogeny:

Order Myliobatiformes

Suborder Hexatrygonoidei

Family Hexatrygonidae (Hexatrygon Heemstra and Smith)

Suborder Myliobatoidei

Suborder Myliobatoidei incertae sedis (†Asterotrygon, n.gen.)

Superfamily Urolophoidea, new combination

Family Urolophidae (Urolophus Müller and Henle, Trygonoptera Müller and Henle)

Superfamily Plesiobatoidea

Family Plesiobatidae (Plesiobatis Nishida)

Superfamily Urotrygonoidea, new combination

Family Urotrygonidae (Urotrygon Gill, Urobatis Garman)

Superfamily Myliobatoidea, new combination

Family Heliobatidae (†Heliobatis Marsh)

Family Potamotrygonidae (Potamotrygon Garman, Paratrygon Duméril, Plesiotrygon Rosa, Castello, and Thorson)

Genera incertae sedis (Dasyatis Rafinesque [including Urolophoides Lindberg], Himantura Müller and Henle [Indo-west Pacific species], Pastinachus Rüppell, Pteroplatytrygon Fowler, Urogymnus Müller and Henle, Taeniura Müller and Henle, “amphi-American Himantura”)

Family Gymnuridae (Gymnura Kuhl, Aetoplatea Valenciennes)

Family Myliobatidae (Myliobatis Cuvier, Aetomylaeus Garman, Pteromylaeus Garman, Aetobatus Blainville, Rhinoptera Kuhl, Mobula Rafinesque, Manta Bancroft)

Origin of the Green River Stingrays

None of the equally most parsimonious trees recovered in our phylogenetic study allows for a single invasion of Fossil Lake by an ancestral stingray form which subsequently underwent speciation to result in the two known stingray genera (see area cladogram in fig. 48). In other words, †Asterotrygon and †Heliobatis do not from a monophyletic unit in any of the resulting trees (figs. 43, 45), nor do they form successive sister-groups to the node Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae)), a topology that would still allow for a single invasion of Fossil Lake by an ancestral stingray taxon. Quite the contrary, the Green River stingrays are phylogenetically placed far apart from each other in all of the minimum-length trees recovered. It may seem counterintuitive that more than one stingray lineage entered and successfully adapted to the freshwater environment of Fossil Lake, especially given that the lake is hypothesized to have existed for “only” some 5 million years (which is not insignificant if compared to the much more species-diverse and younger Lake Victoria in eastern Africa, however; cf. Verheyen et al., 2003), and that Fossil Lake is considered to have been of modest size throughout its history by most estimates compared to the other extinct Green River lakes (even though its size fluctuated through time, Fossil Lake was never much greater than perhaps 70 km north–south by 35 km west–east; Grande and Buchheim, 1994). But the extensive actinopterygian fauna of Fossil Lake is by no means monophyletic either, nor is it expected to be. †Asterotrygon and †Heliobatis are presently unknown from the other extinct lakes of the Green River lake complex (Lake Uinta and Lake Gosiute), so their presence in Fossil Lake cannot be considered a remnant of a previously more widespread continental distribution. The presence of stingrays in the Green River Formation is therefore more similar to the separate freshwater colonizations by endemic living species of Dasyatis and Himantura in Africa, Australia, and southeast Asia, and it is not analogous to the single (monophyletic) potamotrygonid radiation in South American freshwaters.

A phylogenetic analysis of our matrix (table 5) with †Asterotrygon and †Heliobatis constrained as a monophyletic unit resulted in 58 minimum-length trees that are only a single step longer than our original trees summarized in figures 43–45, 47 (this analysis was run in Nona; length = 83 steps; CI = 0.67, RI = 0.81). However, resolution in the strict consensus is dramatically decreased in this constrained analysis, with only a few nodes present: all stingrays except Hexatrygon, Urotrygonidae, Green River stingrays, Potamotrygonidae, and a node uniting Gymnuridae + Myliobatidae; all other stingray genera form a large polytomy together with all of these clades. The only character that is optimized as uniting †Asterotrygon and †Heliobatis in the strict consensus of this constrained analysis is the absence of the dorsal fossa of the scapular process, which is hypothesized to have been lost in their common ancestor (i.e., its presence united all stingrays except Hexatrygon in the constrained analysis). Evidence for the monophyly of the Green River stingrays is consequently meager (individual vertebrae extending to the distant tip of the tail is another putative synapomophy, but is not optimized in this fashion in all minimum-length trees obtained in this analysis). We are therefore confident, given our present understanding of their morphology, that the genera of Green River stingrays are not sister-groups, which unequivocally discards the possibility that they evolved within Fossil Lake from a common ancestor. Note that their phylogenetic positions (figs. 43, 48) still allow for a single invasion of Fossil Lake (such as through a common biogeographical event), but both stingray genera (or their direct ancestors) were phylogenetically separated when the colonization of Fossil Lake occurred.

The phylogenetic placement of †Asterotrygon close to the base of our strict consensus tree (fig. 43), or its placement as the sister-group to urolophids (Urolophus + Trygonoptera) in 21 of the original 35 minimum-length trees (fig. 45C–E), is an indication of a strong Indo-west Pacific association. Both Hexatrygon and Plesiobatis are widely (but sporadically) distributed in the Indian and Pacific Oceans (Last and Stevens, 1994). Urolophids (Urolophus, Trygonoptera) are somewhat more restricted, occurring off southern Japan and Korea (East China Sea) and in the South China Sea, but they are more diverse and common around Australia (Last and Stevens, 1994). Because the most basal living stingrays are all Indo-west Pacific (with more or less restricted distributions), and †Asterotrygon is nested among them in our phylogeny (figs. 43, 45; area cladogram in fig. 48), it is reasonable to infer that an Indo-west Pacific relationship is implied by †Asterotrygon. This historical relationship is even more geographically restricted in the trees in which †Asterotrygon forms a clade with urolophids (the majority of our equally most parsimonius trees). This hypothesis would not be contradicted if stingrays even more basal to Hexatrygon are discovered.

A transoceanic Indo-Pacific historical association has also been recognized for other Green River Formation fishes (Grande, 1985, 1989, 1994; Li et al., 1997), as well as for its fossil plants (MacGinitie, 1969). In fact, most Green River fishes that have been included in phylogenetic studies indicate a trans-Pacific relationship (Grande, 1985, 1994), as corroborated by independent studies of the Polyodontidae (†Crossopholis; Grande and Bemis, 1991) and the teleosts Hiodontidae, Osteoglossidae (†Phareodus; Li et al., 1997), Pellonulinae, †Ellimmichthyidae (= †Paraclupeidae; Chang and Grande, 1997), Catostomidae, and, to a lesser extent, the Percopsidae (†Amphiplaga; Patterson and Rosen, 1989; cf. Murray and Wilson, 1999). This trans-Pacific component has been demonstrated, therefore, for groups that are strictly freshwater (e.g., paddlefishes, Polyodontidae, known only from freshwater or freshwater deposits in China and North America; Grande and Bemis, 1991) as well as from those that are primarily marine (such as stingrays). The area relationships imposed by the putative †Asterotrygon + Urolophidae clade recovered in most of our minimum-length trees is congruent with area relationships among some species of †Phareodus, in which the Australian †P. queenslandicus is considered to be the sister-group to the Green River †P. encaustus (Li, 1994; Li et al., 1997).

The phylogenetic position of †Heliobatis, on the other hand, is indicative of a more recent historical affinity with the Americas (fig. 48), possibly (but not necessarily, see above) indicating an independent stingray invasion of Fossil Lake. The clade basal to †Heliobatis in our stingray phylogeny (fig. 43) is Urobatis + Urotrygon, and both genera are distributed in North and Central America with a few species reaching as far south as northern South America (Meek and Hildebrand, 1923; Bigelow and Schroeder, 1953; Miyake and McEachran, 1986; McEachran, 1995). The node above †Heliobatis in our tree is divided into two clades, one containing species distributed in South American freshwaters (Potamotrygonidae), and another, the most species-diverse of all stingray lineages (“Dasyatidae” + (Gymnuridae + Myliobatidae)), with species presently occurring in most temperate and tropical marine regions, as well as some eight exclusively freshwater species (in the genera Dasyatis and Himantura in Africa, southeast Asia, and Australia). Our phylogeny therefore is congruent with the hypothesis that the largest radiation of species of stingrays (the clade “Dasyatidae” + (Gymnuridae + Myliobatidae)) originated only after an American stingray lineage was established (fig. 48).

This early American stingray component may have shared an ancestry with the species of stingrays described from Cretaceous and Paleocene formations from central and eastern North America. Many of these stingrays inhabited the Late Cretaceous Seaway, a large inland sea that occupied an immense trough in western North America that at one point extended from present-day Gulf of Mexico to Alaska, with a maximum width of some 2000 km (Smith et al., 1994; cf. Funnell, 1990; see fig. 52). These stingray species are known from isolated teeth, including forms from the early Late Cretaceous (Cenomanian) of Texas (Meyer, 1974) and Late Cretaceous of Wyoming (Estes, 1964) and Texas (Welton and Farrish, 1993). The stingrays described from the Late Cretaceous of New Jersey (Monmouth group; Cappetta and Case, 1975) occurred at the opposite (eastern) margin of the North American continent during the Maastrichtian (fig. 52), but represent a very similar if not identical stingray fauna, as this coastline was probably contiguous with the Late Cretaceous Seaway over central North America. The Paleocene stingrays from the Cannonball Formation, which inhabited the last marine incursion (the Cannonball Sea) into central North America (Cvancara and Hoganson, 1993), may also be related to this early American stingray lineage. These stingrays (known only from isolated teeth) include two species of Dasyatis (note that one tooth referred by Cvancara and Hoganson [1993: fig. 3, v–x] to their new species †D. concavifoveus probably represents a gymnurid), †Hypolophodon sylvestris, and a myliobatid (see also Best, 1987). Unfortunately, it is not possible to compare any of the above taxa in a meaningful way to the Green River stingrays (and more precisely to †Heliobatis) to determine if they share a more recent common ancestry. However, it is certainly conceivable that such a relationship existed, which would indicate that an ancestral stingray taxon originally present in the Late Cretaceous Seaway, or in some other marine incursion into North America, is historically correlated with †Heliobatis of Fossil Lake.

Origin of the South American Freshwater Stingrays

The origin of the South American freshwater stingrays (Potamotrygonidae) has been the subject of much debate and has consequently become a benchmark problem in historical biogeography (Brooks et al., 1981; Rosa, 1985; Brooks and McLennan, 1993; Brooks, 1995; Lovejoy, 1996, 1997; Hoberg et al., 1998; Marques, 2000). The discussions concern primarily the sister-group of the Potamotrygonidae and the nature of the data used to infer potamotrygonid origin or affinity (for a more detailed summary of the different points of view, see Lovejoy, 1996, 1997). The theory advanced by Lovejoy (1996, 1997) and Lovejoy et al. (1998), that potamotrygonids originated from a common ancestor with amphi-American Himantura from the Caribbean Sea by way of an early Miocene epicontinental seaway (the “proto-Caribbean/Himantura” hypothesis), was recently given support by a molecular phylogenetic analysis of parasites and their potamotrygonid hosts by Marques (2000). This contrasts with the earlier theory of Pacific origin of the family by means of a phylogenetic relationship with “urolophids” (= Urotrygonidae), with the potamotrygonid common ancestor having been confined to freshwater as a result of the orogeny of the Andean Cordilleras, an ongoing process that started in the Late Cretaceous some 90 million years ago (Brooks et al., 1981; Rosa, 1985; Brooks and McLennan, 1993; Brooks, 1995). The parasite data from which much of the latter theory is founded are beyond our concern but have been extensively criticized by Straney (1982), Caira (1990), and more recently by Lovejoy (1997; cf. Hoberg et al., 1998), who provided an arrant discussion of problems pertaining to Brooks et al.'s (1981) data, host/parasite coevolution in general, and the implicit assumptions of their “Pacific/urolophid” hypothesis. This hypothesis is completely contradicted by our phylogenetic results (and previously by those of Lovejoy, 1996) and will not be further explored here (cf. Rosa, 1985). All of the above authors, however, are in agreement that the Potamotrygonidae is monophyletic (even though their different parasite lineages may not be), a crucial point that we fully corroborate and which is congruent with the theory of a single freshwater invasion into South America by the potamotrygonid ancestor. Our focus on the problem here stems solely from inferring potamotrygonid biogeography on the basis of their phylogenetic relationships, which in our view takes precedence over data concerning the affinities of their parasites (also Straney, 1982; Lovejoy, 1997). Our stingray phylogeny (fig. 43) has prompted us to reevaluate the origin of the potamotrygonids, coupled with a radically different assessment of their minimum age.

The phylogenetic relationships of potamotrygonids advocated by Lovejoy (1996) have been reviewed above and compared to our results (cf. figs. 43 and 46B). His novel clade Taeniura + (amphi-American Himantura + Potamotrygonidae), the basis of his biogeographic hypothesis for potamotrygonids, was not present in any of our equally most parsimonius trees (even when the multistate character pertaining to the insertion of the spiracularis muscle was run as ordered, a constraint necessary to obtain this clade in his analysis). Instead, we found support for a clade comprising Potamotrygonidae + (Taeniura, amphi-American Himantura, Pteroplatytrygon, Dasyatis, Himantura + (Gymnuridae + Myliobatidae)), with some evidence (implementing the weighting schemes of Farris, 1969 and Goloboff, 1993) for an additional node uniting Pteroplatytrygon, Dasyatis, Himantura + (Gymnuridae + Myliobatidae)).

The amphi-American Himantura + Potamotrygonidae clade permitted Lovejoy (1996) to postulate that potamotrygonids were derived from a freshwater-invading ancestor that was distributed along the northern coast of South America (similar to the present-day ranges of H. pacifica and H. schmardae). The timing of this derivation was established by reference to the earliest putative potamotrygonid fossils known (from the late Miocene of Peru; e.g., Frailey, 1986). This, in turn, led to the inference that an epicontinental (probably Miocene) seaway was the route of the South American invasion by the potamotrygonid ancestor, which subsequently underwent extensive speciation (some 20 species are presently recognized in the family, with as many as six known undescribed forms; Carvalho et al., 2003). Lovejoy's theory also allowed for a vicariant explanation for the origin of the two species of amphi-American Himantura, which occurred as a result of the formation of the Panamanian Isthmus in the Pliocene, after the initial divergence between amphi-American Himantura and the Potamotrygonidae. Lovejoy et al. (1998) further supported this specific marine-incursion hypothesis and gave it a more precise Miocene timeline, by analyzing rates of divergence of the gene-encoding mitochondrial cytochrome b, DNA sequences from which were used to corroborate the amphi-American Himantura + Potamotrygonidae clade. The divergence rate inferred for the cytochrome b gene was calibrated based on the orogeny of both the eastern cordillera of the Andes and the Merida Andes (which took place some 8 million years ago [mya]; Lovejoy et al., 1998). The resulting molecular clock estimate placed the origin of the Potamotrygonidae at between 15 and 23 mya, in early Miocene to late Oligocene times (Lovejoy et al., 1998). These results were further supported by the molecular phylogenetic analysis of Marques (2000), based mostly on mitochondrial rDNA sequences (12S, 16S, CO I, and again cyt-b). Marques' molecular clock estimates placed the origin of this clade at approximately 19 mya, well within the mostly early Miocene range of Lovejoy et al., but his consideration of molecular clock confidence limits allowed for the potamotrygonid original divergence to have occurred from between 6 and 38 mya, as early as the late Eocene (Marques, 2000).

The application of molecular clock estimates as a method to establish dates of evolutionary events is highly controversial, and many researchers consider molecular clocks to be mostly unreliable (review in Hillis et al., 1996). Molecular clocks must be calibrated by benchmark dates from geology, paleontology, or biogeography (Smith, 1992; Lundberg, 1998), and the rate of genetic divergence in a given gene may vary significantly even within a single taxonomic lineage (Knowleton et al., 1993; Hillis et al., 1996). Furthermore, there is much discrepancy among methods used to infer actual amounts of genetic divergence (see references in Lundberg, 1998), and the confidence limits allowed by rate calibrations may, in many cases, be as great as the age of the actual events being dated (Hillis et al., 1996; precisely the situation concerning potamotrygonids described above). Hillis et al. (1996) and Lundberg (1998) provided cogent remarks on the use and accuracy of molecular clocks, which need not be repeated here. It is our view that the fossil record, even though frequently imprecise, will provide better estimates of divergence times even if these are underestimates (minimum divergence dates), as also argued by Lundberg (1998). There are two prerequisites to employ fossils as benchmarks for dating: they must unequivocally possess derived features that permit them to be placed with confidence in a phylogeny, and their geological dating must be relatively accurate.

Fossil stingrays that are pertinent for establishing a minimum age of the potamotrygonid lineage (or for the invasion of the potamotrygonid common ancestor into South American freshwaters) were added to our strict consensus tree, allowing for a more precise consideration of the relative ages of certain stingray lineages. These taxa stem from the Monte Bolca deposits of northeastern Italy and are represented by well-preserved skeletons (figs. 49, 50) that retain characters indicative of their placement in our stingray phylogeny (fig. 51). These stingray taxa have not been incorporated into our phylogenetic analysis because the Monte Bolca batoid fauna is still under investigation, but their phylogenetic positions in figure 51 are well supported by uniquely derived features.

One of the most remarkable stingrays from Monte Bolca is †Promyliobatis gazolae (figs. 49A, 50). This stingray is unquestionably a myliobatid, as evidenced by its teeth that are coalesced into a horizontally expanded, grinding pavement, as in pelagic stingrays except mobulids (which have reversed to having small, individual teeth). The dentition of †Promyliobatis is more similar to the dentition of Myliobatis or Aetomylaeus, with a much wider central tooth articulating laterally with smaller, hexagonal pavementlike teeth (De Zigno, 1885; fig. 50 here). This contrasts with the dentition of Rhinoptera, where the central tooth is only slightly wider than the lateral ones, which are also hexagonal (or with Aetobatus, in which a single, greatly expanded tooth is present in each horizontal row). †Promyliobatis presents the cartilaginous rod in the tail posterior to the caudal stings in place of individual vertebrae (contrary to the Green River stingrays). The caudal stings in †Promyliobatis are located farther posteriorly than in any other stingray, fossil or extant, and because its tail was very long (which is mostly preserved), this stingray had enormous reach with its caudal stings. On our tree, †Promyliobatis may be more closely related to Myliobatis or simply to an unresolved myliobatid (the most conservative hypothesis of relationship); in any case, the placement of †Promyliobatis at the node with pelagic stingrays is very well supported (fig. 51).

†“Dasyatis” muricata (fig. 49B) is also unquestionably a stingray, but its position on the tree in figure 51 is not as well resolved. Because it presents the cartilaginous rod in the tail, it is phylogenetically positioned at least at the node uniting potamotrygonids, “dasyatids,” gymnurids, and myliobatids. The dorsal fossa on the scapular process is difficult to discern in this fossil, but it appears to be present according to our observations, and consequently †“Dasyatis” muricata belongs at the next more derived node (uniting “dasyatids,” gymnurids, and myliobatids; fig. 51). This species clearly does not present any of the derived characters that unite other monophyletic components positioned at lower nodes in our phylogeny (urolophids, urobatids, potamotrygonids); its phylogenetic position is therefore strongly corroborated.

The Monte Bolca Formation of northeastern Italy represents an extinct coral reef embayment of the tropical Tethys Sea that was frequently exposed to the vagaries of fluvial systems and open lagoons (Sorbini, 1983; Landini and Sorbini, 1996). The deposits of the two main fossil-bearing sites, Pesciara and Monte Postale (the third site, Purga di Bolca has not yielded fossil fishes), are firmly dated as early Eocene (Lutetian, NP 14, Discoaster sublodoensis nanoplankton zone; Barbieri and Medizza, 1969; Patterson, 1993; Landini and Sorbini, 1996), even though previous authors have given a slightly older age to these beds (as topmost Ypresian; Blot, 1969, 1980). These discrepancies in dating are minor, however, and the 50 million-year-old early Eocene age of the Monte Bolca deposits is presently not disputed (note that some authors have interpreted the lowermost Lutetian as middle Eocene; Tyler and Santini, 2002).

Our phylogeny (fig. 51) indicates that potamotrygonids must be at least as old as †Promyliobatis gazolae and †“Dasyatis” muricata, taking into account their relative phylogenetic positions (and using a stepwise correlation between nodes, moving “down” the tree as outlined by Lundberg, 1993, 1998). The Monte Bolca stingrays provide benchmarks by which the origin of the Potamotrygonidae can be dated, imposing a minimum age of at least 50 mya to this lineage (early Eocene). This is much older than their Miocene origin proposed by Lovejoy (1996, 1997) and Lovejoy et al. (1998). This specific age correlation with the Monte Bolca stingrays is not possible in Lovejoy's phylogeny (or in the tree of McEachran et al., 1996), due to existence of the node uniting Taeniura, amphi-American Himantura, and Potamotrygonidae, which could have originated well after the divergence of their common ancestor from the main stem of stingray evolution (fig. 46B); moreover, the clade amphi-American Himantura + Potamotrygonidae could only have come into existence at an even later time. Conversely, this inference is not possible in our phylogeny, as potamotrygonids branch out directly from the main stingray evolutionary branch, without any intervening node (fig. 51; note that this is also the case in the analysis presented in appendix 2, in which potamotrygonids would be at least as old as †“Dasyatis” muricata, but not necessarily as old as †Promyliobatis). Potamotrygonids are consequently as old as the main evolutionary stem of stingrays, which is at least 50 million years old as inferred by the Monte Bolca stingrays.

In his overview of the temporal context for the origin and diversification of Neotropical fishes, Lundberg (1998: 55) posed the question, “[c]ould the Potamotrygonidae be that [Late Cretaceous] old?” According to our study, this certainly seems possible. Our phylogeny supports a divergence age for potamotrygonids that is considerably older than previous estimates—at least 12 million years older than the most generous timeframe allowed by Marques (2000), and at least some 27 million years older than Lovejoy et al.'s (1998) hypothesis. A 50 million-year-old Potamotrygonidae is clearly feasible when the morphological similarities between fossil and living elasmobranchs, including stingrays, are taken into account. A conservative morphology per se is not necessarily evidence that a particular lineage is very ancient, but given that the oldest stingrays known from skeletal remains (52 mya) are morphologically very similar to modern ones (a circumstance present in many elasmobranch groups known from even older fossils, e.g., †Pseudorhina from Late Jurassic of Solnhofen and modern Squatina; †Asterodermus from Solnhofen and modern rhinobatids; Woodward, 1889; Saint-Seine, 1949), a significant correlation between age and degree of morphological conservativeness in stingrays, as evidenced by their fossil record, cannot be discarded.

Certain components of the Neotropical ichthyofauna have pre-Miocene remains in South America, and it is believed that by middle or late Miocene the Neotropical fish fauna was essentially modern (Lundberg, 1998). Some of the more ancient groups mentioned below have a historical association with Africa (predating its Late Jurassic to Early Cretaceous separation from South America, some 150 to 110 mya), and therefore present a pattern of area relationships different from that of potamotrygonids, which are clearly not more closely related to a clade of African stingrays. Late Cretaceous (Maastrichtian) Neotropical fossils, mostly isolated fragments, include relatives of Lepidosiren (Schultze, 1991; Gayet and Meunier, 1998), osteoglossids (possibly Aptian; Lundberg, 1998; Gayet and Meunier, 1998), characiforms (Erythrinidae, Serrasalminae; Roberts, 1975; Gayet, 1991; Vari, 1995; Lundberg, 1998; Gayet et al., 2003), and siluriforms (overview in Lundberg, 1998). Callichthyid catfishes (Cockerell, 1925) and a variety of different characiforms (Gayet and Meunier, 1998) from the late Paleocene, and fossils related to Lepidosiren and to the Percichthyidae (a percomorph family) from the Eocene, are also reported from South America (Arratia and Cione, 1996; Lundberg, 1998; cf. Patterson, 1993). These groups are therefore older than 50 million years. The lack of a pre-Miocene fossil record for potamotrygonids should not be interpreted too decisively, as other Neotropical groups that are inferred to have originated much earlier than the Miocene, such as those with an historical African association, also lack a robust pre-Miocene fossil record (many characiforms, e.g., Roberts, 1975; Vari, 1995; doradoid and loricarioid catfishes; see summary in Lundberg, 1998). The Cretaceous stingray fossils that have been described from South America may even share a common ancestry with potamotrygonids, but this position is presently untenable as these fossils consist of very generalized isolated teeth (e.g., Schaeffer, 1963; see references in the Introduction), and the search for shared, derived dental features is still not a standard employed by fossil teeth specialists (e.g., Cappetta, 1975, 1987).

Our stingray phylogeny (figs. 43, 51) and its implied area cladogram (fig. 48) do not permit the formulation of such a clearcut biogeographic scenario as that presented by Lovejoy (1996) and Lovejoy et al. (1998). Because potamotrygonids are the sister-group of a large clade that contains most of the modern stingray diversity, it is more difficult to accurately pinpoint or restrict to a small area the distribution of their common ancestor. This contrasts with Lovejoy's hypothesis (and to a lesser extent with the “Pacific/urolophid” theory forwarded by Brooks et al., 1981), in which the common ancestor of the amphi-American Himantura + Potamotrygonidae clade was candidly inferred to have been distributed in the proto-Caribbean Sea. Biogeographic hypotheses that deal with “deeper” phylogenetic divergence events, that is, those hypotheses based on relationships among higher level taxa or more inclusive monophyletic groups, will usually encounter this difficulty, especially regarding relatively widespread marine forms (our stingray phylogeny is a case in point). Of course, if the postulated phylogenetic relationships are inaccurate, then a specific area relationship is not more probable just because its biogeographic implications appear to be more alluring. Our phylogeny (fig. 43), in fact, is not at odds with the mechanism proposed by Lovejoy (1996) and Lovejoy et al. (1998) to explain the presence of the potamotrygonid common ancestor in South America (through a marine introgression). Although we refute that amphi-American Himantura is the sister-group to potamotrygonids, the common ancestor of the clade Potamotrygonidae + (“Dasyatidae” + (Gymnuridae + Myliobatidae)) could also have inhabited the proto-Caribbean region (or the eastern Pacific realm, which was continuous with the proto-Caribbean until the completion of the Panamanian Isthmus in the Pliocene). Therefore, the ancestor of the Potamotrygonidae could still have invaded South America through a marine introgression from that area. This invasion is possible because, according to our phylogeny, there was already a large stingray lineage present in the Americas which eventually gave rise to the Urotrygonidae, †Heliobatis, and to the common ancestor of the large clade above (this component originally had a strong historical association with the Indo-Pacific region; see discussion above concerning the origin of the Green River stingrays).

We disagree, however, with Lovejoy et al.'s (1998) hypothesis in relation to the timing of the vicariant event that allowed for the introduction of the potamotrygonid ancestor into South America. A Miocene epicontinental seaway (such as the Pebasian Seaway of Lundberg et al., 1998) occurred far too recently according to our phylogeny, but earlier marine incursions are also known from the proto-Caribbean region (i.e., from the north; Hoorn, 1993; Hoorn et al., 1995; Smith et al., 1994; Lundberg et al., 1998). According to Smith et al. (1994) and Lundberg et al. (1998), extensive marine transgressions over northern South America occurred frequently (at least four times) from the Late Cretaceous (Maastrichtian) to the late Miocene, and the duration of each incursion and its resulting seaway lasted for millions of years. Two epicontinental incursion episodes in particular are congruent with the minimum age of the potamotrygonid lineage (50 mya) derived from our phylogeny: a Late Cretaceous seaway, and another one that lasted from the late Paleocene to the very early Eocene (fig. 52). Either incursion may have trapped the potamotrygonid ancestor in the neotropics, where it slowly adapted to a progressively more freshwater environment.

The biogeographic hypothesis presented here for the origin of the Potamotrygonidae (a pre-early Eocene marine incursion from the north) does not refute the area relationships of other taxa mentioned by Lovejoy (1996) in support of his theory (the “proto-Caribbean/Himantura” hypothesis). Our hypothesis (cf. fig. 48) is congruent with the area cladogram presented by Lovejoy (1997: fig. 4a): Indo-west Pacific + (South American freshwater + (eastern Pacific + Caribbean/northern West Atlantic)). We simply think that a further area relationship (the American component) should be added after the Indo-west Pacific region (the basalmost area). These area relationships largely apply to other groups in which there is a basal Neotropical freshwater species that is the sister-group to an eastern Pacific/Caribbean species pair (such as anchovies of the genera Cetengraulis and Anchovia; Nelson, 1984) or other taxa that may have undergone vicariance and subsequent diversification due to the formation of the Panamanian Isthmus (such as amphi-American Himantura). These historical associations are still possible under our hypothesis; the only adjustment is that their ancestors may have invaded the Neotropical freshwater system much earlier than the Miocene (but not necessarily, at least regarding the engraulids). Without robust dating (i.e., strong fossil evidence), it is not possible to precisely determine which marine transgression (or other mechanism) was responsible for the invasion of the ancestral taxon; it could have been the same epicontinental seaway that carried the potamotrygonid ancestor (or other primitively marine groups), but this event could also have occurred much later. Although amphi-American Himantura is placed in a polytomy with other “dasyatid” genera and Gymnuridae + Myliobatidae, it may eventually prove to be basal to these taxa (as indicated when analyzing our matrix with the weighting schemes of Farris, 1969 and Goloboff, 1993; see above and fig. 47). In addition, our biogeographic theory does not contradict the northwest–southeast derivation sequence or vicariant pattern that has been reported for some Neotropical fishes distributed in northern South America (e.g., the curimatid Potamorhina, Vari, 1984; to a lesser extent Steindachnerina and Creagutus, Vari, 1991; Vari and Harold, 2001; for summaries, see Vari, 1988; Vari and Weitzman, 1990; Lundberg and Chernoff, 1992). In these cases, according to the above authors, basal species are distributed in northern South America and, through vicariance, more derived species have been pushed progressively farther south or southeast from the ancestral area.