A NEW GENUS OF RHYNCHOSAUR FROM THE MIDDLE TRIASSIC OF SOUTH-WEST ENGLAND

Authors


Abstract

Abstract:  We present a description of new cranial and postcranial material representing a new genus of rhynchosaur (Diapsida, Archosauromorpha) from the Otter Sandstone Formation (Mid Triassic) of Devon, south-west England. The taxon had been named Rhynchosaurus spenceriBenton, 1990, but cladistic analysis of the clade, and one autapomorphy, show that it does not belong to Rhynchosaurus, and a new generic name is required. We propose the name Fodonyx for this genus. A cladistic analysis of the Rhynchosauria confirms the main discoveries of previous analyses, and that Fodonyx is sister group to the Hyperodapedontinae, the clade of Late Triassic rhynchosaurs. The new cladistic analysis, for which many more characters were coded for Fodonyx than before (a rise from 39 to 75 per cent), counter-intuitively produced less well-resolved results: the new codings of previously uncoded characters introduced conflict so that Fodonyx turns out to be less like the Late Triassic rhynchosaur clade than had been assumed before.

Rhynchosaurs were terrestrial, herbivorous basal archosauromorphs of the Early–Late Triassic. Large claws on the hind feet suggest the ability to dig well and the heavy jaws and interlocking tooth-plates imply a diet of tough vegetation and tubers (Benton 1983). Rhynchosaurs enjoyed a near-global distribution, having been found in Great Britain (Benton 1983, 1990), South Africa (Dilkes 1998), Zimbabwe (Raath et al. 1992), Tanzania, Madagascar (Buffetaut 1983), India (Chatterjee 1974), Brazil (Langer and Schultz 2000), Argentina, Canada and the United States (Lucas et al., 2002; Nesbitt and Whatley 2004). There are currently some 15 valid species of rhynchosaur, recognised after recent revision and recognition of synonymies, as well as a further five unnamed, or tentatively named, taxa based on incomplete remains from other locations (Benton 1983, 1990; Hunt and Lucas 1991; Langer and Schultz 2000).

Cladistic analyses of the phylogeny of rhynchosaurs (Benton 1983, 1985, 1990; Evans 1988; Dilkes 1995, 1998; Wilkinson and Benton 1995; Langer et al. 2000a; Langer and Schultz 2000) have revealed a common pattern: two basal forms, Mesosuchus and Howesia from the Lower Triassic of South Africa, then a number of Mid Triassic forms, principally Rhynchosaurus and Stenaulorhynchus, and a crown clade of Late Triassic forms, species of Hyperodapedon and Scaphonyx.

The Devon rhynchosaur was named Rhynchosaurus spenceri by Benton (1990), a new species assigned to the existing genus RhynchosaurusOwen, 1842. The new specific name was established since R. spenceri differed from all other then-known rhynchosaurs, but the remains were too incomplete to determine whether it belonged to a distinct genus or not. Indeed, R. spenceri was excluded from the cladistic analysis by Benton (1990) because so many parts of the skeleton were unknown, and hence uncoded in the character matrix. This approach was criticized by Wilkinson and Benton (1995): excluding a taxon from cladistic analysis simply because it is incompletely coded is inappropriate. On including R. spenceri in a repeat of Benton’s (1990) cladistic analysis, they found that it fell between the other two species of Rhynchosaurus, R. articeps and R. brodiei, on the one hand, and the Late Triassic clade on the other. Langer and Schultz (2000) found the same thing, and indicated that R. spenceri was definitively not a member of the genus Rhynchosaurus.

The aims of this paper are to present new material of the Devon rhynchosaur, concentrating in particular on a new partial skeleton and a new skull, and to place this in the context of a thorough phylogenetic analysis of rhynchosaur relationships. Our conclusion is that the Devon rhynchosaur cannot remain a species of Rhynchosaurus and has to be assigned to a new genus.

Institutional abbreviations.  BRSUG, Bristol University, Department of Geology; EXEMS, Royal Albert Museum, Exeter.

Systematic Palaeontology

Subclass DIAPSIDA Osborn, 1903
Infraclass ARCHOSAUROMORPHA Huene, 1946
Order RHYNCHOSAURIA Osborn, 1903
Family RHYNCHOSAURIDAE Huxley, 1859

Genus FODONYX gen. nov.

Fodonyx spenceri (Benton, 1990)
Text-figures 1–7
Figure TEXT‐FIG. 1..

Fodonyx spenceri (Benton, 1990) from the Otter Sandstone Formation (Middle Triassic, Anisian) of Devon, south-west England. Skull (BRSUG 27200) in A, dorsal, B, ventral (excludes separate right squamosal), C, right lateral, D, posterior, and E, anterior views.

Figure TEXT‐FIG. 2..

 Reconstruction of Fodonyx spenceri (Benton, 1990) from the Otter Sandstone Formation (Middle Triassic, Anisian) of Devon, south-west England, based on skull BRSUG 27200, and with additional information from the holotype, EXEMS 60/1985.282, in A, dorsal, B, ventral, C, right lateral, D, posterior, and E, anterior views. For explanation of abbreviations, see Appendix. Some sutures cannot be determined from the new skull and are highly variable in other rhynchosaurs, and thus cannot be estimated with confidence and are not included. The posterior part of the lower jaw and parts of the squamosal and quadratojugal are restored based on existing material of F. spenceri (compare with Text-fig. 1).

Figure TEXT‐FIG. 3..

Fodonyx spenceri (Benton, 1990) from the Otter Sandstone Formation (Middle Triassic, Anisian) of Devon, south-west England. Partial postcranial skeleton (EXEMS 79/1992). Scale bar represents 200 mm split into divisions of 10 mm. For explanation of abbreviations, see Appendix.

Figure TEXT‐FIG. 4..

 Reconstruction of axial material of Fodonyx spenceri based on EXEMS 79/1992. A, dorsal view of a single vertebra. B, left lateral view of a single vertebra. C, anterior view of a partial vertebra (centrum not shown). D, chevron. E, dorsal rib.

Figure TEXT‐FIG. 5..

 Detail of lower left hindlimb, in ventral view, of Fodonyx spenceri from EXEMS 79/1992. For explanation of abbreviations, see Appendix.

Figure TEXT‐FIG. 6..

 Reconstruction of lower left hindlimb, in dorsal view, of Fodonyx spenceri based on EXEMS 79/1992. For explanation of abbreviations, see Appendix.

Figure TEXT‐FIG. 7..

 Reconstruction of left tibia of Fodonyx spenceri based on EXEMS 79/1992, in A, proximal, B, anterior, C, left lateral, D, posterior, and E, right lateral views.

[Full synonymy list to 1990 in Benton 1990]

1993 Rhynchosaurus spenceriBenton, 1990; Benton et al., pp. 167–170.

1994 Rhynchosaurus spenceriBenton, 1990; Benton et al., pp. 145–146.

1995 Rhynchosaurus spenceriBenton, 1990; Wilkinson and Benton, pp. 141–147.

1997 Rhynchosaurus spenceriBenton, 1990; Benton, pp. 145–147.

2000 ‘RhynchosaurusspenceriBenton, 1990; Langer and Schultz, pp. 647–648.

2000 ‘RhynchosaurusspenceriBenton, 1990; Langer et al., pp. 120, 123–127.

Derivation of name.  The new generic name Fodonyx is modelled on the name Scaphonyx, given to the Brazilian rhynchosaur by Woodward (1907), and meaning ‘spade claw’: Latin, fodere, to dig, and Greek, onyx, claw, hence meaning ‘digging claw’, in reference to the large unguals present on the foot.

Holotype.  EXEMS 60/1985.292, a partial skull and mandible, including the floor of the orbit and the palate of the right side, a partial palate of the left side, the posterior right-hand angle of the skull, and both mandibles (Benton 1990, figs 28–29, 31, 36i–j, 37f–g).

Referred material.  Material listed by Benton (1990, pp. 221–222) as well as EXEMS 79/1992 (a partial postcranial skeleton; Benton et al. 1993), BRSUG 27200 (a nearly complete skull), EXEMS 60/1985.20 (isolated right jugal); EXEMS 60/1985.48 and BRSUG 26874 (two isolated parietals); BRSUG 26875 (partial dorsal vertebra); EXEMS 60/1985.324 (partial toe with ungual).

Locality and horizon.  Various sites along the South Devon coast from Ladram Bay to Port Royal, Sidmouth (maps and map references in Benton 1990, 1997; Benton et al. 1993, 1994), all in the upper half of the Otter Sandstone Formation (Sherwood Sandstone Group). The unit is dated as Anisian on the basis of faunal comparisons and matching with palynologically dated units in the English Midlands (Benton et al. 1994). This has since been confirmed by magnetostratigraphy (Hounslow and McIntosh 2003), who found that the lower parts of the Otter Sandstone Formation correspond to the early and mid Anisian, and the upper parts, which contain the majority of the macrofossils, correlate with late Anisian and latest Anisian magnetozones on the marine standard.

Revised diagnosis. Fodonyx has a single apomorphy that distinguishes it from other rhynchosaurs: the paraoccipital processes, when seen in posterior view, angle ventrally. In all other rhynchosaurs, these processes are either horizontal or are raised dorsally. The cladistic analysis also confirms that Fodonyx is distinct from any other known rhynchosaur. Fodonyx differs from Rhynchosaurus (as represented by R. articeps and R. brodiei) in having the orbits orientated more dorsally than laterally, a major diagonal crest on the jugal that reaches the anterior portion of the orbit, and ornamentation on the external surface of the jugal with crests and bosses dorsal to this diagonal crest, in having the frontal longitudinal groove almost the same depth throughout, a basipterygoid process that is broader than long, and chevron bones that taper distally. Fodonyx lacks the characters of the Late Triassic Hyperodapedontinae (see below).

Morphology of the Skull

Introduction

A new rhynchosaur skull was found in spring 1999 by Mark Hounslow at Pennington Point, about 20 m east of the River Sid outfall (National Grid reference SY 130873). It was in situ (many fossils have come previously from fallen blocks) at the base of the cliff, and found at a time when much of the beach gravel had been swept away. Stratigraphically, the skull was found about 3 m below the top of Andy Newell’s Unit C, the Pennington Point Member of Gallois, and layer 21 of Hounslow and McIntosh (2003). According to the magnetostratigraphy of these authors, the age of this horizon is within the late Illyrian (latest Anisian), equivalent to the lower part of the Tethyian P. trammeri (conodont) Zone.

This skull (BRSUG 27200) was prepared by Remmert Schouten in the Bristol Palaeontology Laboratories in 2004 and 2005. The sandstone was removed from the dorsal and right-hand sides and the whole of the palate was exposed. The orbital and temporal regions on the right-hand side were prepared deep into the specimen. Sandstone was left in parts of the left-hand side in order to keep the specimen stable.

The skull is typically rhynchosaurian, triangular in shape in dorsal view, with large subcircular temporal fenestrae and orbits, a large midline naris, hooked ‘tusk’-like premaxillae, and broad maxillary tooth plates. The skull was probably rather lower than assumed by Benton (1990), whose reconstruction was based on materials that lacked the skull roof. It is nearly complete and has the lower jaws in place (Text-figs 1–2). The posterior ventral corners of the skull, including both quadrates, the quadratojugals and parts of the jugals are missing. The skull roof has been depressed slightly, causing movement between skull roof bones at suture lines. The braincase is intact but has been disarticulated and moved slightly forward within the cranium. The base of the lower jaw has been eroded, probably before preservation, and is also missing parts of the angular and surangular.

Skull roof (Text-figs 1A, C, E, 2A, C, E)

The premaxilla is long and runs up over the snout to contact the prefrontal, close to the orbit. At its tip, the premaxilla is small and triangular in cross-section. The two premaxillae do not quite contact at their anterior tips, probably as a result of the post-mortem crushing. As a pair they form the classic rhynchosaurian ‘beak’. The premaxillae separate from one another rapidly backwards to provide space for the nasal bones and the boundaries of the heart-shaped single naris. Towards the distal end, there is a deep facet posterolaterally worn out by the anterior tip of the dentary when the jaws are closed. Just above this is a change in angle and texture in the premaxilla, with more marked longitudinal striations in the dorsal half of the element. This presumably marks where skin covered the dorsal half, and perhaps a keratinous sheath covered the distal, smoother half of the premaxilla.

The maxilla is broad and plate-like, forming much of the anterior side of the snout, as well as carrying the massive tooth plate. It contacts the premaxilla for three-quarters of the length of the latter. Posteriorly, the maxilla contacts the prefrontal, lacrimal and jugal, and it runs back to terminate below the posterior portion of the orbit. A number of foramina are visible, both for nerves and capillaries, in a row about 10 mm above the curved ventral tooth-bearing margin, and possibly indicating the limit of gum tissue. On the ventral edge, a number of unworn posterior teeth are visible in lateral view. The ventral view is obstructed by the mandibles, but the maxillary tooth plate was clearly an elongate triangle, divided by a midline groove, as seen in numerous isolated maxillae (Benton 1990, pp. 274–276). The occlusal surface of the maxillary tooth plate shows an anterior part, where bone and teeth are worn smooth by use, and a posterior part where up to five lateral teeth remain unworn and bear their enamel caps (stained black in preservation). Hidden from view are the medial tooth rows, and the subsidiary medial groove, seen in isolated maxillary tooth plates.

The nasal bones are large and paired, with a zigzag suture line down the midline of the skull. As with most rhynchosaurs, the nasals are shorter than the frontals, but they are similar in width. The nasals form the curved posterior border of the naris, and they extend back with the lateral margin in contact fleetingly with the premaxilla, and mainly with the prefrontal. The posterior margin is pointed, forming a firm zigzag suture with the narrow anterior margin of the frontal.

The lacrimal is small and placed on the anterior edge of the orbit. The lacrimal ducts are clearly visible along the inner edge of the orbital opening.

The prefrontal is large and borders the anterodorsal margin of the orbit, forming a thick, angular ‘eyebrow ridge’. Medially, the prefrontal is bounded by the nasals and frontals, but it does not meet the postfrontal.

The jugal is a complex four-branched element forming the ventral margin of the orbit and the anterior and ventral margins of the lower temporal fenestra. The anterior branch is bounded ventrally by the maxilla, and contacts the lacrimal anterodorsally. There is a deep, smooth depression in this branch beneath the orbit, and the jugal expands laterally to swing into the posterior process beneath the lower temporal fenestra. This portion is missing on both sides, something seen in many incomplete rhynchosaur skulls. Based on the distribution of this character in all other derived rhynchosaurs, the posterior branch of the jugal met the quadratojugal at the back of the lower temporal opening. The dorsal branch of the jugal forms the thin anterior margin of the lower temporal fenestra, and underlies the ventral branch of the postorbital, forming a strong pillar behind the orbit. The crest along the anterior margin of this pillar is more pronounced than seen in other rhynchosaurs. The medial branch of the jugal, which lies above and behind the maxillary tooth plate in the palate, is incompletely preserved. The jugal carries some small vessel openings and rugosities along the sloping ventral margin of the anterior process where it lies above the maxilla, but this rugose ridge is much less pronounced than in other large rhynchosaurs, such as Hyperodapedon (Benton 1983).

The frontals are long and together form an extended diamond-shaped portion of the skull roof. They are joined to each other firmly down the midline along a suture line that is straight posteriorly, and has small zigzags anteriorly. The frontals contact the nasals in front, the prefrontal and postfrontal laterally, with a short exposure in the dorsal orbital margin, and it tapers back to connect with the parietals nearly at a point. The anterior portion of the frontals is roughly horizontal, but they are dished posteriorly, and rise to a low, smooth ridge along the line of the frontal/postfrontal suture.

The postfrontal is roughly triangular, forming the posterodorsal margin of the orbit. There is a long, straight, posteromedial contact with the frontal, and posteriorly with the parietal. This posterior parietal contact dips ventrally, and is overlain by the pointed medial branch of the postorbital, which sits in a deep ‘socket’ of the postfrontal.

The parietals are fused, and together form a T-shaped element running from the frontals and postfrontals back to the occiput, and laterally to the squamosals. They form the medial borders of the upper temporal fenestrae. The midline portion of the parietals is triangular in cross-section, with a high, narrow ridge dorsally. The posterior margin of the parietals is slightly concave in dorsal view, and quite deep and flat in occipital view, where the lateral wings extend for virtually the width of the upper temporal fenestrae, lying beneath the medial branches of the supratemporals and squamosals.

The postorbitals are roughly T-shaped, with medial, posterior and ventral branches. The medial branch runs to a point that sits in a deep ‘socket’ formed by the underlying postfrontal. The medial and posterior branches form the anterior and lateral margins of the upper temporal fenestra and the dorsal margin of the lower temporal fenestra. This branch sends a long, broad, but thin, tongue back over the main body of the squamosal, and terminates in line with the posterior margins of the temporal fenestrae. The ventral branch is triangular in cross-section, forms a deep posterior margin of the orbit, and sits on a broad attachment with the jugal.

The presence or absence of the supratemporal has long been debated in Rhynchosaurus and other rhynchosaurs (Benton 1990, p. 228). Here it is clearly present and well defined, a slender element lying in a gulley along the posterior margin of the squamosal and contacting the parietal medially. The supratemporal is present in other Early and Mid Triassic rhynchosaurs, but is clearly absent in Late Triassic forms such as Hyperodapedon and Scaphonyx.

Only the right squamosal is preserved, and that only in part. It is a three-branched element that forms much of the posterior margin of the skull, as well as the posterior margins of both temporal fenestrae. The medial and ventral branches outline the bulk of the dorsolateral and lateral margins of the occiput. The medial branch lies over the lateral branch of the parietal, and forms a concavity for the supratemporal. The ventral branch descends along the posterior margin of the lower temporal fenestra, and its posterior margin curves forwards around the conch-like depression over the quadrate and quadratojugal, which it overlies. This detail is not seen in the present specimen, but was described in EXEMS 60-1/985.292 (Benton 1990, fig. 28c–d). In BRSUG 27200 the medial branch of the squamosal extends forwards at least half-way between the temporal fenestrae, beneath the postorbital.

The quadratojugal and quadrate are largely missing on both sides, with only a small part of the dorsal portion of the quadrate adhering beneath the squamosal. An isolated right quadrate/quadratojugal unit is present in the holotype skull of F. spenceri (EXEMS 60/1985.292; Benton 1990, fig. 28). The separation might be because, although the quadratojugal and quadrate form a massive column in the posterior corner of the skull, their connection to the rest of the skull, through the posterior branch of the jugal, the ventral branch of the squamosal, and the quadrate wing of the pterygoid, is weak and is likely to break in a defleshed specimen.

Palate and hyoids (Text-figs 1B, 2B)

Much of the palate is intact, though parts have fragmented, probably before preservation, and some is preserved with calcitic matrix that could not be prepared away without risk to the underlying material. The palate is sharply triangular, bounded by the massive maxillary tooth plates on either side in the anterior half, and with the pterygoids swinging posterolaterally from the midline to the quadrates in the posterior half. The anterior lateral parts of the palate are partly obscured by the mandibles.

The paired vomers are relatively long with a simple suture between them. They are composed of very thin bone that has degraded somewhat. Anteriorly, they meet at a point just behind the premaxillae, and laterally fuse with the maxillae. Posteriorly, they narrow to enclose the anterior part of the choana.

The paired palatines meet the vomers anteriorly in the midline, and they contact each other along a straight midline suture. The palatine forms most of the medial, posterior and lateral borders of the choana and part of the infraorbital foramen. Posterolaterally the palatine sits below the lateral wing of the pterygoid, and meets the ectopterygoid.

The pterygoids are the largest elements in the palate, and they have three main processes. The anterior processes meet each other with a short midline suture in a V-shape just behind the palatines, but the pterygoids then open up for most of the midline, forming a clear interpterygoid space. The lateral process of the pterygoid is broad and plate-like, meeting the jugal above and each palatine anterolaterally, and lying above the ectopterygoid. The lateral process sweeps in a broad, partial spiral into the deep, plate-like posterolateral, or pterygoid, process that traverses the posterior portion of the palatal view of the skull to meet the quadrate in the posterolateral corner of the skull. This contact is partially preserved here, but the rest is seen in EXEMS 60/1985.282 (Benton 1990, fig. 28d). Medially, the pterygoids form shallow pockets for the basipterygoid processes of the braincase.

The ectopterygoid is small and spans behind the maxillary tooth plate and sits behind the lateral wing of the pterygoid, meeting the jugal laterally. The ectopterygoids cannot be seen in palatal view (Text-fig. 1B) as they are covered by matrix and hyoid elements.

Hyoids have been reported before in Hyperodapedon and Scaphonyx (Benton 1983, pp. 637–638), but understandably they are rarely seen. In the current specimen, anterior portions of a right and left hyoid element (hy, Text-fig. 2B) are preserved below the ectopterygoids, in contact with the splenial on either side, and running posteromedially for 15 mm or so. Each element is circular in cross-section, about 4 mm across, and longitudinally striated.

Braincase (Text-figs 1B, D, 2B, D)

The endocranium is largely intact and in situ. It has slightly dissociated from the skull and moved down and to the right, but maintains its orientation. The left-hand side is eroded, but the centre and right are complete. Calcitic matrix has made some areas impossible to prepare fully. Overall, it is typical of rhynchosaurs in size and shape (cf. Benton 1983).

The basioccipital is short, carrying the hemispherical occipital condyle. It meets the basisphenoid on a straight suture across the tubera spheno-occipitales. The basisphenoid narrows dramatically in front of the tubera, with a deep depression in the ventral middle portion, before expanding anterolaterally into the basipterygoid processes. These extend ventrally as somewhat trumpet-shaped processes, and have been rammed deep into the receiving pits in the pterygoids, presumably by taphonomic dislocation. Although the end has slightly eroded, the rather more complete right basipterygoid process is 62 mm in length, and therefore c. 50 per cent longer relative to skull length than in Rhynchosaurus and 100 per cent larger in absolute size.

The paroccipital process on the right side shows the typical exoccipital and opisthotic. The cranial nerve foramina, metotic foramen, fenestra ovale and exoccipital-opisthotic suture cannot be distinguished because of the difficulties of preparation. The paroccipital process runs slightly ventrally, and its distal end expands to form a deep, dorsoventrally orientated distal portion where it contacts the squamosal and quadrate (Text-fig. 1D).

The prootic is hard to make out because of adhering calcitic matrix. The supraoccipital can be determined in rough outline: it is a roof-like element, forming the dorsal margin of the foramen magnum, and extending up to meet the underside of the parietal in a point in the midline.

Lower jaw (Text-figs 1B–C, E, 2B–C, E)

Only the anterior portions of both lower jaws are preserved, the posterior elements having been largely removed; the ventral margins of both mandibles have also been eroded, since they stuck out of the block in which the skull was found. Enough of the lower jaw is preserved to show its typical rhynchosaur shape, and the complete mandibles in the holotype (EXEMS 60/1985.282; Benton 1990, figs 28, 31) allow a full reconstruction (Text-fig. 2C).

The dentary forms more than the anterior half of the mandible. It carries teeth on its expanded dorsal face, but these are largely obscured, except for four or five unworn posterior jaw-crest teeth on the left mandible, because the jaws are tight shut. The dorsal margin of the dentary curves up to an anterodorsally extended, rounded anterior tip. This anterior tip fits neatly in a facet of the premaxilla when the jaws are closed. The dentary extends from a thick tooth-bearing dorsal portion downwards as a thin plate around the cavity occupied in life by Meckel’s cartilage. The lateral plate meets the splenial ventrally, and the angular and surangular behind. The dentary bears a 4-mm-deep ‘smooth’ zone below the tooth row, and below that are major vessel openings, especially towards the front of the jaw.

The splenial is complete on both sides (except for the skimmed-off ventral portion of the mandibles). It forms the medial wall of the channel for Meckel’s cartilage, and the ventral part of the anterior mandible. It is seen in lateral view only towards the front of the jaw, where it expands to support the strong symphyseal plates.

A portion of the right surangular is preserved, attached to the posterior dorsal margin of the dentary. Only the anterior part of the angular may be seen on the eroded ventral face of the mandible where it extends between dentary and splenial in the floor of the cavity for Meckel’s cartilage. The prearticular and articular are absent.

Morphology of the Postcranial Skeleton

The description of the postcranial skeleton that follows is based on the partial skeleton EXEMS 79/1992, found in Ladram Bay (National Grid reference SY 098852) in 1992, and described briefly by Benton et al. (1993). Further preparation since then has revealed more detail, and the following elements may now be identified: portion of the mandible (ventral portion of angular?), 17 dorsal and anterior caudal vertebrae, nearly all separated into centra and neural arches, three chevrons, numerous ribs, including a nearly complete anterior dorsal rib, much of the anterior gastral basket, left scapula, left humerus, left ilium, ?both ischia, and partial left hindlimb (tibia, fibula, ankle and foot elements). The specimen is seen in ventral view (Text-fig. 3), as shown by the gastralia and ventral views of the dorsal centra. This is the original orientation of the specimen in the rock (Benton et al. 1993), so the carcass must have flipped over, belly uppermost (perhaps floating in the river that deposited it, the belly inflated with digestive gases). Most of the elements to the right of the main blocks with the backbone and ribs are then from the left side of the animal. The hindlimb is bent back, and seen in ventral view, with the foot trailing behind the carcass.

Most of the remains (vertebrae, ribs, gastralia, chevrons, left ilium) sit in three sandstone blocks that overlap and fit together firmly. The left hindlimb and foot is in three blocks that may be linked to the main blocks via some rib ends. The small blocks with the left scapula, left humerus and mandible probably lie in front. A further eight small blocks with isolated vertebrae, ribs, the ?ischia, and other remains have lost their association, but are part of the specimen.

Vertebrae and chevrons

In total 17 vertebrae are present and appear to be all dorsal and caudal, with no cervical or sacral vertebrae present, even in the pelvic area. They are all disarticulated and lie scattered between the shoulder and hip regions, surrounded by ribs; in most cases, the centra and neural arches lie separated. For this reason, the centra and neural arches are described separately.

The preserved dorsal centra are uniform in size and shape and are roughly equal in height and diameter (Text-fig. 3). They range in length from 16 or 17 mm for anterior dorsal centra to 14 or 15 mm for posterior dorsals. They are deeply amphicoelous and spindle-shaped with a pronounced narrowing towards the centre. The centra are deep and lack a ventral keel. The articular ends are roughly circular in anterior and posterior views, and each articular end is surrounded by a pronounced lip that curves over and forms a major projection from the lateral and ventral sides of the centrum. The dorsal surface of the centrum is exposed in many cases (Text-fig. 4A–C) and shows a deep midline excavation in the floor of the neural canal, sometimes with narrow, sharp-crested, longitudinal ridges in the base. The neural arch facets are broad and flat, and run down each side of the dorsal surface of the centrum. Generally, these longitudinal facets are traversed by shallow transverse ridges and grooves that match grooves and ridges on the base of the neural arches.

Several disarticulated neural arches provide information from dorsal, ventral and lateral views. The neural arch of dorsals (Text-figs 3, 4A–B) is broad, as in most larger rhynchosaurs (Benton 1983, 1990), supporting the zyapophyses, transverse processes and the dorsal spine. The neural canal is roughly circular in cross-section. In anterior view, the neural arch sits with broadly expanded facets on top of the centrum, and these facets nearly meet in the midline above the centrum. They extend upwards into narrow pillars on either side of the neural canal that support the broad prezygapophyses. The postzygapophyses extend ventrolaterally from the posterior margin of the neural spine, and support the articular faces. The articular faces of pre- and postzygapophyses are broad and nearly circular. In many cases, they appear to be orientated sub-horizontally, tilted up by only 10–20 degrees laterally: this would allow considerable lateral bending of the vertebral column, but very little dorsoventral flexion. The transverse processes are anteroposteriorly broad, and located anteriorly, just behind, and partly below, the prezygapophyses. They project at right angles to the midline of the vertebra, and are broad and oval, with smooth edges. The base of the neural spine is broad and occupies the posterior half of the neural arch. The spine itself is not preserved in many vertebrae. In an anterior dorsal, near the humerus, the spine is 17 mm high, measured from the postzygapophyseal facet, and it narrows from 12 mm in anteroposterior length at the base, to 10 mm distally. There is no spine table.

A series of three chevrons is preserved, only the anterior of which may be seen in anterior and lateral views (Text-fig. 4D), the other two being partially concealed behind it and in the matrix. The anterior chevron is 24 mm long, but its distal end is missing. The chevron is the usual Y shape, but the dorsal facets are fused into a single element, 9 mm across, bridging over a triangular-shaped opening. The ventral projections of the chevron outline the triangular opening, and come together as a keel that expands distally to be 5 mm deep anteroposteriorly, but the distal termination is unknown.

Ribs and gastralia

Numerous ribs are present, though most are fragments of the main shaft, and these show little more than that the ribs were robust and deep, and bore a longitudinal groove on the broader lateral face. One or two posterior dorsal ribs appear to show their proximal end, a single capitulum 5 or 6 mm wide.

One moderately complete anterior dorsal rib, lacking only the distal tip, is preserved (Text-fig. 3 (r), 4E), probably from one of the first three or four vertebrae of the trunk. It is some 98 mm long, and 22 mm wide anteriorly, narrowing to 8 mm in the middle, and 10 mm distally. The broad capitulum has an oval articular facet c. 12 mm wide that attached to the transverse process. A narrower tuberculum, c. 7 mm wide, is found on a ventral process attached to the side of the centrum. This rib is remarkably straight, and lacks the normal curvature one might expect to encompass the rather curved flanks of a bulky herbivore; in life, it probably angled backwards from an anterior dorsal vertebra, rather as seen in the better articulated specimens of Hyperodapedon gordoni figured by Benton (1983, fig. 27).

An isolated rib head close to the mandibular fragment (r*, Text-fig. 3) is probably also an anterior rib, but the shaft is much more curved, and this might come from behind the anterior ‘straight’ ribs. The shaft, some 20 mm from the head, is 5 mm wide, and it narrows distally to 4 mm, but the whole specimen is no more than 30 mm long. The dichocephalous head is beautifully shown, spanning 19 mm. The articular face of the capitulum is oval in outline, with a posterior twist, terminating the oval with a tail shaped like the number 6. The face measures 4 × 10 mm in maximum dimensions. The tuberculum articular face is more circular in outline, with a diameter of 4 mm. The capitulum and tuberculum heads are separated by a deep, V-shaped lamina of bone, and the tuberculum extends 5 mm, and the capitulum 3 mm, from the base of the V.

Some 15–20 anterior gastralia (g, Text-fig. 3) are seen in a bunch below the anteriormost dorsal vertebrae and the anterior ‘straight’ dorsal rib. The gastralia appear to be in three segments, the long, narrow lateral elements overlapping a middle segment, some 25 mm long. Lateral gastralia are 60–70-mm-long, thin, strap-like elements, some 2 mm wide and 0.1 mm deep. The complete gastral basket is rather disarticulated, but the orientation of the elements is roughly straight from side to side, with little evidence of a forward or backward V-shape.

Shoulder girdle and forelimb

Both scapulae are present. The left scapula (L.sc., Text-fig. 3) is incomplete around all but the anterior margin. The blade broadens dorsally from a narrow point of 20 mm at the base of the blade to at least 40 mm wide measured anteroposteriorly at the distal margin. The anterior margin has a prominent boss (partially missing), against which the clavicle may have attached (cf. Benton 1990, fig. 17c). The ventral part of the scapula, with the facet for attachment to the coracoid, and the glenoid facet, is incomplete. The right scapula is beneath the main blocks, behind the mid-dorsal ribs, and shows a medial view of the blade.

The left humerus (L.h, Text-fig. 3) is incomplete at both proximal and distal ends, but it is similar in size and shape to the left humerus described by Benton (1990, fig. 33). The bone is some 65 mm long, as preserved, 35 mm across the proximal end, and 33 mm across the distal end, although these measurements are probably rather less than for the complete bone. The posterior wing of the proximal end is broad and thin, while the deltopectoral crest, largely missing, has a broad base. The shaft of the humerus is oval in cross-section, and 7 mm across at its narrowest point. The distal end, set at some 90 degrees to the proximal end, shows the ventral depression seen by Benton (1990), but the supinator crest and articular margins are incomplete.

Pelvis and right hindlimb

The pelvis is represented by incomplete remains of the ilium and ischium, which are identified by their overall shape, in comparison with these elements in other rhynchosaurs, and their location relative to the posterior dorsal vertebrae, and the partial right hindlimb. Two parts of the ilium are present, both showing the medial surfaces. Deep growth lines in a radiating pattern from the centre are clearly visible in both. The left ilium (il, Text-fig. 3) lies behind some posterior dorsal vertebrae and ribs, and shows a medial view of the dorsal blade, but it is difficult to orientate.

Putative remains of both ischia are in a small block. The left ischium (L.is, Text-fig. 3) is most complete, showing its dorsal thickened and rounded margin, the buttress behind the major facet for attachment to the ilium, and most of the curved, rounded blade. The right ischium is more embedded in the matrix and is harder to describe.

The preserved left hindlimb lacks only a femur (Text-figs 3, 5–6). It consists of almost all the bones of the leg and foot except the femur. The foot is split across two overlapping blocks and the break-line runs across the tarsals and metatarsals. All five digits are present. The tibia is broken but largely intact, though the dorsal surface is eroded. The fibula is missing its distal end and is mostly hidden behind the tibia. Almost all of the ankle bones and distal tarsals appear to be present but are partially concealed by other bones and the matrix; they are also somewhat worn and so difficult to identify. At least four metatarsals are present, but they are broken, and parts of the foot are lost across the break between the blocks. The tarsals and unguals are almost all present, though some are reduced to fragments or impressions in the matrix.

The tibia (Text-figs 5–7) is largely complete; it is 84 mm long, and anteroposteriorly quite thin, with a broad-faced head and shaft but with a relatively narrow base. The cross-section is a compressed oval measuring 13 × 6 mm at the narrowest point. The shaft shows a distinct twist, highlighted by a diagonal ridge that probably lay behind the fibula in life. The proximal end is 34 mm across; the articular surface shows heavy wear, and no fine detail is preserved. The distal end is 21 mm across.

The fibula (Text-figs 5–7) is obscured by both the tibia and matrix and the distal end is missing, though an impression of the end is present in the matrix (with some bone fragments). It appears to be similar in length to that of the tibia and relatively robust, although more slender than the tibia, with an oval cross-section measuring 8 × 5 mm at the distal end. Both the tibia and fibula have been rotated prior to preservation so that they lie in a position that is the reverse of what would normally be expected.

The proximal tarsals (Text-figs 5–7) are difficult to identify, partly because of the rotation of the limb, but mainly because they are of similar size and shape. All three are present, but are heavily worn, rounded and distorted. One is located beneath the fibula and is probably the centrale. Another lies immediately below, and a third is present behind the tibia on the underside of the matrix, presumably the astragalus and calcaneum respectively. Three distal tarsals are identified, DI, DII and DIII (Text-fig. 6). DIII is only visible on the underside of the matrix. Again, these are worn and/or broken.

The metatarsals have all been broken across their width, and the proximal ends of metatarsals (mt) III–V are partly hidden beneath the tibia. Mt I is short and broad, 13 mm wide at the proximal end and 15 mm long. Mt II, III and IV are long and flat; the breakages mean that their true lengths are hard to determine, but they measure between 32 and c. 50 mm long. Mt V is badly broken, but the 18-mm-wide proximal end is seen well under the slab. The broad proximal end of mt III heavily overlaps II and IV.

The phalanges are largely intact and show the expected number of elements per digit: 2, 3, 4, 5, 4. Most phalanges retain their articulation, although some have rotated within their digits. All are broader at the proximal end than the distal; they become narrower and taller distally, partly so that the penultimate phalanx in each digit supports the high ungual. Four unguals are present (that from digit IV is missing, but an impression is left in the matrix). Ungal I is displaced, being on a separate block, but there is a clear mark on the first tarsal where it would attach. The unguals are large, ranging in length from 20–23 mm in digits 1 and 2 to 10 mm in digit 5. The unguals are at most 4 mm broad proximally at the base, the ventral line is curved slightly downwards, and the dorsal margin much more so. Overall, the ungual shape is a rather rounded-ended structure without major recurvature, as seen in other rhynchosaurs. Each ungual bears a shallow groove along the sides distally, presumably for locking of the keratin sheath, and the ungual bone is pitted in the dorsal area.

Cladistic Analysis of the Rhynchosauria

Taxa

In attempting to re-analyse the relationships of rhynchosaurs with data from the new material of Fodonyx spenceri, we produced a data matrix as comprehensive as possible, including all rhynchosaur taxa and all available phylogenetically informative characters. We excluded Scaphonyx fischeriWoodward, 1907 because the taxon is ill defined, a probable nomen dubium (Langer and Schultz 2000).

Taxa were coded as follows: Mesosuchus (Dilkes 1998), Howesia (Dilkes 1995), Rhynchosaurus articeps, R. brodiei, and Fodonyx spenceri (Benton 1990 and pers. obs.), Otischalkia (Hunt and Lucas 1991), Ammorhynchus (Nesbitt and Whatley 2004), Stenaulorhynchus (Huene 1938 and pers. obs.), Mesodapedon (Chatterjee 1980), Hyperodapedon (Isalorhynchus) genovefae (Buffetaut 1983; Langer et al. 2000a), ‘Scaphonyxsulcognathus (Azevedo 1982; Schultz 1986; Azevedo and Schultz 1987), Hyperodapedon gordoni (Benton 1983 and pers. obs.), H. huxleyi (Chatterjee 1974), H. huenei (Langer and Schultz 2000), H. mariensisTupi-Caldas, 1933 (Langer and Schultz 2000), ‘Mariante rhynchosaur’ (Schultz and Azevedo 1990), Zimbabwe Hyperodapedon (Raath et al. 1992), Wyoming Hyperodapedon (Lucas et al. 2002), H. sanjuanensis (Huene 1929; Sill 1970; Azevedo 1984; Benton and Kirkpatrick 1989; Langer et al. 2000b), Supradapedon (Chatterjee 1980), and the Nova Scotia rhynchosaur (Chatterjee 1980; Benton 1990; pers. obs.).

As outgroup we selected three other archosauromorphs, Proterosuchus (coded from Cruickshank 1972, ankle from Carroll 1976, and braincase from Clark et al. 1993), Prolacerta (coded from Gow 1975, with braincase from Evans 1986, and skull from Modesto and Sues 2004), and Trilophosaurus (coded from Gregory 1945). These three represent the clades Archosauria (= Archosauriformes), Prolacertiformes and Trilophosauridae respectively, three major clades within Archosauromorpha that are close relatives of Rhynchosauria (Benton 1985; Evans 1988; Dilkes 1998).

We retained all taxa in the analysis initially, and did not delete any simply because they were coded for only a small proportion of the possible characters. Then, following the principles of safe taxonomic deletion (Wilkinson and Benton, 1995), we excluded eight taxa that were coded identically to more fully coded taxa: Mesodapedon (= Stenaulorhynchus and others); Otischalkia (= Rhynchosaurus articeps); Hyperodapedon genovefae, Zimbabwe Hyperodapedon, and Supradapedon (= Hyperodapedon gordoni and others); Wyoming Hyperodapedon and Nova Scotia rhynchosaur (= Hyperodapedon sanjuanensis).

Characters

Numerous characters of potential significance in determining the phylogeny of rhynchosaurs have been proposed and debated (Chatterjee 1974, 1980; Benton 1983, 1985, 1990; Dilkes 1995, 1998; Langer and Schultz 2000; Langer et al. 2000a, b). We carried out a review of all characters proposed hitherto and sought further characters. Our procedure was to: list all characters and variants in classic anatomical order from the tip of the snout to the last toe bone; combine synonymous characters used by different authors, and retain the clearest character definition in each case; consider the wording of each character definition carefully, and to revise and reword where necessary for clarity; code every character in the revised list for all 20 taxa; further revise the character definitions in the light of uncertainties in coding from specimens and from the literature; weed out characters that are ambiguous and/or uncodable.

In the end, 75 characters were retained that are codable (Appendix, first section). Of these, characters 13 and 52 are currently phylogenetially uninformative, so they were excluded from the cladistic analyses. They remain in the list though, since they cannot be coded for each rhynchosaur taxon and might in the end show variations in state.

The characters weeded from the master list include some that were used by previous authors. For example, the following six characters from Benton (1983, p. 709) were excluded because they could not be defined and coded in any meaningful way, or were invariant across all taxa: ‘dorsum sellae high’, ‘exoccipitals take part in occipital condyle’, ‘spheno-occipital tubera bifurcate ventrally’, ‘reduced presacral vertebrae 8 and 9’, ‘compressed scapula’, and ‘large lateral glenoid fossa’. Two further characters were excluded because they are redundant with others: ‘pterygoids directed mainly posteriorly’ (= character 29); adductor fossa extends more than halfway along jaw (= character 36).

Dilkes (1995, pp. 683–694) accepted 12 of Benton’s (1990) 26 characters, modified six, and excluded eight. We accept five of Dilkes’ (1995) modifications; the sixth was a concern about an unquantified definition of the relative sizes of the jugal and maxilla. This character (9 here) has been reworded for clarity. All of the eight characters excluded by Dilkes (1995) are retained:

  • 1Character 13, a frontal that is broader than long, is autapomorphic in Stenaulorhynchus; we argue that autapomorphies should be retained in character lists, and excluded at the analytical stage because future finds may confirm them in other taxa.
  • 2Ch. 22, absence of the parietal foramen, was said to be autapomorphic in Mesosuchus, but that is retention of a primitive state seen in the outgroup; Dilkes (1995, p. 684) argued that absence of the parietal foramen was plesiomorphic for rhynchosaurs, but its absence in archosaurs, Trilophosaurus, and from some specimens of Prolacerta is more likely a parallelism of the apomorphic state (basal amniotes all have the foramen).
  • 3Ch. 23, absence of the supratemporal, was excluded because of uncertainties in the literature, but these are somewhat clearer now (Dilkes 1998; this study).
  • 4Ch. 24, the shape of the ventral process of the squamosal, was indeed poorly expressed, but the character is quantified here.
  • 5Ch. 29, the occipital condyle lying roughly in line with the quadrates, was seen to be variable in the outgroup, but, with redefinition (Dilkes 1998; this study), this character can be retained.
  • 6Ch. 44, the relative widths of the medial and lateral portions of the maxillary tooth plate, is indeed an inapplicable character for the outgroups and for Mesosuchus and Howesia since they lack a tooth plate with a major groove; this character is retained here, however, since it may help sort out the derived rhynchosaurs.
  • 7Ch. 64, the relative lengths of the humerus and femur was said to be an autapomorphy of Hyperodapedon; we find a wider distribution of the derived state and so retain this character.
  • 8Ch. 70, relative size of the centrale, was a poorly defined character, but it has been redefined and distinguished from ch. 66.

The final tally of 75 characters (Appendix, first section) includes characters introduced by Chatterjee (1980), Benton (1985, 1990), Dilkes (1995, 1998) and Langer and Schultz (2000), although the wording is occasionally modified (see Table 1 for details of sources and modifications). One new character, 34, the shape of the supraoccipital, was mentioned by Dilkes (1995), but has not been used hitherto in phylogenetic analysis.

Table 1.   Outline classification of the rhynchosaurs, based on the cladistic analysis (Text-fig. 8B), and including the deleted taxa that, so far as codable, match more fully coded taxa. These are indicated with an asterisk (*) and placed immediately below one of the taxa with which their coded characters are identical.
Order Rhynchosauria Osborn, 1903
 Mesosuchus browni Watson, 1912
 Howesia browni Broom, 1905
 Family Rhynchosauridae Huxley, 1859
 Stenaulorhynchus stockleyiHaughton, 1932
 *Mesodapedon kuttyiChatterjee, 1980
 Rhynchosaurus articepsOwen, 1842
 Rhynchosaurus brodieiBenton, 1990
 *Otischalkia elderaeHunt and Lucas, 1991
 *Ammorhynchus navajoiNesbitt and Whatley, 2004
 Fodonyx spenceri (Benton, 1990)
 Subfamily Hyperodapedontinae Chatterjee, 1969
 ‘ScaphonyxsulcognathusAzevedo and Schultz, 1987
 Hyperodapedon hueneiLanger and Schultz, 2000
 Hyperodapedon gordoniHuxley, 1859
 *Hyperodapedon genovefae (Buffetaut, 1983)
 Hyperodapedon huxleyi Lydekker, 1881
 *Zimbabwe Hyperodapedon (Raath et al. 1992)
 *Supradapedon stockleyi (Boonstra, 1953)
 Hyperodapedon mariensis (Tupi-Caldas, 1933)
 ‘Mariante rhynchosaur’ (Schultz and Azevedo 1990)
 *Nova Scotia rhynchosaur
 Hyperodapedon sanjuanensisSill, 1970
 *Wyoming Hyperodapedon (Lucas et al. 2002)

Codings for a number of characters have been changed from codings in previous analyses:

  • 1Ch. 3, the open lower temporal bar, was coded confidently as ‘1’ for Rhynchosaurus articeps and Stenaulorhynchus by Benton (1990), even though that area of the skull is not preserved in any specimens, as indeed noted by Benton (1990, p. 231) and Dilkes (1995) in R. articeps and Huene (1938, p. 88) in Stenaulorhynchus; re-coded as ‘0’.
  • 2Ch. 4, the presence of a single median naris in Howesia, was coded as ‘?’ by Benton (1990), although its presence is indicated by Dilkes (1995, p. 668); re-coded as ‘1’.
  • 3Ch. 6, the premaxillary beak in the Nova Scotia rhynchosaur, was coded as ‘?1’ by Benton (1990), but there is no evidence for this; re-coded as ‘?’.
  • 4Ch. 9, the enlarged jugal in Scaphonyx sanjuanensis, was reported as ‘?0’ by Benton (1990), but this taxon shows the derived state (Sill 1970); re-coded as ‘1’.
  • 5Ch. 19, the relative length of the parietal and frontal in Howesia, was coded as ‘?1’ by Benton (1990), but ‘0’ by Dilkes (1998); the latter view is accepted.
  • 6Ch. 20, fusion of the parietals in Mesosuchus, was coded as ‘0’ by Benton (1990), and ‘1’ by Dilkes (1998); the latter view is accepted.
  • 7Ch. 23, absence of the supratemporal, was coded in Howesia as ‘?’ by Benton (1990) and ‘0’ by Dilkes (1998), in Stenaulorhynchus as ‘1’ by Benton (1990) and ‘?’ by Dilkes (1998), and in Scaphonyx fischeri as ‘1’ by Benton (1990) and ‘?’ by Dilkes (1998); the view of Dilkes (1998) is accepted for the first two, but not for the third, based on Huene (1929, p. 10).
  • 8Ch. 24, a broad ventral process of the squamosal, was coded as ‘?’ by Benton (1990) and ‘1’ by Dilkes (1998) in Howesia; the latter view is accepted. It was also coded as ‘0’ by Benton (1990) in F. spenceri, but is actually broad (pers. obs.); re-coded as ‘1’.
  • 9Ch. 29, the occipital condyle in line with the quadrates in F. spenceri, was coded as ‘?0’ by Benton (1990), but ‘1’ by Dilkes (1998); the latter view is accepted.
  • 10Ch. 35, the deep mandible in Howesia, was coded as ‘0’ by Benton (1990) and ‘?’ by Dilkes (1998); the dentary is indeed incomplete (Dilkes 1995), so Dilkes’ (1998) coding is accepted.
  • 11Ch. 36, the elongate dentary in Mesosuchus, was coded as ‘?0’ by Benton (1990), but is clearly ‘0’ (Dilkes 1998).
  • 12Ch. 39, ankylothecodont teeth, had been suggested in Trilophosaurus (Gregory 1945; Dilkes 1995, p. 673) but not in other members of the outgroup until Modesto and Sues (2004, pp. 347–348) argued strongly for ankylothecodont implantation in Prolacerta; both re-coded as ‘1’.
  • 13Ch. 40, absence of premaxillary teeth, was assumed to be the case in the Nova Scotia rhynchosaur by Benton (1990), but cannot be determined for Fodonyx, so it is re-coded as ‘?’ in both Hyperodapedon genovefae and Otischalkia; Benton (1990) recorded ‘?1’ for this character; it is confirmed by Buffetaut (1983) and Hunt and Lucas (1991) and so is re-coded as ‘1’ for both.
  • 14Ch. 45, lingual teeth on maxilla in Mesosuchus, coded as ‘?0’ by Benton (1990) and ‘0’ by Dilkes (1998); the latter is accepted.
  • 15Ch. 46, numbers of dentary tooth rows, coded in Mesosuchus as ‘0’ by Benton (1990) but ‘2’ by Dilkes (1998), and in Scaphonyx sanjuanensis coded as ‘?0’ by Benton (1990), and confirmed as ‘0’ by Langer et al. (2000b); the latter coding accepted in both cases.
  • 16Ch. 49, vomerine teeth absent in Rhynchosaurus articeps, coded as ‘1’ by Dilkes (1998), but cannot be determined for Fodonyx, so is coded ‘0’, as in Benton (1990).
  • 17Ch. 50, palatine teeth absent in Rhynchosaurus articeps, coded as ‘1’ by Dilkes (1998), but cannot be determined for Fodonyx, so is coded ‘0’, as in Benton (1990).
  • 18Ch. 51, pterygoid teeth absent, coded in Howesia as ‘0’ by Benton (1990), ‘1’ by Dilkes (1998), and in R. articeps as ‘?1’ by Benton (1990), ‘1’ by Dilkes (1998); in both cases, Dilkes’ (1998) codings are accepted; this character was coded as ‘?1’ by Benton (1990) in R. spenceri, but cannot be determined for Fodonyx, so is coded ‘?’ here.
  • 19Ch. 52, postaxial intercentra absent in Stenaulorhynchus, coded as ‘0’ by Langer and Schultz (2000), and ‘1’ by Dilkes (1998); retained as ‘1’.
  • 20Ch. 57, tapering chevrons in Scaphonyx fischeri, is coded as ‘?’ by Dilkes (1998), but is retained as ‘1’ here, based on Huene (1942).
  • 21Ch. 60, the proximal anchor-like shape of interclavicle in Howesia, coded as ‘1’ here, as indicated by Broom (1906, p. 595), although Dilkes (1995, p. 676) noted that the shoulder girdle region is now lost, and Dilkes (1998) coded as ‘?’.
  • 22Ch. 61, expansion of the posterior branch of the interclavicle in H. gordoni, coded as ‘0’ by Dilkes (1998) but it is clearly ‘1’ (Benton 1983, fig. 29d).
  • 23Ch. 64, the humerus longer than the femur, coded in Howesia as ‘0’ by Benton (1990) and ‘?’ by Dilkes (1998), but data in Broom (1906) and Dilkes (1995) indicate quite clearly that the femur is longer than the humerus, and so it is coded ‘0’ here.
  • 24Ch. 65, a femur with a wide distal end, coded in Howesia as ‘1’ by Dilkes (1998), but data in Broom (1906) and Dilkes (1995) suggest that the ratio of distal femur width to length is 0.31 (17/55), and should then be coded as ‘1’; it is coded ‘0’ for Otischalkia, but this is true only if the ascribed femur (Hunt and Lucas 1991) really is from this rhynchosaur.
  • 25Ch. 66, the presence of three proximal tarsals in Mesosuchus and Howesia, was coded as ‘1’ by Benton (1990), and ‘0’ by Dilkes (1998), based on the reinterpretation of the rhynchosaur ankle by Dilkes (1995); the latter view is accepted here.
  • 26Ch. 70, the presence of a large centrale in Mesosuchus and Howesia, was coded as ‘0’ by Benton (1990), and ‘1’ by Dilkes (1998), based on the reinterpretation of the rhynchosaur ankle by Dilkes (1995); the latter view is accepted here.
  • 27Ch. 71, absence of a contact between the centrale and distal tarsal 4 in Mesosuchus, was coded as ‘1’ by Dilkes (1998), even though the contact can be seen in Dilkes (1998, fig. 22), so it is coded as ‘0’ here.
  • 28Ch. 72, a small distal tarsal 4 in R. articeps, was coded as ‘0’ by Langer and Schultz (2000), although distal 4 is similar in size to distals 1–3 (Benton 1990, fig. 20), so it is coded as ‘1’ here.

Phylogenetic results

The phylogenetic analysis yielded 36 most parsimonious trees (MPTs), with a length of 114, consistency index (CI) of 0.68, retention index (RI) of 0.85, and rescaled consistency index (RC) of 0.58. The majority-rule consensus tree (Text-fig. 8A) shows reasonable evidence for the stem positions of Mesosuchus and Howesia respectively. Stenaulorhynchus is then the basalmost member of core Rhynchosauridae, followed by the two species of Rhynchosaurus, R. articeps and R. brodiei. Next comes Fodonyx, with some confidence, followed by a clade of species of Hyperodapedon and ‘Scaphonyx’. ‘Scaphonyxsulcognathus and Hyperodapedon huenei appear to be basal taxa to a crown unresolved group of several Hyperodaepdon species, with H. gordoni and H. huxleyi possibly paired.

Figure TEXT‐FIG. 8..

 Cladistic analysis of the relationships within Rhynchosauria. A, majority rule consensus tree of 36 MPTs. Figures indicate the proportions of source trees that include the node above. B, phylogeny of the Rhynchosauria, with names of major clades indicated. Bootstrap percentages (based on 1000 replicates) are indicated for each node. This tree (showing only nodes supported by more than 50% of bootstrap replications) is identical to the strict consensus tree of all 36 source trees, except that for ‘Scaphonyxsulcognathus.

Bootstrapping of the majority rule consensus (Text-fig. 8B) gave a less-resolved tree, with strong support for all clades indicated before, except the pairing of the two species of Rhynchosaurus (this clade is found in only 53 per cent of bootstrap replicates), the position of Rhynchosaurus between Stenaulorhynchus and Fodonyx (found in only 55 per cent of bootstrap replicates), and the crown clade of six species of Hyperodapedon (found in only 51 per cent of bootstrap replicates. The position of Fodonyx, between Rhynchosaurus and the Hyperodapedontinae, as defined by Langer and Schultz (2000), is reasonably secure (found in 88 per cent of bootstrap replicates).

One of the aims of this project had been to see whether fuller coding of the characters of Fodonyx would improve our understanding of its phylogenetic position, and of the Rhynchosauridae as a whole. Whereas only 29 out of the 75 characters (39 per cent) were coded for ‘Rhynchosaurusspenceri by Benton (1990), the new material allowed us to code 56 of the 75 characters (75 per cent) as well as to correct three of the previously published codings. Removing Fodonyx from the analysis and replacing it with ‘Rhynchosaurusspenceri, coded as by Benton (1990), and without the benefit of the new skull and postcranial material, gave rise to a similar cladogram. Fewer MPTs were generated (11), tree length is shorter (L = 110), and the consistency index and other tree statistics are better (CI = 0.70; RI = 0.87; RC = 0.61).

Discussion

More complete, and corrected coding, of characters might be expected to improve the cladistic result: there should perhaps be fewer MPTs (the data matrix is the same size both before and after), these ought perhaps to be shorter, and the consensus tree might be expected to show higher bootstrap values at relevant nodes. Our observation is that none of these occurred. Fuller, and corrected, codings for Fodonyx made the tree weaker in all those measures.

Evidently, the improved data recording for Fodonyx has increased incongruence. In the ‘old’ analysis, ‘Rhynchosaurus’ spenceri is linked to Hyperodapedontinae in 99 per cent of bootstrap replicates, whereas this falls to 88 per cent in the ‘new’ analysis. This suggests that the algorithm substituted hyperodapedontine-like values for some of the missing 61 per cent of data in the 1990 analysis. Even though Fodonyx shares its overall size and its skull shape with the hyperodapedontines, it retains plesiomorphic features that have been revealed by the new material and draw it down the cladogram, away from the crown Hyperodapedontinae.

Both analyses confirm that the Devon rhynchosaur is not a species of Rhynchosaurus, as had been assumed by Benton (1990), but that it represents a separate taxon, as suggested by Benton et al. (1993), Wilkinson and Benton (1995), and Langer and Schultz (2000). The bootstrap value of 88 per cent (Text-fig. 8B) is not overwhelmingly strong, but perhaps sufficient to justify its placement in the phylogeny. Furthermore, the Devon taxon cannot be hoisted into the crown clade, Hyperodapedontinae (Hyperodapedon + ‘Scaphonyx’), and the new analysis shows that it is in many regards more primitive than had been thought, based on less complete material. The cladistic analysis (Text-fig 8A–B) provides weak support for the pairing of R. articeps and R. brodiei as species of one genus, Rhynchosaurus, although the second species is so incompletely known that there are serious problems of missing data.

Within Hyperodapedontinae, as defined by Langer and Schultz (2000), the species of Hyperodapedon are hard to resolve. They found that the species of Hyperodapedon and ‘Scaphonyx’ formed a polytomy, and that ‘Scaphonyx’ sanjuanensis should be regarded as a species of Hyperodapedon. They found a pairing of the type species of Hyperodapedon, H. gordoni and H. huxleyi, and that H. huenei was a basal species. We also find a tentative pairing of H. gordoni with H. huxleyi, and H. huenei basal (Text-fig. 8A), but these divisions are not robust and not supported by bootstrap analysis, nor do they appear in the strict consensus tree (Text-fig. 8B). We find, as did Langer and Schultz (2000), that ‘Scaphonyxsulcognathus is the basal hyperodapedontine and, because the type species of the genus ‘ScaphonyxfischeriWoodward, 1907 is perhaps a nomen dubium (Langer and Schultz 2000), ‘S.sulcognathus requires a new generic name.

Classification

Taken conservatively, and including the previously excluded taxa, a complete classification of rhynchosaurs is indicated in Table 1. Character transformations at each node in the cladogram (Text-fig. 8B) are indicated in the Appendix.

Conclusions

As long suspected, the ‘Devon rhynchosaur’, called Rhynchosaurus spenceri by Benton (1990), belongs to a new genus, placed midway between other species of Rhynchosaurus and the Late Triassic hyperodapedontines Hyperodapedon and ‘Scaphonyx’. New materials of the Devon rhynchosaur, assigned to the new genus Fodonyx here, have yielded a great deal more information about the skull and skeleton. They confirm what had been determined before but add anatomical detail of the skull roof, snout, occiput, braincase, vertebral column, ribs, limb girdles, and hindlimb (Text-fig. 9).

Figure TEXT‐FIG. 9..

 Life restoration of Fodonyx spenceri, based on the skeletal material described here, depicted among the low plants that may have formed its diet. Drawing by Andrea Cobbett.

There is a time difference of perhaps 5–8 myr between the species of Rhynchosaurus and Fodonyx. The Helsby Sandstone Formation/Tarporley Siltstone Formation boundary of the Cheshire Basin, source of Rhynchosaurus articeps, is early Anisian in age, based on biostratigraphic and mapping evidence (Benton et al. 1994), whereas the Otter Sandstone Formation of Devon, source of Fodonyx spenceri, has been dated as latest Anisian on the basis of biostratigraphy (Benton et al. 1994) and magnetostratigraphy (Hounslow and McIntosh 2003). This age difference might explain why Fodonyx is more derived than Rhynchosaurus.

Acknowledgments

Acknowledgements.  We thank Malcolm Hart and Mark Hounslow, discoverers of the new skeleton and skull of Fodonyx respectively. We also thank David Hill (then Bristol City Museum) and Remmert Schouten (Department of Earth Sciences, University of Bristol) for preparing the skeleton and skull so skilfully. The partial skeleton has formed the centrepiece of exhibitions in museums in Plymouth, Exeter, and Sidmouth, and the skull is on show in the Department of Earth Sciences, University of Bristol: thanks to curators in all of these institutions for varied help.

Appendix

Characters used in phylogenetic analysis

The characters are listed in standard anatomical order. The plesiomorphic state for each character is coded as ‘0’, and derived states as ‘1’ and ‘2’, as appropriate.

Skull shape (5)

  • 1Maximum skull width relative to midline skull length: longer than broad (0), broader than long (1). [Benton 1983, p. 709; 1985, p. 133; 1990, character 1]
  • 2Skull height (maximum, measured in lateral view) relative to midline skull length: <50% (0), >50% (1).
  • 3Lower temporal fenestra: closed ventrally (0), open ventrally (1). [Dilkes 1998, ch. 4]
  • 4External nares: separate (0), single medial naris (1). [Benton 1985, p. 131; 1990, ch. 4, reworded]
  • 5Orientation of orbits – maximum area when seen in lateral or dorsal views: mainly lateral (0), mainly dorsal (1). [Langer and Schultz 2000, ch. 3] Possible linkage to ch. 1.

Dermal skull elements (23)

  • 6Premaxilla ventral margin: horizontal (0), down-turned and forming (1). [Benton 1985, p. 132; 1990 ch. 3, modified according to Dilkes 1995]
  • 7Premaxilla and prefrontal contact: absent (0), present (1). [Dilkes 1998, character 7]
  • 8Maxilla ventral margin: horizontal (0), convex (1). [Dilkes 1998, ch. 16]
  • 9Jugal area in lateral view: smaller than maxilla (0), larger than maxilla (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 5, reworded]
  • 10External surface of jugal: not ornamented (0), ornamented with crests and bosses dorsal to major diagonal crest (1). [Langer and Schultz 2000, ch. 5]
  • 11Major diagonal crest on jugal: short, and does not reach anterior portion of orbit (0), long, and reaches anterior portion of orbit (1). [Benton 1983, p. 709; Langer and Schultz 2000, ch. 4]
  • 12Width of jugal/postorbital bar: less than 40% (0) or more than 40% (1) of maximum orbital diameter. [Langer and Schultz 2000, ch. 6, reworded]
  • 13Frontal shape: longer than broad (0), broader than long (1). [Benton 1983, p. 709; 1990, ch. 6]
  • 14Frontal longitudinal groove: much deeper posteriorly than anteriorly (0), almost the same depth throughout (1). [Langer and Schultz 2000, ch. 2]
  • 15Shape of dorsal surface of frontal next to sutures with postfrontal and parietal: flat to slightly concave (0), depressed with deep pits (1). [Dilkes 1998, ch. 20]
  • 16Shape of dorsal surface of postfrontal: flat to slightly concave (0), depressed with deep pits (1). [Dilkes 1998, ch. 21]
  • 17Postfrontal enters border of upper temporal fenestra: no (0), yes (1). [Benton 1983, p. 709]
  • 18Length of anteroventral process of postorbital relative to posterodorsal process: longer (0), shorter (1). [Dilkes 1998, ch. 23, reworded]
  • 19Midline length of parietal relative to frontal: shorter (0), longer (1). [Benton 1983, p. 709; 1990, ch. 7, reworded]
  • 20Parietals: separate (0), fused (1). [Benton 1985, p. 131; 1990, ch. 8]
  • 21Parietal table: broad (0), constricted and with sagittal crest (1). [Dilkes 1998, ch. 26]
  • 22Parietal foramen: always, or sometimes, present (0), always absent (1). [Benton 1985, p. 132; 1990, ch. 9]
  • 23Supratemporal: present (0), absent (1). [Benton 1985, p. 133; 1990, ch. 10]
  • 24Breadth of ventral process of squamosal at mid-height, relative to breadth of lower temporal fenestra: <50% (0), >50% (1). [Benton 1990, ch. 11, reworded]
  • 25Quadratojugal anterior process: present (0), absent (1). [Dilkes 1998, ch. 35, reworded]
  • 26Ectopterygoid and palatine contribute to lateral border of suborbital fenestra: with maxilla (0), maxilla excluded (1). [Dilkes 1998, ch. 41]
  • 27Suture between ectopterygoid and pterygoids: complex overlap (0), simple overlap (1). [Dilkes 1998, ch. 142, polarity reversed]
  • 28Ectopterygoid reaches lateral corner of transverse flange of pterygoid: no (0), yes (1). [Dilkes 1998, ch. 42, reworded]

Braincase (6)

  • 29Relative position of occipital condyle: approximately in line with quadrates (0), markedly anterior to quadrates (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 12]
  • 30Relative midline length of basioccipital and basisphenoid: basisphenoid longer (0), basioccipital longer (1). [Benton 1983, p. 709; Langer and Schultz 2000, ch. 10, reworded]
  • 31Basipterygoid process: longer than broad (0), broader than long (1). [Benton 1983, p. 709; Langer and Schultz 2000, ch. 9]
  • 32Orientation of basipterygoid processes: anterolateral (0), lateral (1). [Dilkes 1998, ch. 43]
  • 33Club-shaped ventral ramus of opisthotic: present (0), absent (1). [Dilkes 1998, ch. 46, polarity reversed]
  • 34Supraoccipital shape: plate-like (0), inverted V-shape (1). [Dilkes 1995 1998]

Mandible (4)

  • 35Maximum depth of mandible relative to length: less than one-quarter (0), from one-quarter to one-third (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 13, reworded]
  • 36Length of dentary relative to length of mandible: half, or less than half (0), over half (1). [Benton 1990, ch. 14, reworded and polarity reversed, as in Dilkes 1995]
  • 37Jaw symphysis: formed largely by dentary (0), formed only by splenial (1). [Dilkes 1998, ch. 71]
  • 38Divergence of dentaries in front of symphysis: absent (0), present (1). [Dilkes 1998, ch. 72]

Teeth (13)

  • 39Tooth implantation: subthecodont or thecodont (0), ankylothecodont (1). [Benton 1985, p. 132; 1990, ch. 15]
  • 40Premaxillary teeth: present (0), absent (1). [Benton 1985, p. 131; 1990, ch. 2, modified]
  • 41Number of rows of teeth on maxilla: single row (0), multiple rows (batteries) of teeth (1). [Benton 1985, p. 132; 1990, ch. 16, modified]
  • 42Tooth occlusion: single sided overlap (0), flat occlusion (1), blade and groove jaw apparatus, where dentary blade(s) fit precisely into maxillary groove(s) (2). [Benton 1985, p. 133; 1990, ch. 17, modified according to Dilkes 1995]
  • 43Number of grooves on maxilla: none (0), one (1), two (2). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 18]
  • 44Relative width of tooth-bearing areas lateral and medial to maxillary groove: medial wider than lateral (0), lateral wider than medial (1). [Chatterjee 1980; Benton 1983, p. 709; 1990, ch. 19, reworded]
  • 45Lingual teeth on medial face of maxilla: absent (0), present (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 20, reworded]
  • 46Number of rows of teeth on dentary: one (0), two (1), more than two full rows (2). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 21, modified]
  • 47Lingual teeth on medial face of dentary: absent (0), present (1). [Langer and Schultz 2000, ch. 20, reworded]
  • 48Numbers of dentary teeth in tooth row: more in anterior half (0), more in posterior half (1). [Langer and Schultz 2000, ch. 18, reworded]
  • 49Vomerine teeth: present (0), absent (1). [Dilkes 1998, ch. 66]
  • 50Palatine teeth: present (0), absent (1). [Dilkes 1998, ch. 67]
  • 51Pterygoid teeth: present (0), absent (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 22]

Postcranial (24)

  • 52Postaxial intercentra: present (0), absent (1). [Langer and Schultz 2000, ch. 22]
  • 53Slender and tapering cervical ribs: present (0), absent (1). [Benton 1985, pp. 116–117; Dilkes 1998, ch. 77, polarity reversed]
  • 54Cervical rib accessory process: present (0), absent (1). [Dilkes 1998, ch. 78, polarity reversed]
  • 55Neural arches of mid-dorsals: deeply excavated (0), shallowly excavated (1). [Dilkes 1998, ch. 84, polarity reversed]
  • 56Ratio of lengths of caudal transverse processes to centra: >1.0 (0), <1.0 (1). [Dilkes 1998, ch. 89, polarity reversed]
  • 57Distal width of chevron bones relative to width of proximal area: same width or wider (0), narrower/tapering (1). [Langer and Schultz 2000, ch. 23]
  • 58Posterior process on coracoid: present (0), absent (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 23]
  • 59Coracoid foramen: restricted to coracoid (0), shared between coracoid and scapula (1). [Langer and Schultz 2000, ch. 24]
  • 60Shape of proximal end of interclavicle: broad diamond (0), gracile anchor (1). [Dilkes 1998, ch. 96]
  • 61Posterior stem of interclavicle: little change in width along entire length (0), expands distally (1). [Dilkes 1998, ch. 98, reworded]
  • 62Relative lengths of anterior and posterior blades of ilium: posterior blade longer (0), anterior blade longer (1). [Benton 1983, p. 709; 1985, p. 128; Langer and Schultz 2000, ch. 27]
  • 63Anterior apron of pubis: absent (0), present (1). [Dilkes 1998, ch. 104]
  • 64Relative length of femur and humerus: femur longer than humerus (0), humerus same length, or longer than, femur (1). [Benton 1983, p. 709; 1985, p. 133; 1990, ch. 24, modified]
  • 65Relative proportions of femur; distal width/total length: <0.3 (0), >0.3 (1). [Dilkes 1998, ch. 111]
  • 66Number of proximal tarsals: two (0), three, by incorporation of centrale (1). [Benton 1985, p. 131; 1990, ch. 25]
  • 67Perforating foramen: between astragalus and calcaneum (0), between distal ends of tibiale and fibula (1). [Dilkes 1998, ch. 115]
  • 68Relative length of astragalar facets for tibia and centrale: tibial facet longer (0), centrale facet longer (1). [Langer and Schultz 2000, ch. 28]
  • 69Lateral tuber of calcaneum: present (0), absent (1). [Benton 1985, p. 115; Dilkes 1998, ch. 116, polarity reversed]
  • 70Size of centrale relative to width of calcaneum: smaller (0), similar size (1). [Benton 1985, p. 131; 1990, ch. 26]
  • 71Contact between centrale and distal tarsal 4: present (0), absent (1). [Dilkes 1998, ch. 118, reworded]
  • 72Size of fourth distal tarsal relative to other distal tarsals: twice as large (0), approximately same size (1). [Langer and Schultz 2000, ch. 29]
  • 73Ratio of lengths of metatarsals I and IV: >0.4 (0), <0.4 and >0.3 (1), <0.3 (2). [Dilkes 1998, ch. 123; see also Modesto and Sues 2004, p. 348]
  • 74Ratio of lengths of digits 3 and 4: <0.81 (0), >0.81. [Dilkes 1998, ch. 124; recoded according to Modesto and Sues 2004, p. 348]
  • 75Length of metatarsal I relative to first phalanx: longer (0), same length (1). [Benton 1983, p. 709, reworded]

Matrix of characters used in phylogenetic analysis

Characters are present in states 0 and 1, with some in state 2, as defined in the descriptions of characters above. A question mark (?) indicates that the character cannot be coded in existing material, and ‘N’ indicates that the character is non-applicable and so cannot be coded. Non-applicable characters were treated as uncertain (‘?’) for the cladistic analyses.

 10203040506070
Proterosuchus0000000000000000000000000000000000010000000N0000000000000001000000000000100
Prolacerta0010010000000000000000001000000000000010000N0000000?00000000001000000000000
Trilophosaurus010000?0?0000000??0011???N??0100?0010011000N0000111?000?0000100000000000000
Mesosuchus0011011000000011100110001000000000000000010N02000001001000?0101000001100100
Howesia00110?110000001110011100?0000000?0?0??1?110N121??011???000???01010000100??0
Rhynchosaurus articeps00?1011100000011011111000111000011011111122?1210??1???010001100001101111110
Rhynchosaurus brodiei00?101111000001101???????1?1??????0?111112201210??????0????11??????????????
Stenaulorhynchus00?1011100001011011111?001?100001100111112201210111111100001100011100110211
Mesodapedon???????1??????????????????????????????1?12201210???????????????????????????
Fodonyx spenceri001111110110011011011100?11100100101111112201210111???1??????????1??11??21?
Rhynchosaurus spenceri11?1?1?1111????????????0?1110?????01111112201210??1???1????????????????????
Otischalkia?????11????????????????????????????????1???????????????????????00??????????
Ammorhynchus???????1??????????????????????????????1?121?1?1????????????????????????????
Hyperodapedon genovefae?????1?1??????????????????????????????111210011?????????????????11?1?1?????
Scaphonyx sulcognathus1101111111110011111111110111111111111111122002111111?????10????0???1???????
Hyperodapedon huenei110111111111011111111111011111111111111112200111111????????????????????????
Hyperodapedon gordoni110111111111011111111111011111111111111112100111111111111111110111111111211
Hyperodapedon huxleyi110111111111011111111111011111111111111112110111111111111101110111111111211
Hyperodapedon mariensis1???1????111?1????????11????111???11????12110111???1????110??0?0???1???1???
Mariante rhynchosaur110111111111011111111111011111111111111112110211111????????????????????????
Zimbabwe Hyperodapedon?????1?1??????????????????????????????11121?011???1????????????????????????
Wyoming Hyperodapedon?????1?1??????????????????????????????111211000???1????????????????????????
Scaphonyx fischeri110111111111011111111111011111111111111112110001111111111111100011110111111
Scaphonyx sanjuanensis110111111111?1?????111?10???111??1011111121100011111????111??0?0?0?1?0?1???
Supradapedon???????1??????????????????????????????1?12110??????????????????????????????
Nova Scotia rhynchosaur???????1??????????????????????????????1?1211000????????????????????????????

Character transformations

In this list, unequivocal characters are not annotated, convergences are marked with an asterisk (*), reversals with a negative sign (-), and characters that may be placed at different nodes, depending on whether accelerated transformation (A) or delayed transformation (D) is assumed. Derived states for multistate characters are indicated in parentheses after the character number. Nodes are indicated in Text-figure 9.

Node 1. Rhynchosauria (= Mesosuchus + all taxa above)

4, 6, 7, 8, 15, 16, 17*, 20, 21, 42(1), 46(2)*, 55, 61, 70, 73(1)

Node 2. Howesia + all taxa above

22, 33A, 37A, 38A, 39, 40A, 41, 45, 47, 49A, 51, 53A, 54A, 60, 65, 73(2)A, 74A

Node 3. Rhynchosauridae (= Stenaulorhynchus + all taxa above)

-17, 18, 19, 26, 27A, 28, 33D, 34, 37D, 38D, 40D, 42(2), 43(2), 49D, 50, 53D, 54D, 63, 66, 67, 71, 73(2)D, 74D, 75

Node 4. Rhynchosaurus + all taxa above

27D, 36, 56, 69*, 72

Node 5. Rhynchosaurus

-55, -65, -73(2->1)*, -75

Node 6. Fodonyx + all taxa above

5, 10, 11, 14, 17*, 31, 57, 58A, 68A

Node 7. Hyperodapedontinae (= ‘Scaphonyx’ sulcognathus + all taxa above)

1, 2, -3, 9*, 12, 23, 24, 29, 30, 32, 35, -45, 48, 58D, 68D

Node 8. Hyperodapedon huenei + all taxa above

-46(2->1)

Node 9. Other species of Hyperodapedon and ‘Scaphonyx

43(1), 44, 59

Mesosuchus: 25, 69*

Rhynchosaurus brodiei: 9*

Fodonyx: -16, -19, -33

Scaphonyxsulcognathus: -14

Hyperodapedon mariensis: 46(2)*, -59*

Hyperodapedon huxleyi: -59*

Hyperodapedon gordoni: -44

H. huxleyi + H. gordoni: 46(1), 62, 64

Scaphonyx sanjuanensis: -35, -66, -70

Anatomical abbreviations

a, angular; ar, articular; as, astragalus; bo, basioccipital; bpt, basipterygoid process; ca, calcaneum; ce, centrale; ch, choana; cv, chevron; d, dentary; ec, ectopterygoid; eo, exoccipital; f, frontal; fi, fibula; g, gastralia; h, humerus; hy, hyoid; I, digit one; il, ilium; is, ischium; iof, infraorbital fenestra; j, jugal; l, lacrimal; L, left; m, maxilla; man, mandible; n, nasal; op, opisthotic; p, parietal; pf, postfrontal; pl, palatal; pm, premaxilla; po, postorbital; prf, prefrontal; pt, pterygoid; q, quadrate; qj, quadratojugal; r, rib; R, right; sa, surangular; sc, scapula; so, supraoccipital; sp, splenial; sq, squamosal; st, supratemporal; ti, tibia; v, vomer; vt, vertebra; V, digit five; ?, unidentified bone; 1, 2, 3, 4, distal tarsals 1, 2, 3, 4.

Ancillary