A new Berriasian species of Goniopholis (Mesoeucrocodylia, Neosuchia) from England, and a review of the genus

Authors

  • MARCO BRANDALISE DE ANDRADE,

    Corresponding author
    1. Department of Earth Sciences, Faculty of Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK
      E-mail: marcobranda@yahoo.com.br
    Search for more papers by this author
  • RICHARD EDMONDS,

    1. Jurassic Coast World Heritage Team, Dorset County Council, County Hall, Colliton Park, Dorchester, Dorset DT1 1XJ, UK
    Search for more papers by this author
  • MICHAEL J. BENTON,

    1. Department of Earth Sciences, Faculty of Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK
    Search for more papers by this author
  • REMMERT SCHOUTEN

    1. Department of Earth Sciences, Faculty of Sciences, University of Bristol, Wills Memorial Building, Queens Road, Bristol BS8 1RJ, UK
    Search for more papers by this author

E-mail: marcobranda@yahoo.com.br

Abstract

A new species, Goniopholis kiplingi sp. nov., based on an exceptionally preserved skull from the Lower Cretaceous of England is described in detail. It shows great similarity with Goniopholis simus and Goniopholis baryglyphaeus, but can be distinguished by the presence of longer lachrymals, smooth (not edged) dorsal surface of the quadrate, and proportionally longer rostrum. A comprehensive phylogenetic analysis of Mesoeucrocodylia (104 taxa; 486 characters) focused on goniopholidids (14 species) places G. kiplingi as sister-group of G. simus, and as part of a monophyletic group also containing G. baryglyphaeus. The relationships of Nannosuchus gracilidens and three undescribed European taxa are explored, and preliminary analyses of Denazinosuchus kirtlandicus (Upper Cretaceous, USA) and ‘Goniopholis’ phuwiangensis (Lower Cretaceous, Thailand) are presented. The assignment of taxa to the genus Goniopholis is discussed. Goniopholis, in its traditional sense, is considered paraphyletic and a restricted updated definition is proposed, with comments on the evolution of other goniopholidids. Morphological characteristics of fragmentary material attributed to Goniopholis are not considered sufficient to secure their generic/specific assignment, and provide no support for the presence of Goniopholis in Gondwanan and/or Upper Cretaceous sedimentary units. Currently Goniopholis is restricted to the Upper Jurassic−Lower Cretaceous of Europe.

© 2011 The Linnean Society of London, Zoological Journal of the Linnean Society, 2011, 163, S66–S108.

INTRODUCTION

Goniopholis Owen, 1841 was a semi-aquatic neosuchian, superficially resembling modern crocodylians in general aspect and mode of life, although retaining ‘mesosuchian’ choanae and several other primitive characteristics, and specializations. Even though known from England for over a century (Owen, 1841, 1842, 1878, 1879), the morphology and evolution of Goniopholis has been poorly investigated, and the taxon has been used as a basket genus for a variety of species, and a plethora of fragmentary material. As a result, the biochronological and palaeobiogeographical distributions of the genus are poorly known.

Despite the breadth of other recent work on Goniopholis (Salisbury et al., 1999; Salisbury, 2002; Schwarz, 2002), no comprehensive cladistic analysis has tackled the validity of the genus and the family, compromising our understanding on their relationships and diversity. So far, the following works comprise only a part of the sample of known species: Tykoski et al. (2002) include four taxa; Jouve et al. (2006), Jouve (2009), Turner & Buckley (2008), and Pol, Turner & Norell (2009) five taxa; Young & Andrade (2009) six taxa; and Lauprasert et al. (2007) eight taxa. In all of them, Goniopholis simusOwen, 1878 and EutretauranosuchusMook, 1967 are present, but many taxa have never been included in a phylogenetic analysis [e.g. Amphicotylus lucasiiCope, 1878; Nannosuchus gracilidensOwen, 1879; Goniopholis baryglyphaeusSchwarz, 2002; Denaziosuchus kirtlandicus (Wiman, 1932)].

Goniopholis is indeed a key element in the phylogeny of Mesoeucrocodylia, but its use to establish the position of goniopholidids within Neosuchia is hardly appropriate if considered alone. Goniopholidids show a broad range of morphological adaptations that cannot be represented solely by G. simus and/or Eutretauranosuchus, especially because this last taxon has never been described in detail. An ampler sample of goniopholidids is needed to improve our knowledge of this particular group, and tackle questions on the group diversity, validity of taxa, and their distribution. Moreover, a larger sample of goniopholidids will provide a better understanding of the broad interrelationships of mesoeucrocodylians, including the evolution of extant forms.

Here we describe a new species of Goniopholis (Fig. 1), based on an exquisitely preserved specimen from the earliest Cretaceous Purbeck Limestone Group (Fig. 2), and test its relationships through a comprehensive phylogenetic analysis (104 taxa; 486 characters), including a large sample of goniopholidids (15 species). This analysis introduces for the first time several goniopholidids, including three undescribed specimens from Europe (Hulke, 1878; Dollo, 1883; Hooley, 1907; Fig. 3), which have been assigned to Goniopholis, but considered as new taxa on the basis of several distinctive characteristics (see Salisbury et al., 1999; Salisbury, 2002; Schwarz, 2002). Special attention is given to the comparison of the new Goniopholis with G. simus (Fig. 4) and G. baryglyphaeus (Fig. 5).

Figure 1.

Goniopholis kiplingi sp. nov. DORCM 12154 in its original slab (A), showing general aspect and material directly associated (indicated by white pointers): B, turtle scute; C, coprolite; D, bivalve. Although DORCM 12154 was extracted from the original matrix, the slab was preserved to show the context of the specimen with the sediment. Scale bars = 20 mm.

Figure 2.

Provenance and geological settings of Goniopholis kiplingi sp. nov. The holotype DORCM 12154 was found in a crumbling cliff (right) at the north end of Durlston Bay, near Swanage. The outcrop is part of the Jurassic Coast of England, and is considered a Natural Heritage Site. This is the first British specimen of Goniopholis to have its precise stratigraphical provenance identified to the exact bed (129b, Intermarine Beds), which is indicated by an arrow, on the right.

Figure 3.

Sample of European goniopholidids included in this study. A, reconstruction of Goniopholis simus, as interpreted from our examination of specimens. B, Nannosuchus gracilidens BMNH 48217 (type). C, Hooley's specimen BMNH R3876, erroneously referred to Goniopholis crassidens. D, Dollo's specimen IRSNB R47, erroneously referred to G. simus. E, reconstruction of ‘Mr Willett's crocodile’ based on BMNHB 001876, herein referred to as ‘Hulke's specimen’, a specimen that was earlier also considered to be a G. simus. F, Vectisuchus leptognathus SMNS 50984 (type), often referred to Goniopholididae, but actually showing closer affinities with Elosuchus (see the phylogenetic analysis herein). Note that maxillary depressions are absent in Nannosuchus and Vectisuchus. Scale bars = 20 mm. Drawings not to scale.

Figure 4.

Goniopholis simus, type specimen BMNH 41098 (A) and complementary slab BMNH 41098a (B), showing general features of the skull, including the palatines and perichoanal structures (C), and the frontal morphology at the transfrontal crest (D). Note the dorsal projection of the frontal at the medial line of the skull, indicated in D; and also the large amount of bony material in the complementary slab BMNH 41098a, holding part of the perinarial and periorbital crests, as well as a substantial part of the ornamentation. Abbreviations: FoST, supratemporal fossa; Fr, frontal; Mx, maxilla; Orb, orbit; Pal, palatine; Pt, pterygoid.

Figure 5.

Goniopholis baryglyphaeus MG ‘Gui Croc 1’ (type). The skull is heavily damaged and composed of fragments, but the bony surface is well preserved. The ornamentation is intense, and almost completely composed of pits. Note that the skull is divided in anterior and posterior sections, and only parts of the premaxillae are preserved. The transfrontal crest is indicated. Scale bar = 20 mm. Abbreviations: FLT, laterotemporal fenestra; FoMX, maxillar fossa; Fr, frontal; Ju, jugal; Mx, maxilla; Par, parietal; Pmx, premaxilla; Po, postorbital; Qj, quadrate-jugal; Sq, squamosal.

With this comprehensive analysis at hand, the relationships of goniopholidids are tested, as is their placement in the broad context of crocodylomorph evolution, and key characters and clades are discussed. We also preliminarily explore the relationships of ‘GoniopholisphuwiangensisBuffetaut & Ingavat, 1983 (Lower Cretaceous, Thailand) and Denazinosuchus kirtlandicus (Wiman, 1932) (Upper Cretaceous, USA) to evaluate their original placement in Goniopholis. Finally, we comment on the distribution of Goniopholididae, which should be further explored in future works.

Institutional abbreviations

BMNHB, Booth Museum of Natural History, Brighton, England; BMNH, Natural History Museum, London, England, United Kingdom; BRSUG, Department of Earth Sciences, University of Bristol, Bristol, England, United Kingdom; DORCM, Dorset County Museum, Dorchester, England, United Kingdom; MG, Museu Geológico, Lisboa, Portugal; PMU, Paleontological Museum, University of Uppsala, Sweden; SGM, Serviço Geológico e Mineralógico do Brasil.

GEOLOGY

Locality

The specimen (DORCM 12154, a skull minus mandible; Figs 1, 6–11) was discovered by R. E. whilst site monitoring on the Jurassic Coast (Dorset, England), where rocks from the Purbeck Limestone Group (PLG) are very well exposed. The specimen was found at the northern end of Durlston Bay, Swanage area (Fig. 2), at point GR SZ 038785 of the Ordnance Survey National Grid, roughly corresponding to global positioning system (GPS) coordinates 50°36′22.49478″N, −01°56′49.47318″W (latitude = 50.6060917195521, longitude = −1.94706264315903).

Figure 6.

Goniopholis kiplingi sp. nov. DORCM 12154 after extraction from its original slab (left), and reconstruction of the specimen (right), with interpretation of sutures and main structures of the skull. Note the evident ornamentation, almost completely composed of pits, and the elongated lachrymals. Scale bar = 20 mm. Abbreviations: Boc, basioccipital; CPN, perinarial crest; CLPO, lower periorbital crest; CUPO, upper periorbital crest; Eoc, exoccipital; FLT, laterotemporal fenestra; FoLac, lachrymal fossa; FoMX, maxillar fossa; FoPN, postnasal fossa; Fr, frontal; FST, supratemporal fenestra; Ju, jugal; Lac, lachrymal; Mx, maxilla; Na, nasal; Par, parietal; Prf, prefrontal; Pmx, premaxilla; Po, postorbital; Qj, quadrate-jugal; Qu, quadrate; Sq, squamosal.

Figure 7.

Premaxillae and perinarial structures of Goniopholis kiplingi sp. nov. DORCM 12154. A, oblique view of the specimen, making evident the deep anteronarial notch, perinarial crests and postnarial fossa. B, dorsal and C, ventral view, showing the naso-oral fossa/incisive foramen and alveoli. Narial notch and postnarial fossa indicated in A, and incisive foramen indicated in C. Scale bars = 20 mm.

Figure 8.

Dorsal view of skull at orbits, showing the periorbital structures of Goniopholis kiplingi sp. nov. DORCM 12154. The extreme elongation of the lachrymals (left) is one of the autapomorphic characteristics of this species. On the right, detail of the orbit and surrounding structures (indicated only at the lower image). In the left image, a putative bite mark is indicated at the anterior end of frontal, near the triple contact with nasal and prefrontal. Note the lachrymal-jugal suture next to the orbit, and inside the lachrymal indicated by the white pointer. Scale bars = 20 mm. Abbreviations: BPO, postorbital bar; CLPO, lower periorbital crest; CUPO, upper periorbital crest; Fr, frontal; FST, supratemporal fenestra; Ju, jugal; Lac, lachrymal; Mx, maxilla; Na, nasal; Par, parietal; Po, postorbital; Prf, prefrontal.

Figure 9.

Maxillary depressions of Goniopholis kiplingi sp. nov. DORCM 12154, with detail of the right structure in lateral view, showing its internal structure. Note that the area around the fossa is smooth, unornamented; and the interior is complex (see Andrade, 2009). The dorsal exposure of the right fossa is taphonomic; the structure faces laterally in other goniopholidids, and also on the left side of DORCM 12154. Common neurovascular foramina are indicated by white pointers. Scale bar = 20 mm. Abbreviations: Ju, jugal; Mx, maxilla.

Figure 10.

Distal end of skull of Goniopholis kiplingi sp. nov. DORCM-12154. The quadrate (left) is robust, but its dorsal surface smoothly progresses towards the posteromedial surface, where a foramen aerum can be seen, as indicated by the white pointers. A secondary opening to the foramen, at the dorsal surface of quadrate, is also indicated. In occipital view (right), the specimen is very similar to Goniopholis simus, with the cranioquadrate passage mostly covered by the paraoccipital process, as indicated by the arrow. Scale bars = 20 mm. Abbreviations: Ex, exoccipital; PaCQ, cranioquadrate passage; Par, parietal; PPoc, paraoccipital process; Qj, quadrate-jugal; Sq, squamosal.

Figure 11.

Dentition of Goniopholis kiplingi sp. nov. DORCM-12154. A, ventral view of the right alveolar margin at the premaxilla−maxilla notch, showing the position of the fifth premaxillary alveolus (as indicated), which is lateral relative to the first maxillary tooth. B, detail of the fourth maxillary crown (right side). C, detail of third premaxillary crown (left side). Both teeth are not completely erupted, and have a mesial−distal keel. The intense ornamentation creates a crenulated edge, as it covers the keel, producing a false-ziphodont serration, in this case visible even without the aid of optical resources. Scale bars = 10 mm.

The skull was broken in two pieces, the anterior part (rostrum, part of the skull table and most of the orbits) encased in a fallen block, whereas the posterior section (quadrates, occiput, most of skull table and part of the right orbit) remained in the cliff. Collection in situ provided an accurate stratigraphical assignment to the section. The lithostratigraphy of the section corresponds to the Intermarine Member sensuClements (1993), between beds 125 and 133, often referred to as ‘Shingle’. The specimen was preserved between alternating layers of sandy limestone and calcareous mudstone, fitting in all aspects bed 129b in Clements' (1993) lithostratigraphical scheme. Currently, the Intermarine Member of Clements (1993; beds 112–145) can be referred to simply as the Intermarine beds (Benton & Spencer, 1995), corresponding to a subsection of the Stair Hole Member sensuWesthead & Mather (1996) in the lower Durlston Formation.

Although thoroughly studied and reviewed (see Clements, 1993; Westhead & Mather, 1996; Ensom, 2002), many stratigraphical proposals have been adopted for the PLG. Here we follow the functional divisions of Westhead & Mather (1996), because of their basin-wide mappability, and the bedding divisions of Clements (1993), which provide a comprehensive distribution of sediments in the area of collection (Durlston Bay). For an overview of the stratigraphy of the ‘Purbeck beds’, see Ensom (2002), and other works within Milner & Batten (2002) for detailed palaeontological and geological aspects.

Taphonomy and preservation of DORCM 12154

The new specimen is a well-preserved skull, although neither the mandible nor postcranial remains were found in association. However, most sutures, ornamentation, and foramina can be easily identified. Fully erupted teeth are completely absent from the skull and were also not found in the matrix. The bony elements of the skull are all preserved in their original places, indicating that there was not enough time for full disarticulation. The absence of teeth, palpebrals, and mandible, combined with the completeness of the skull, suggest that definitive burial took place after the initial phases of decomposition of the carcass.

A series of shallow circular marks can be seen on the surface of the skull. These are here considered as the possible result of interspecific antagonistic interactions, which in crocodylians commonly occur in the form of bites, either lethal or nonlethal (e.g. Buffetaut, 1983; Avilla, Fernandes & Ramos, 2004; Vasconcellos & Carvalho, 2007). Similarly, the most complete skull of Nannosuchus (BMNH 48217, type) also shows bite marks on the rostrum, although in this case the perforations are much deeper. The postulated bite marks in DORCM 12154 are so shallow that they are probably not the cause of death, but the lack of secondary ossification suggests that they occurred close to the end of the individual's life.

After burial, the specimen endured strong dorsoventral flattening and light lateral deformation, the skull being only slightly asymmetric. As a result, a few structures are damaged and displaced (e g. maxillae, palate, postorbital bar), and the skull has lost part of its three-dimensional structure. This is made evident by comparison of both maxillae in dorsal view, the orientation of the posterior-most alveoli in the computed tomography (CT) scan, and the position of the pterygoid and nasopharyngeal duct, partially visible through the supratemporal fenestrae.

The skull was collected still embedded in the rock matrix, in horizontal position, concordant with the orientation of the sediment layers, but upside down (palate facing upwards, skull table downwards). Therefore, the underside of the slab actually revealed the dorsal surface of the skull. The matrix shows centimetric to millimetric layers of fine sediment, mostly greyish to whitish in colour. In addition, a darker and harder layer was directly associated with the bony surface in all areas that were prepared mechanically. The particular way the sediment was found inside the alveoli suggests that a weak aqueous flux was present during the depositional process. A strong current is unlikely, as it would have turned over the skull prior to definitive burial.

The surrounding matrix includes further fossil material, distributed in different layers, within bed 129b. The underside of the slab is crowded with unidentified bivalve shells, and includes one turtle bony plate (Fig. 1). A few infaunal bivalves were found in close association with the specimen, but it is likely that they were buried in the sediment after the death of this Goniopholis specimen and the deposition of those layers. There are neither encrusting organisms on the skull, nor any obvious borings, both of which support the idea of rather rapid burial. Plant foliage was collected in a clay layer set directly above the specimen and donated to the BMNH. Another plant specimen was discovered in situ and awaits description by Paul Ensom. Massive amounts of wood were present in bed 133 (the ‘Red Rag’), also in the Intermarine beds.

SYSTEMATIC PALAEONTOLOGY

Superorder Crocodylomorpha Walker, 1970

Order Crocodylia Gmelin, 1789 (sensu Martin & Benton, 2008; = CrocodyliformesBenton & Clark, 1988)

Suborder Mesoeucrocodylia Whetstone & Whybrow, 1983

Infraorder NeosuchiaBenton& Clark, 1988

Family Goniopholididae Cope, 1875

GenusGoniopholisOwen 1841

Type species: Goniopholis crassidens Owen, 1841

1841 Goniopholis crassidens Owen

1878 Goniopholis simus Owen

1999 Goniopholis crassidens (Owen) Salisbury et al.

1999 Goniopholis simus (Owen) Salisbury et al.

2002 Goniopholis simus (Owen) Salisbury et al.

2002 Goniopholis baryglyphaeus Schwarz

2008 Goniopholis cf. simus Andrade et al.

Goniopholis kiplingisp. nov.

Etymology: Specific name after Rudyard Kipling, British novelist, author of ‘The Jungle Book’ amongst others and an important disseminator of natural sciences through literature, from the end of the 19th century to the beginning of the 20th century.

Holotype: DORCM 12154, well-preserved skull, dorsoventrally flattened, lacking mandibles and most teeth (Figs 1, 6–11).

Type-locality: Cliff at GR SZ 038785, north end of Durlston Bay, Swanage, Jurassic Coast, Dorset, England, UK; GPS coordinates: 50°36′22.49478″N, −01°56′49.47318″W.

Type-stratum: Bed 129b (Clements, 1993), Intermarine beds sensu Wimbledon inBenton & Spencer (1995; = Intermarine Member sensuClements, 1993), Stair Hole Member sensuWesthead & Mather (1996), Purbeck Limestone Group; Berriasian, Lower Cretaceous (Salisbury et al., 1999; Milner & Batten, 2002). Sediment directly associated with the specimen preserved and catalogued under the same number as the type.

Diagnosis for the species (autapomorphies indicated by ‘*’): Goniopholidid mesoeucrocodylian with the unique combination of the following features: skull sublongirostrine, with rostrum between 66–70% of skull length (*); premaxillae in the shape of an axe head, in dorsal view, with deep notches at premaxilla−maxilla suture; naris orientated fully dorsally; perinarial crests present; narial border high throughout, but anterolateral border with deep notch on both sides; nasals widely excluded from narial border by posterordorsal branches of premaxillae; naso-oral foramen present and fully open, formed by premaxillae and maxillae; naso-oral fossa diamond-shaped; maxillary depressions with all borders well defined, and three internal chambers, the first being the largest, and chambers decreasing in size posteriorly and bearing a neurovascular foramen at the bottom; lachrymals proportionally long and narrow, extending anteriorly beyond the maxillary depressions, but not tapering anteriorly (*); upper periorbital crest strong and divided into a longer anterior (lachrymal) and shorter posterior (prefrontal) sections; lachrymal fossa with extensive participation of jugal, but not of prefrontal; prefrontal extending posteriorly to the posteromedial border of the orbit; frontal participating at posteromedial (primary) border of the orbit; frontoparietal suture well within the intertemporal bar; supratemporal fossae subpolygonal, at least as large as orbit; frontal−postorbital suture complex; postorbital with anterolaterally directed process, short and robust; quadrate with two foramina aerum, one at posteromedial corner, one at dorsal surface; dorsal and posteromedial surfaces of quadrate separated by a smooth ridge (not acute crest); teeth crowns keeled, with enamel ornamented by thin, well-defined basi-apical ridges, non-anastomosed (base of the crown) to poorly anastomosed (appex), creating crenulations (false-serrations) on the keels.

Description ofGoniopholis kiplingi DORCM 12154

General features of the skull

The skull (Fig. 6) is sublongirostrine (rostrum 68.5% of skull length; RL/SWPo = 1.76) and its relative length is slightly longer than in G. simus and G. baryglyphaeus. Perinarial, periorbital, and transfrontal crests are present. As in most other neosuchians, there is no antorbital fenestra or fossa. Orbits are orientated laterally and dorsally, with only a small anterior component, as in G. simus. The length of the DORCM 12154 skull (475.6mm; premaxilla to occipital surface) gives a total body length estimate of 3.47 metres based on ‘body length/head length’ regressions provided by Sereno et al. (2001).

The laterotemporal fenestra is triangular and faces laterodorsally, as in most Mesoeucrocodylia. The supratemporal fenestra is ample, but still slightly smaller than the orbit. A supratemporal fossa surrounds the fenestra itself, and has the approximate size of the orbit. The subpolygonal outline of the supratemporal fossa is more evident on the left side, whereas the border of the fenestra on the right side looks rounder because of preservation/preparation. A post-temporal fenestra is present and evident, wider than high and close to the medial line of the skull. The squamoso-otoccipital fenestra was obliterated by the dorsoventral deformation of the specimen.

The rostrum was relatively narrow and moderately high, prior to deformation. It broadens posteriorly, smoothly fitting the skull at orbits, with rostrum limits poorly defined, as in most other related genera (e.g. Sunosuchus, Eutretauranosuchus, Siamosuchus), and in contrast to the morphology in Pholidosaurus. The dorsal surfaces of frontal, parietal, postorbital, and squamosal composing the skull roof are flat, forming the skull roof.

Postnarial fossa: DORCM 12154 has a small, poorly defined fossa immediately posterior to the narial opening (Fig. 7; character 41). The internal surface of this depression is also ornamented. The depression itself is mostly shallow and with poorly defined limits. Sampling this potentially informative character is difficult because it is frequently overlooked and assumed to be taphonomic. It requires direct observation, explicit mention in the text, or imaging techniques that provide three-dimensional information (stereophotography, tomography data). Direct examination of specimens showed that at least G. simus, G. baryglyphaeus, and the undescribed ‘Hulke's goniopholidid’ share the same feature, and it is absent in Sarcosuchus, Elosuchus, and dyrosaurids. Calsoyasuchus completely lacks such a depression (Tykoski et al., 2002), as shown by tomography data. Images of Amphicotylus (D. Pol, pers. comm. 2010) suggest the presence of the fossa in this taxon, and indeed the structure may be present in several other goniopholidids, but it is not properly noted for taxa such as Denazinosuchus and Eutretaurnosuchus.

Lachrymal fossa: Although the antorbital fossa is absent in DORCM 12154, a small fossa is present immediately anterior to the orbit, in the lachrymal area (Fig. 8; character 53). In this fossa, the bone surface is unornamented and extremely concave. The limits of this fossa are made evident by the anterior end of the periorbital crests. Within Goniopholididae, the fossa is present in most taxa (Goniopholis, Amphicotylus, Nannosuchus), but it is unknown in Sunosuchus and Siamosuchus, and absent in Calsoyasuchus. The undescribed goniopholidids (BMNHB 001876, BMNH R3876, IRSNB R47) included in the study also have a small, well-defined fossa. Most other groups of crocodylomorphs (e.g. sphenosuchians, protosuchians, notosuchians, metriorhynchids, atoposaurids) have an even larger and deeper fossa, whereas Bernissartia and eusuchians seem to lack this structure completely.

The naming of this fossa is confusing because of its position, and also its variable composition within Mesoeucrocodylia. This fossa may be erroneously identified as the antorbital fossa, as in many cases these structures occupy the same general area, anterior to the eye (e.g. Sphagesaurus, Notosuchus; see Pol, 2003; Andrade & Bertini, 2008a, b, c). Nonetheless, they cannot be considered homologous structures because both are present as separate elements in several taxa (e.g. metriorhynchids, Hsisosuchus, Mahajangasuchus). The use of names such as preorbital fossa or postantorbital fossa would only add more confusion, and so Young & Andrade (2009) defined this as the prefrontal−lachrymal fossa whilst revising Geosaurus giganteus, as in this taxon the prefrontal constituted almost half of the concavity. However, in the case of Goniopholis and other goniopholidids, the prefrontal does not participate in the anterior border of the orbit (character 160). As in other European Goniopholis and related forms (Hulke's, Dollo's, and Hooley's goniopholidids), the jugal of DORCM 12154 broadly participates in the ventral part of the fossa, contributing to the anterior border of the orbit. In Amphicotylus, Sunosuchus, and Calsoyasuchus, neither prefrontal nor jugal seem to advance on the lachrymal region. Despite these problems, the fossa is always located immediately anterior to the orbit and the lachrymal bone always participates in the fossa. Therefore, the topological position of the fossa is taken as a reference, not its composition, and this structure is here designated as the lachrymal fossa, simply meaning that this is a fossa located in the lachrymal region, immediately anterior to the orbit.

Ornamentation and crests: Considering the dorsal elements of the skull, ornamentation does not occur inside the narial opening, alveolar margin of maxillae and maxillary depressions, temporal fossae, quadrate, medial section of the quadratojugal, and postorbital bar. All other surfaces are heavily ornamented with a modified version of the same pattern usually found in eusuchians, characterized by pits and grooves. In this specimen, however, no grooves (sensuBuffrénil, 1982) are present, and pits are no more than ellipsoid (Figs 6–8). Despite most of the ornamentation constituting an irregularly distributed fabric of pits, there are long rows of subcircular pits on the lateral surface of the jugal. Overall, the ornamentation is greatly similar to that found in most goniopholidids, Elosuchus, and Vectisuchus.

Unlike the morphology found in eusuchians and sebecians, the bony surface next to the alveolar margin is mostly smooth, lacking ornamentation, both in the premaxilla and maxilla (Figs 7, 9). In DORCM 12154, the ornamentation only approaches the ventral rim of the lateral maxillary surface at the fourth and fifth maxillary alveoli. Therefore, the alveolar margin resembles the condition found in several notosuchians and basal crocodyliforms. The smooth surface completely surrounds the maxillary depressions. The skull ornamentation does not progress ventral to the ventral-most neurovascular foramina of the premaxilla and maxilla, and the same occurs in the mandible (Ballerstedt casts BMNH R5260/R5261). This is common to all Goniopholis specimens examined, as well as to the undescribed goniopholidids and Nannosuchus (even though it lacks the maxillary depressions). This pattern of ornamentation has never been recognized in the group, although it must be noted that several of the specimens are partially embedded in matrix and mostly exposed dorsally, which presumably hampered observations.

Perinarial and periorbital crests are common in goniopholidids (Figs 7, 8). Upper orbital crests are also found in other neosuchians, and are particularly well developed in certain alligatorids (e.g. Melanouchus, Caiman; see Brochu, 1999). In Goniopholis, perinarial and periorbital crests are always robust and evident, and in many cases an interorbital crest links both upper periorbital crests at their posterior ends, dividing the frontal anterior process from its main body. However, the presence of a sagittal interorbital crest is not as frequent amongst goniopholidids, and is only reported for Siamosuchus (see Lauprasert et al., 2007) and Sunosuchus (e.g. Wu, Brinkman & Russel, 1996; Fu, Ming & Peng, 2005; Schellhorn et al., 2009).

The perinarial crests of DORCM 12154 (Fig. 7; character 29) begin medial to the distal end of the narial opening and progress anteriorly, encircling the naris for at least three-quarters of the narial perimeter, being absent from the anterolateral and anterior borders of the naris. At their posteromedial end, they are small and narrower, gradually increasing in size and width to the lateral margin of the naris, and then smoothly decreasing again towards their anterior ends. These crests are in contact posterior to the naris, but not truly continuous because of the presence of the medial suture between both premaxillae. The area immediately posterior to the medial contact of the perinarial crests is smoothly concave, forming a postnarial fossa. Perinarial crests seem to be absent from North American taxa and Sunosuchus, but occur in Amphicotylus (D. Pol, pers. comm. 2010). However, perinarial crests are present in Hulke's goniopholidid, although no contact occurs between both, posterior to the narial opening.

The lower orbital crest (Fig. 8) is simply the dorsally projecting margin of the anterior jugal ramus, which runs lateral and ventral to the orbit, reaching the area immediately anterior to the orbit (lachrymal area). At this point, the dorsal edge of the jugal curves laterally, breaking from the jugal edge, and becomes a truly projecting blade that transects part of the dermal surface of the jugal. As the blade projects, the edge becomes evident as a crest, creates a notch at the anterior border of the orbit, and delimits the ventral end of the lachrymal fossa. This morphology is not exclusive to DORCM 12154, and can also be found at least in Amphicotylus (D. Pol, pers. comm. 2010), G. simus, and the goniopholidids reported by Hulke, Dollo, and Hooley.

The upper periorbital crests (Fig. 8) are preserved on both sides of DORCM 12154, set immediately medial to the dorsal border of the orbit. They provide extensive support to the palpebrals. In DORCM 12154 the dorsal periorbital crests are partially divided by a deep notch, where the lachrymal−prefrontal crosses the bony surface towards the orbit. The lachrymal section of the crest is elongated, whereas the prefrontal section is knob-like. Similar crests with a shallower notch also occur in most specimens of G. simus and G. baryglyphaeus, but the crest of Nannosuchus has no notch and is not as robust. As the robustness and the notch are more extreme in larger specimens, the differences between the crests seen in these specimens may reflect the ontogenetic stage. Upper periorbital crests provide extensive support to a single palpebral element. These crests are absent in Hulke's, Dollo's and Hooley's specimens and also in Sunosuchus and Eutretauranosuchus, but are present in Goniopholis stovalli and possibly Amphicotylus.

DORCM 12154 shows a transversally orientated crest, transecting the frontal medial to the orbits and connecting with the posteromedial ends of the upper orbital crests (Fig. 8). Although continuous, the limits between these crests remain fairly recognizable. Whenever it occurs, the transfrontal interorbital crest is poorly arched, with the concavity facing anteriorly. At least G. baryglyphaeus, Nannosuchus, G. stovalli and possibly Amphicotylus also share this crest. As observed by Hooley (1907), both BMNH R3876 and Hulke's specimen lack the crest completely. In DORCM 12154, as well as in G. simus, this crest isolates the main body of the frontal from its anterior process, and also delimits a difference in the depth of the dorsal surfaces of both, a condition absent in Nannosuchus and unknown in other taxa that have the transfrontal crest.

A similar transverse crest is present only in caimans (Caiman latirostris, Melanosuchus), but is not as robust and is positioned on the rostrum, medial to the cranial border of the orbits, or anterior to it (see Brochu, 1999). The interorbital crest present in Goniopholis and other goniopholidids are here interpreted as different structures from those in alligatorids, although presumably analogous in their biomechanical function, and possibly contributing to the general robustness of the skull.

At the medial section of the crest, DORCM 12154 has a buttressed area, dorsally projected, flatter anteriorly and roughly triangular, in dorsal view (character 139). The same structure is present at least in G. simus (type BMNH 41098 and the Ballerstedt casts BMNH R5259 and R5262; see Fig. 4), but unreported in most other taxa. The undescribed Hulke's goniopholidid bears a low, buttressed ‘hump’ at the same position, but this is poorly defined, with a circular profile (dorsal view). Furthermore, it is low, feebly projecting dorsally. Goniopholis baryglyphaeus has a buttressed area, but it expands transversally with the transfrontal crest and neither assumes the triangular profile nor truly projects dorsally. No other crocodylian apart from these British goniopholidids seems to have such a swollen area or projection medial to the orbits.

Rostrum

Nares: The narial opening (Fig. 7) is dorsally orientated, as its posterior border is dorsal to most of its anterior border, and is exclusively surrounded by the premaxillae. As in most Goniopholis, there is a perinarial crest at the rim of the naris (see ‘Ornamentation and crests’, above). The narial opening is proportionally wide (∼30% of premaxillae width), as in G. simus, G. baryglyphaeus, and Amphicotylus lucasii, but unlike Hulke's specimen. The narial opening is also slightly wider than long (AP/ML = 0.89), as in G. simus. It has the same heart-shaped profile that Salisbury et al. (1999) attributed to G. simus, but the posterior projection of the anterior border is not as pronounced as a result of preservation. The naris of both DORCM 12154 and G. simus differ substantially from the naris of Hulke's specimen, which has a nearly perfect circular profile (AP/ML = 1.02).

Premaxillae: Premaxillae (Fig. 7) contact only the maxillae and nasals. Next to the premaxilla and maxillary suture there is a pronounced notch (for occlusion of a putative enlarged fourth dentary tooth) at the lateral alveolar margin. The lateral-most border of the premaxilla is directed posterolaterally, partially bounding the notch. This gives the premaxillae the appearance of a wide axe blade in dorsal view. The morphology is very similar to that in G. simus, G. baryglyphaeus, Hulke's specimen, G. stovalli, Amphicotylus lucasii, and even the pholidosaurid Meridiosaurus, but unlike Nannosuchus, Siamosuchus (unknown in Hooley's specimen and Denazinosuchus), which are paddle-like in dorsal view.

Posterior to the naris, the premaxillae meet medially broadly, excluding the nasals from the narial border. The anterior rami of the premaxillae also form a high vertical wall at their medial contact. Immediately lateral to this contact there is a pronounced notch (character 34), at the anterolateral border of the narial opening. The overall morphology is very similar to that in G. simus, but no other specimen is known to have such a deep narial notch. The condition is unknown or undescribed in Eutretauranosuchus, Sunosuchus, Denazinosuchus, and Dollo's and Hooley's specimens (where the premaxilla is not preserved). Shallow unornamented areas are present in the same position in Hulke's specimen, and are considered homologous. Paired anterior narial notches are probably present also in Amphicotylus (D. Pol, pers. comm. 2010).

The posterior ends of the premaxillae extend between and dorsal to the maxillae, and contact the nasals, creating a bifurcated profile. These posterior extensions are relatively long (∼50% of anteroposterior length posterior to premaxilla−maxilla notch; ∼60% of anteroposterior length posterior to naris). From the narial opening, the naso-oral fossa is visible. The naso-oral fenestra (= incisive foramen) is single, longer than wide, and has a sinusoidal outline. It is located at the posterior end of the diamond-shaped naso-oral fossa. Tomography data confirm the presence of the naso-oral fenestra, and also the participation of the maxillae in its distal border (a summary of CT procedures is available in the Supporting Information File S1).

The premaxillary surface next to the alveolar margin is smooth and faces slightly ventrally, with an edge clearly separating the ornamented and unornamented surfaces. Six to seven neurovascular foramina are present at this edge, in each premaxilla. The whole alveolar set is projected ventrally relative to the palatine surface of the premaxilla, but neither as evident as in Sarcosuchus or Terminonaris (see Sereno et al., 2001), nor at a lower level than the maxillary alveolar margin. Most alveoli are preserved in the specimen, but the first left and the second right alveoli are partially broken and deformed. There are five alveoli in each ramus, the third and fourth being particularly enlarged, and the fifth the smallest. The fifth premaxillary alveoli are at a more lateral position than the first maxillary alveoli. Overall, the alveoli are weakly procumbent, but the anterior dentition was most likely not, because of the curvature of the crowns. It is possible to recognize two occlusion pits for dentary teeth at the palatal ramus of each premaxilla, immediately posterior to the projected alveolar margin. The medial pit is the deepest, and is posterior to alveoli 1 and 2. The second one is evidently shallower and is located posterior to alveoli 3 and 4.

Maxilla: In dorsal aspect (Fig. 6), the maxillae are mostly horizontal laminae, vertically placed only near their lateral end (although this feature is enhanced by the taphonomic dorsoventral deformation). They contact the premaxillae anteriorly, nasals dorsally, and lachrymal and jugal posteriorly. Ventrally, the maxilla probably also contacted the palatine and ectopterygoid.

The maxillary (or rostral) depressions occur in a wide range of taxa, including all true Goniopholis, as well as Amphicotylus, Calsoyasuchus and Hulke's and Hooley's specimens (Fig. 3). Similar structures also occur in the basal mesoeucrocodylian Hsisosuchus (Jurassic, China; see Gao, 2001; Peng & Shu, 2005), but apparently extend slightly posteriorly, reaching the jugal (Gao, 2001). The rostral depressions found in goniopholidids and hsisosuchids are here considered as potentially homologous, because of their similar structure and location. At least in goniopholidids these structures seem to be specialized neurovascular foramina and have a specific morphology that may be related to sensorial or glandular functions (Andrade, 2009). In DORCM 12154 the morphology of the maxillary depressions (Fig. 9) is similar to the type of G. simus (BMNH 41098) and the cast material for the Ballerstedt specimens (BMNH R5259–R5262), as well as to G. baryglyphaeus and Hulke's specimen. The border is evidently well defined throughout the perimeter of the depression and the concave surface has a complex internal structure. At least three internal chambers were present, delimited by shallow and mostly vertical ridges. In DORCM 12154 and most other specimens there is evidence of an enlarged neurovascular opening at the bottom of each chamber (Fig. 9). Amongst the specimens examined, only the type specimen of G. simus shows, on its right side, what seems to be an incipient fourth chamber, between and above chambers 1 and 2. This seems to be an aberrant condition, as this ‘chamber’ is much shallower, bears no neurovascular opening, and is not present on the left side. Salisbury et al. (1999) reconstruct the maxillary depression of G. simus with five internal chambers, all anteroposteriorly aligned. This, however, is not the condition seen in the holotype, or in the casts of the Ballerstedt specimens, which show only three internal chambers.

Although the chambers are not preserved in Hulke's specimen, they are clearly visible in Hooley's specimen (BMNH R3876), also in the number of three. The maxillary depressions are completely absent from Nannosuchus, Denazinosuchus, and Vectisuchus. In the case of Nannosuchus, it is not clear if this absence may be a result of its ontogenetic stage.

Nasals: The nasals (Fig. 6) are mostly parallel and elongated, tapering only in the anterior-most section, as in most neosuchians and basal notosuchians. They neither reach the narial opening, nor get close to its border. Instead, the nasals extend forward and contact the premaxillae, separating the posterior ends of these elements. They also contact the lachrymals, prefrontals, and frontal. The nasals widely separate the anterior sections of the frontal and prefrontals, and slightly separate prefrontals from lachrymals. The nasals only contact the medial edge of the lachrymals. The dorsal surface is flat throughout the entire medial contact (character 75), with no evident medial groove (as in thalattosuchians and several notosuchians) or deep trench (Hsisosuchus, Calsoyasuchus). The ornamentation is almost exclusively ornamented with pits, lacking the typical elongated grooves of eusuchians. At least two well-defined circular bite marks are visible at mid-rostrum, and a third shallower mark is on the posterior-most section of the right nasal, close to the prefrontal.

Periorbital elements:

The periorbital elements of Goniopholis kiplingi DORCM12154 were preliminary described, with emphasis in the comparison with other goniopholidid taxa (Andrade & Hornung, 2011). The following section focuses on the morphology of G. kiplingi itself.

Lachrymals: In DORCM 12154, the lachrymals contact the maxillae, nasals, prefrontals and jugals, and take part in the anterior border of the orbit. As in all other Goniopholis, the lachrymals are relatively wide laminae that extend alongside the rostrum, separating completely the maxilla from the prefrontals. Unlike most other goniopholidids, they do not taper progressive to the anterior end, but end rather abruptly. In DORCM 12154 they are proportionally long (AP/ML of lamina anterior to orbits = 3.15), when compared to all other Goniopholis (AP/ML ≅ 1.5) and even Hulke's specimen (AP/ML = 2.85), extending well beyond the anterior-most border of the prefrontals and past the anterior-most tip of the rostral depressions (Figs 6, 8).

Prefrontals: These are relatively small, longer than wide, and well ornamented. The prefrontal contacts the lachrymal, nasal, frontal, and palpebral. The prefrontal certainly does not contact the postorbital in the dorsal part of the orbit, but preservation prevents observation on the ventral surface, where a contact would be possible. The posterior section of the prefrontal extends posteriorly and constitutes the primary medial border of the orbit, as described for G. simus by Salisbury et al. (1999). The prefrontal also provides extensive support for the palpebral (Andrade & Hornung, 2011).

Frontal: The frontal of DORCM 12154 is a single element, with both sides being completely fused through most of its length (character 132). However, a sagittal suture is retained at the medial line of the anterior process (Figs 6, 8). The main body of the frontal is wide, flat, and has strong ornamentation (characters 133–134). It is well developed posterior to the orbit, where it expands laterally anterior to the supratemporal fenestra and posteriorly into the intertemporal bar. This gives the frontal an overall ‘T-shaped’ outline. Its anterior process is narrow and long, progressing anterior to the orbits and slightly anterior to the prefrontals. The anterior-most border of the anterior frontal process is truncated, with a diminutive intrusion from the nasal, on the medial line.

Dorsally, the frontal contacts the nasals and prefrontals (anteriorly), postorbitals and palpebrals (laterally), and parietal (posteriorly). The frontoparietal contact is medial to the supratemporal fenestra. The frontal posterolateral edge constitutes the full length of the anterior border of the supratemporal fossa, preventing the contact between postorbital and parietal, in dorsal view (although this contact may occur ventral to the frontal, inside the fenestra). The frontal of DORCM 12154 reaches the primary border of the orbit, but its participation is restricted to the posteromedial corner (character 141).

Postorbital: The postorbital is very similar to that of G. simus, bearing a short and robust process dorsal to the bar, anterolaterally orientated. The dorsal surface (and the process) are intensely ornamented. The suture with the frontal is complex, and the contact with the parietal occurs deep inside the supratemporal fossa. Posteriorly, the contact with the squamosal is hardly identifiable, and in dorsal aspect both elements seem to be almost fused. A ventral ramus, clearly distinct from the main body, projects ventrally and contributes to the postorbital bar.

Jugal: The jugal is a triradiate element, with anterior, posterior, and ascending rami. Both anterior and posterior rami are flattened and intensely ornamented. The anterior ramus expands dorsoventrally and anteriorly. The ascending ramus contrasts with the remaining rami, as it is unornamented and mostly cylindrical.

The anterior edge extends dorsomedially, contacting the lachrymal. The dorsal edge is a robust crest, projecting anteriorly and posteriorly. At the anterior end, the crest curves anterolaterally, breaking the contour of the anterior ramus and creating a notch, ending at the dorsolateral surface of the ramus. Posteriorly, the crest bounds the lateral end of the ascending ramus. The jugal participates extensively in the lachrymal fossa (Fig. 8). Ventrally, the anterior jugal ramus has at leacst two neurovascular foramina next to the maxilla, facing ventrally (character 180). This is shared with other goniopholidids examined, except for Nannosuchus, where the anterior end of the jugal is not well preserved. Such foramina are also present at least in Pholidosaurus, dyrosaurids, Hylaeochampsa, and Bernissartia, and are easily identifiable in all crown crocodylians. Many notosuchians (e.g. Sphagesaurus) share a similar state, but the foramen is a single enlarged foramen, anteriorly orientated (see Andrade & Bertini, 2008a).

The posterior ramus is not as expanded. It extends to the quadratojugal in a diagonal contact. A vascular foramen is present close to the ventral end of the ascending ramus, on both sides in DORCM 12154, at the anteroventral corner of the laterotemporal fenestra. At least on the left side, the dorsomedial surface of the posterior jugal ramus shows no fewer than two other openings along its dorsomedial surface. This feature is not evident in other taxa, but its presence can be easily masked by preservation.

Postorbital bar: The postorbital bar is a composite structure composed of the descending ramus of the postorbital, ascending ramus of jugal, and often by a dorsal ramus of ectopterygoid. In DORCM 12154 the structure is poorly preserved on both sides following taphonomic dorsoventral deformation. Nonetheless, it is possible to recognize that it was subdermic and not ornamented, robust, and cylindrical (subcircular cross-section). The area of the jugal−postorbital suture is destroyed, and it is impossible to identify the contact. As the dorsal ramus of the jugal projects posteromedially on both sides in DORCM 12154, the inclination of the bar is perfectly identifiable. The left ectopterygoid, as preserved, shows that it took part in the medial face of the bar, but it is not possible to recognize whether it contacted the postorbital inside the bar or not. However, it was most likely similar to the condition found in Hulke's specimen, where the ectopterygoid extends dorsally, taking part in the medial face of the bar, even reaching the ventral ramus of the postorbital.

Unfortunately, it is not possible to confirm the presence of a vascular opening on the postorbital close to the dorsal end of the bar, because of poor preservation of the area in DORCM 12154. However, the surface immediately posterior to the dorsal end of the bar is smooth and depressed, and it is quite possible that vascular foramina were present in that fossa. A similar situation is seen in Hulke's specimen, which shows an even deeper fossa.

Palpebrals: Palpebrals were not preserved, but there are evident scars for the attachment of an anterior (single) palpebral. These scars are deep, long, and are preserved on both sides in the orbital margin. Each scar ranges from the anteromedial to the posteromedial border of the orbit, encompassing most of the lateral surface of the upper periorbital ridges. These should have been ‘delta-like’, massive, although not nearly similar to the wide palpebrals found in baurusuchids. Amongst goniopholidids, only Nannosuchus and Hulke's specimen are reported to have palpebrals preserved (Hulke, 1878; Joffe, 1967), which are fairly consistent with the delta-like profile inferred for DORCM 12154, although this structure is evidently more gracile in Nannosuchus. Hooley's and Dollo's specimens have narrower and more elongated palpebrals, robust and integrated to the orbital border. In DORCM 12154, as in most other goniopholidids, there is no evidence of a posterior palpebral. The morphology of the posterior orbital border and comparison with related forms show that its presence is unlikely.

In the absence of palpebrals, the orbits of DORCM 12154 assume a triangular profile. Overall, the morphology of the orbit is similar to G. baryglyphaeus and most G. simus, but specimens of this later taxon show an intraspecific variability, with the orbit being either triangular or subcircular (Andrade & Hornung, 2011).

Skull roof

The skull roof is wide, flat, and intensely ornamented. Pits composing the ornamentation tend to be larger at the medial line of the skull roof, and never elongated (see Buffrénil, 1982), therefore not forming the grooves commonly seen in eusuchians. The intertemporal bar is relatively wide, showing a double concave profile, contrasting with G. simus, where the intertemporal bar has straight and parallel lateral borders.

Overall, the lateral margins of the skull roof (= upper temporal bars) are parallel in DORCM 12154, as in most goniopholidids (character 148). The temporal bars are strongly sinusoidal (character 149), as in G. simus, G. baryglyphaeus, and Nannosuchus, contrasting with the straight profile seen in Hulke's specimen, most North American forms, pholidosaurids, dyrosaurids, and possibly Siamosuchus. The sinusoid profile is also shared by Amphicotylus (D. Pol, pers. comm. 2010; contraMook, 1942). The bar anterior to the supratemporal fossa is broad, almost as wide as the bar posterior to the fossa.

Supratemporal fenestrae and fossae: In DORCM 12154, the left fossa shows better preservation. It is subcircular to polygonal, with straight anterior and posterior margins, angled lateral margin, and a convex round medial margin. The lateral margin is formed by the intersection of two straight margins, which meet at an angle (∼80°) in the middle of the supratemporal bar. From the anterolateral to posteromedial borders (postorbital, squamosal, and parietal), the internal margin projects slightly over the fenestra. Most of the anterior to medial margins (frontal to anterior end of parietal) do not project at all, but are acute and well defined. However, in the anteromedial corner, the transition from the fossa to the skull table is somewhat smoother. The same morphology is present in other Goniopholis specimens, as well as in several other related forms studied (i.e. Nannosuchus, Hulke's specimen, Pholidosaurus, Sarcosuchus).

Parietal and squamosal: The parietal of DORCM 12154 is single, well sculpted, and composes all the posteromedial margin of the skull table. Its posterior border is well preserved and mostly straight, not projecting significantly over the occipital surface. The suture with the frontal is evident at the middle of the intertemporal bar, and is mostly ‘v’-shaped (concavity facing posteriorly).

The squamosal contacts the parietal medial to the supratemporal fenestra. The contact with the postorbital is not evident, but is likely to be lateral to the external angle of the supratemporal fenestra. In dorsal view, the element is flattened, intensely ornamented, and extends posterolaterally over the quadrate, as an ornamented horizontal lobe (giving the skull table its sinusoid profile). There is, however, no unsculpted lamina departing from this lobe and reaching the quadrate, or from the lateral edge of the supratemporal bars.

Quadratojugal: The quadratojugal is firmly attached to the jugal and quadrate, and the sutures with these elements are difficult to recognize. Nonetheless, the contact with the posterior jugal ramus is slightly displaced, showing that the suture is ventral to the laterotemporal fenestra and orientated anterodorsally−posteroventrally. As in most mesoeucrocodylians, the lateral end is intensely ornamented. The suture with the quadrate curves laterally near the condyles, next to the ornamented surface of the quadratojugal, and does not reach the articular surface. The unornamented ascending process narrows as it progresses medially. Its anterior edge probably held a long quadratojugal spine as in G. simus, but the area is not preserved in DORCM 12154.

Quadrate: The quadrate is robust, and is posterodistally orientated (Fig. 6). The proximal end of the quadrate is partially covered by the skull table on both sides, as a result of the dorsoventral deformation of the skull, obliterating the tympanic area and preventing observation of possible structures (i.e. preotic siphoneal foramen). CT data show that the quadrate is massive in structure, and is neither perforated nor internally pneumatic (a characteristic restricte d to certain notosuchians and basal groups of crocodylians). The quadrate distal condyles are level with each other, and are positioned at a more posterior position than the occipital condyle. The posteromedial margin is steep throughout, and separated by the dorsal surface by a smooth and poorly defined edge, similar to Pholidosaurus. In opposition, Hulke's specimen, G. simus, and at least G. stovalli have an acute edge, separating the dorsal and the posteromedial surfaces.

As for most crocodylians, including modern forms, DORCM 12154 has a foramen aerum on the dorsomedial face of the quadrate, next to the quadrate condyle. However, a second opening is present at the dorsal surface of the quadrate, at a short distance (laterally and posteriorly) from the ‘primary’ opening of the foramen aerum (Fig. 10), a feature that is preserved on both sides and is clearly visible. This is unexpected, because crocodylomorphs only possess one foramen aerum on each quadrate, being either on the dorsomedial face (most crocodilians so far described, including Sphagesaurus and Sunosuchus) or the dorsal surface (alligatoroids, mostly; see Brochu, 1999), and so far only DORCM 12154 shows this condition.

Occipital surface

The observation of the supraoccipital in DORCM 12154 is constrained by the dorsoventral deformation of the skull (Fig. 10). However, it is still possible to recognize that the element does not reach the skull roof.

The exoccipitals (Fig. 10) are mostly well preserved, and the only damaged area is dorsal to the foramen magnum, where both laminae meet. In this section, the exoccipitals loosened as a single lamina from their main body, and became horizontal. Consequently, the surface actually faces dorsally, rather than posteriorly, resembling the horizontal and posteriorly projecting lamina that covers the foramen magnum in Hsisosuchus, although in DORCM 12154 this is clearly an artificial condition caused by the dorsoventral deformation.

The lateral edges of the exoccipitals (= paraoccipital processes) are evident but they do not extend lateral to the skull roof or the squamosals. Their morphology is very similar to G. simus, G. baryglyphaeus, Pholidosaurus, Sarcosuchus, Bernissartia, and the specimens noted by Hulke (1878), Dollo (1883), and Hooley (1907). The ventral edge of the paraoccipital process projects ventrally, shielding the distal end of the cranioquadrate canal, which is not evident on the occipital surface. In G. simus and G. baryglyphaeus the cranioquadrate canal is only partially enclosed (dorsally, medially, and ventrally), but laterally exposed, and runs as a sulcus from the auditory meatus to the occipital surface (Salisbury et al., 1999; Schwarz, 2002). A laterally exposed cranioquadrate canal is also present at least in the Hulke specimen, and some non-goniopholidid neosuchians (e.g. Allodaposuchus, Hylaeochampsa; see Delfino et al., 2008). In DORCM 12154 the area is obliterated due to dorsoventral deformation, but preliminary observation of the CT data seems to corroborate the same morphology for this specimen. In any case, it is evident that the paraoccipital process of DORCM 12154 does not contact the quadrate extensively, lateral to the passage (character 308).

Palate, choanae, and pterygoid

Although the ventral side of the specimen remains mostly hidden by matrix, and detailed description must await proper analysis of CT data, it is possible to present preliminary information on the palatal structures. Further information on the CT data on the palate, including a simplified reconstruction, is available in the Supporting Information File S1.

The secondary palate of DORCM 12154 is fully formed, with palatine rami of premaxillae and maxillae present and in contact on the medial line. The nasopharyngeal duct is complete and opens in the posterior half of the skull, between the palatines and pterygoids. There is no evidence of a ventral exposure of the vomer in the palate. As in most mesoeucrocodylians, including related forms such as Hulke's specimen, the maxillary palate progresses anteriorly to the premaxillary palate. This contrasts with the condition found in eusuchians, where the premaxillary palate progresses posteriorly over the maxillary palate. Although not perfectly preserved, the internal naris has an overall morphology similar to that in G. simus, where the choanal opening is ample, longer than wide, and with a lanceolate profile (Fig. 4).

The naso-oral fenestra (= incisive foramen, foramen incisivum) is present and fully open. The passage is deep and the surrounding fossa resembles closely that in G. simus (although in G. simus the fossa does not form a fenestra; see Salisbury et al., 1999). As in most mesoeucrocodylians, the palatine rami of the maxilla take part in the posterior border of the naso-oral fenestra.

Maxillo-palatine fenestrae (= palatine fenestrae; Gasparini, 1971) are known from a series of taxa assigned to Goniopholididae, such as Eutretauranosuchus, Sunosuchus, and Calsoyasuchus, where they are also referred to as the ‘primary choanae’ (as in Tykoski et al., 2002). They are however absent from DORCM 12154, as in G. simus and Hooley's and Hulke's specimens.

Unfortunatelly, it is impossible to verify precisely the relationships amongst palatine, ectopterygoid and pterygoid, but it seems clear that DORCM 12154 had neither a palatine−ectopterygoid contact, nor a palatine bar, as in all neosuchians. Therefore, the pterygoids reach the distal end of the suborbital fenestra, separating the ectopterygoids from the palatines.

Dentition

The dentition of DORCM 12154 was almost entirely lost prior to burial, and only a few unerupted crowns remain preserved and accessible. Although most of the ventral side of the specimen is covered by matrix, the alveolar margin was exposed on the right side, showing the overall distribution of teeth (Fig. 11). Further, a small section of the left side was also exposed near the ectopterygoid−jugal contact, showing the last few alveoli.

There were five alveoli in the premaxilla, the third and fourth being largest, and the fifth the smallest. Another 19 alveoli are present in the maxilla, the fourth and fifth being largest, but with a second wave of large teeth at about the 11th alveolus. This comprises a total of 24 upper teeth. Alveoli vary in size along the premaxilla and maxilla, as expected for a goniopholidid (Table 1). The maxilla has two ‘waves’ of enlarged teeth (festooning), but its lateroventral margin does not project ventrally/laterally, coincident with the second set of enlarged alveoli. The morphology agrees with G. simus and G. baryglyphaeus, and contrasts with the festooning seen in peirosaurids and related forms, and also with Bernissartia and Crocodylus, where the feature is particularly evident. Although the mandible was not preserved, the dentary most likely supported 20–24 teeth, with a large tooth at the symphysis occluding in the premaxilla−maxilla notch. The first dentary tooth was probably large, fitting into the occluding pits seen in the premaxillae, near the medial line. The fifth premaxillary alveolus is positioned at a more lateral position than the first. There is a considerable gap between both teeth at the premaxilla−maxilla notch. Although premaxillary teeth seem to have been at the same level as the maxillary teeth before the flattening of DORCM 12154, it is clear that the premaxillary dentition projects ventrally relative to the palate.

Table 1.  Alveoli measurements for Goniopholis kiplingi sp. nov., specimen DORCM 12154, compared with alveoli from several other neosuchians, including extant
 AlveoliPremaxillaMaxilla
1234512345678910111213141516171819221
1234567891011121314151617181920212223242526
  1. All data collected upon first hand examination of specimens, except for Siamosuchus, taken from Lauprasert et al. (2007). Measurements in mm; inline image, alveolus not preserved or not accessible; X, alveolus not present; ?, presence of alveolus is uncertain; †, fossil taxon.

Taxa Goniopholis kiplingi 11.511.5819.615.68.259.6213.513.22.119.415.612.112.714.013.815.813.4611.51.09.79.79.59.09.0 X X
DORCM 12154
Goniopholis baryglyphaeus inline image inline image 6.57.8 inline image 6.36.67.110.010.07.77.07.38.67.06.55.7 inline image inline image 6.3 ? ? 5.14.8 X X
‘Gui Croc 1’
Siamosuchus 7.07.011.011.06.07.09.01.013.813.88.58.08.08.08.07.07.07.07.07.07.06.06.05.0 X X
SDRC PPC 1
Hooley's (1907) specimen inline image inline image inline image inline image inline image 8.011.012.018.315.09.07.5 inline image inline image 13.4 inline image inline image 7.2 inline image inline image ? ? ? ? ? ?
BMNH R3876
Hulke's (1878) specimen9.711.712.514.38.01.511.112.717.719.215.46.915.816.213.214.213.114.014.012.6 inline image 1.09.28.57.46.7
BMNHB 001876
Dyrosaurus maghribensis 9.59.313.66.9 X 9.517.29.59.59.09.012.011.711.71313.011.712.310.611.610.78.9 X X X X
IRSNB R146
Bernissartia 1.751.651.652.3 inline image 0.81.72.82.73.3 inline image 1.642.72.55 inline image 2.93.34.44.54.42.9 X X X X X
IRSNB R46
Alligator 7.238.2710.6213.679.469.4213.014.3516.4813.2210.710.111.911.4614.713.910.789.859.418.87 X X X X X X
BMNH 1973-2-2-21
Crocodylus niloticus 13.610.6521.6727.6613.712.614.4715.822.13213.815.717.821.923.621.318.214.911.2 X X X X X X X
BMNH 1949-1-1-2
Crocodylus acutus 6.976.0810.9712.987.69.7710.2810.7113.0916.4311.029.989.7411.7712.7712.510.359.488.5 X X X X X X X
BRSMG Ad5637
Crocodylus porosus 18.6227.032.016.0 X 13.015.3217.8529.542.118.5219.1415.518.6726.8720.8317.3215.110.32 X X X X X X X
BMNH 1947-3-5-33

Only the third premaxillary (left side) and fourth maxillary (right side) crowns are preserved and exposed, although partially (Fig. 11). As these are non-erupted teeth, wear or damage from use did not affect these crowns, which are well preserved. They are very similar to crowns described and attributed to G. simus, G. baryglyphaeus, Hulke's and Hooley's specimens, and evidently more robust than the teeth of Nannosuchus. The crown itself is bulbous, as it is slightly inflated, with an acuminate apex (although the anterior tooth is slightly more slender when compared with the posterior one). The crown is subcircular in cross section, without evident lateral compression, but the lingual and labial surfaces are asymmetric. The labial face is strongly arched, whereas the lingual is not as much. Enamel ornamentation is present on both lingual and labial surfaces, in the form of basi-apical ridges. These are well defined, conspicuous, and closely packed, but low. Overall, the ornamentation is non-anastomosed at least at the base and mid-crown, whereas apically there is a reasonable degree of anastomosis. A distinct keel runs on mesial and distal faces of the crown. As the anastomosed ornamentation extends toward the apex and keel, a crenulated surface is formed, leading to a false-ziphodont condition. This pattern of tooth crown morphology seen in DORCM 12154 (Fig. 11) is very similar to the one found in G. simus and G. baryglyphaeus, but is also consistent with the crown morphology seen in taxa from other related groups, such as Pholidosaurus and Machimosaurus.

PHYLOGENETIC ANALYSIS

The present analysis is the result of a new data set, strongly based on the literature (e.g. Clark, 1994; Ortega et al., 2000; Sereno et al., 2001; Gasparini, Pol & Spalletti, 2006; Jouve et al., 2006; Turner & Buckley, 2008; Young & Andrade, 2009), and includes 165 new (or deeply revised) characters. The matrix (486 characters; 104 terminals) is used for a main analysis, as well as for three additional exploratory runs. In the main (‘core’) analysis, 471 characters are parsimony-informative.

The revision of mesoeucrocodylian relationships and extension of the data set are taken from Andrade (2010), and include further aspects that will be published elsewhere. Nonetheless, a more complete account of procedures can be found in the Supporting Information File S1 (including taxa, character list, matrix, line commands) and S2 (output files).

Taxon sampling in Goniopholididae

Most taxa included in the analysis are non-eusuchian mesoeucrocodylians (75%), and 59% of the total sample was examined directly. A complete list of taxa used in the analysis can be found in the Supporting Information (File S1), providing sources of bibliographic data and specimens examined.

The matrix includes 15 taxa that either pertain to Goniopholididae, or are often referred to the family, including: five species frequently included in the genus Goniopholis; three unnamed taxa previously classified as either G. simus or G. crassidens (Hulke, 1878; Dollo, 1883; Hooley, 1907); three species of Sunosuchus; the monospecific Eutretauranosuchus, Siamosuchus, and Calsoyasuchus; and the putative goniopholidid Vectisuchus. The impact of Nannosuchus (= Goniopholis gracilidens) in the phylogeny was evaluated by means of an additional run, where topologies (including and excluding the taxon) were compared. The phylogenetic affinities of Denazinosuchus and ‘G.phuwiangensis were tested through additional exploratory runs. A summary review of goniopholidid taxa included in the analysis is presented below.

Goniopholis and taxa previously referred to the genus

Goniopholis simus is well known from several specimens (e.g. BMNH 41098, IPB R359, GZG 0061), comprising skull, mandible, and postcranium (usually disarticulated, but preserved in association). Material comes either from the Purbeck Limestone Group (PLG; formerly, ‘Purbeck beds’) of England, or from the Obernkirchen Sandstone (Bückeberg Formation, mid to late Berriasian of Germany. It is possible that G. simus is conspecific with G. crassidensOwen, 1842, as pointed out in the literature (Salisbury et al., 1999; Salisbury, 2002; Schwarz, 2002) and supported by new postcranial evidence (Hornung, Andrade & Reich, 2009). Although G. crassidens is the first member of its family, it is generally accepted that the taxon is limited to its type specimen BMNH 3798, a disarticulated collection of postcranial elements and partial mandible, also from the PLG (see Salisbury, 2002). Indeed, the incompleteness of BMNH 3798 makes comparison with other specimens of very limited value. Here we second earlier morphological studies (Salisbury et al., 1999; Salisbury, 2002; Schwarz, 2002), and consider BMNH 3798 as nondiagnostic material, referring only to G. simus (as in other phylogenetic works, e.g. Gasparini et al., 2006; Lauprasert et al., 2007; Turner & Buckley, 2008). Therefore, G. crassidens was not included in the present phylogenetic analysis. Unfortunately, the majority of specimens of G. crassidens and G. simus are historical, lacking an accurate lithostratigraphical provenance as pointed out by Salisbury (2002). This author traced the origin of specimens recovered from Dorset, based mainly on the works of Owen (1842, 1878, 1879), and indicated the origin of the Goniopholis specimens from Swanage derived from the Intermarine beds (= ‘Intermarine Member sensuClements, 1993; formerly ‘Upper Building Rocks’). This layer corresponds to the Stair Hole Member (Durlston Formation; Berriasian, Lower Cretaceous). Other reports of Goniopholis from different localities range from the Middle Jurassic to Lower Cretaceous, but are limited to nondiagnostic remains that cannot be assigned to a genus (Benton & Spencer, 1995; Salisbury et al., 1999; Salisbury, 2002).

Goniopholisgracilidens (Fig. 3) comes from the ‘Beccles Residuary Marls’ (untraced bed, either 108 sensuClements, 1993, or below; PLG), Berriasian (Lower Cretaceous) of England. This taxon was originally described as N. gracilidens by Owen (1879), but has been considered as immature specimens of either G. crassidens or G. simus (Kalin, 1933; Joffe, 1967; Clark, 1986). More recently, Salisbury (2002) supported the assignment of G. gracilidens to the genus Goniopholis, but regarded it as a valid independent taxon. Indeed, the status of the lectotype (BMNH 48217) as an immature specimen (e.g. frontals/parietals not fused in the medial line) is not sufficient to invalidate the species, and to assign the material to either G. crassidens or G. simus. Therefore, we second Salisbury (2002) in recognizing its specific status. However, this taxon has never been included in previous analyses, and its generic assignment is not secure. As a result, we prefer to refer to the taxon in its original form, Nannosuchus, and to elaborate on its generic assignment in the light of the phylogenetic relationships of the group. Therefore, the inclusion of Nannosuchus in the analysis allows a test of the assignment to Goniopholis, proposed by Salisbury (2002).

Goniopholis baryglyphaeus Schwarz, 2002 comes from the Guimarota coal mine, in Portugal, corresponding to the Alcobaça Formation (Kimmeridgian, Upper Jurassic). The sole specimen of G. baryglyphaeus includes a reasonably complete skull (Fig. 5), as well as postcranial material, overall well preserved but fragmentary in nature. All material, originally deposited in Germany as IPFUB Gui Croc 1, was returned to Portugal and is currently housed in the Museu Geológico in Lisbon (M. M. Ramalho, pers. comm. 2010), where it is now referred to multiple catalogue numbers (MG 26002 to MG 26045, and MG 26110 to MG 26115), even though it belongs to a single individual. The skull is divided into two main sections, and it is also referred to multiple numbers that correspond to individual fragments (MG 26014 and MG 26019, skull table, braincase, and orbits; MG 26015, MG 26016, and MG 26017, rostrum). This taxon is relevant for the phylogenetic analysis not only for its completeness, but also because it is the oldest known species of Goniopholis in the stratigraphical record of Europe. The position of G. baryglyphaeus in the genus is well supported, as the phylogenetic analysis placed this species as the sister taxon of G. simus (Lauprasert et al., 2007).

Three unnamed goniopholidid taxa (Fig. 3) come from the Lower Cretaceous of England, and were described in early works on goniopholidids (Hulke, 1878; Dollo, 1883; Hooley, 1907). Regardless of their general resemblance to Goniopholis and previous referral to this genus, their independent specific status has been recognized in the literature (see Salisbury, 1998, 2002; Salisbury et al., 1999; Andrade & Hornung, 2011). As they remain unnamed taxa, for practical purposes they are herein, respectively, referred to as Hulke's, Dollo's and Hooley's goniopholidids. Hooley's specimen (BMNH R3876) comprises a poorly preserved but fairly complete skull, and associated postcranial material, from Berriasian sediments of Atherfield (Isle of Wight). It was initially identified and described as G. crassidens by Hooley (1907), but this was based on the similarities with a skull roof erroneously assigned to the type of G. crassidens. As the correspondence of this skull roof to the type of G. crassidens cannot actually be confirmed (see Salisbury, 2002), no single diagnostic information can be used to assign BMNH R3876 to this taxon. However, as pointed out by Hooley (1907), there are evident differences between BMNH R3876 and G. simus. Dollo's specimen (IRSNB R47), is perhaps the most complete goniopholidid from Europe, consisting of skull, mandibles, and most of its postcranium, including limbs, dermal armour, and distal elements of the tail. It was originally assigned to G. simus (see Dollo, 1883), but actually has great similarity with Hooley's specimen (see Salisbury et al., 1999; Salisbury, 2002). The specimen was recovered from very distinct sediments at Bernissart (Sainte-Barbe Clays Formation), and is mid-Barremian to early Aptian in age (Yans et al., 2006). Hulke's specimen (= ‘Mr Willett's specimen’ in Salisbury, 1998) is a well-preserved skull without mandibles (BMNHB 001876). It was preliminarily described by Hulke (1878) as an additional specimen of G. simus, but actually shows obvious differences from the latter, even in the general morphology (e.g. longer, narrower rostrum, proportionally larger supratemporal fenestrae). Unfortunately, the specimen lacks specific geological information, but other crocodylomorph material from Mr Willett's collection comes from the PLG or the Wealden Group, and we consider that BMNHB 001876 most likely shares the same origin (probably Berriasian−Aptian). Previous to this study, only Hooley's and Hulke's specimens were included in the analysis of Salisbury et al. (2006), but as a single terminal also combining data from other Goniopholis, and therefore providing no clue to the relationships of these forms.

The European fossil record also includes other species of Goniopholis that are not considered in detail here. This is the case for Goniopholis tenuidens Owen, 1879, Goniopholis pugnaxKoken, 1887, and Goniopholis minorKoken, 1887, which were previously recognized to be based on non-diagnostic material, or even nomina dubia (for a detailed review, see Salisbury et al., 1999; Salisbury, 2002).

The North American record of ‘Goniopholis’ is extensive, particularly from the Morrison Formation (Kimmeridgian−Tithonian, Upper Jurassic). Several specimens have been described in cursory detail, and may represent synonymous taxa. Of these, two are regarded here as possibly valid taxa. Amphicotylus lucasiiCope, 1878, was soon moved to Goniopholis by Cope (1888), but had its original name re-instated by Mook (1942). The key specimen representing this taxon is AMNH 5782, from Colorado, a mostly complete and well-preserved skull. The second taxon is G. stovalliMook 1964, which was recovered from different outcrops in Oklahoma. According to Allen (2007), G. stovalli is more related to other North American taxa (e.g. Eutretauranosuchus, Calsoyasuchus) than to any European taxa. Goniopholis affinisLull, 1911 is a very incomplete specimen (MGS 8175; teeth, osteoderms) from the Aptian (Lower Cretaceous) Arundel Clay (Potomac Group; Maryland). These North American specimens referred to Goniopholis require a detailed direct revaluation and description, which is out of the scope of the present work. Few of these definitions remain, as their relationships to the British species have been disputed (e.g. Schwarz, 2002; Lucas & Sullivan, 2003; Allen, 2007). From the North American goniopholidids above, only Amphicotylus is included in our analysis.

Other goniopholidid taxa

Siamosuchus phuphokensis Lauprasert et al. 2007 comes from the Sao Khua Formation (Khorat Group) at Phu Phok (Thailand), currently understood as a pre-Aptian Cretaceous unit. It is based on a reasonably well-preserved skull and fragmentary postcranial remains found in association, all assigned to a single individual (PPC-1). Siamosuchus clearly diverges from other Goniopholis by the presence of a sagittal crest on the dorsal surface of the frontal, a feature that is absent in the latter but is characteristic of Sunosuchus.

Eutretauranosuchus delfsi Mook, 1967 comes from the Morrison formation at Cañon City (Colorado). Although the single specimen CMNH 8028 is a well-preserved and fairly complete set of skull and mandibles, it has never been described in detail, and several aspects of its anatomy are unclear, preventing the scoring of new characters and the assessment of previously scored ones. Nevertheless, the taxon has been included in several phylogenetic analyses (e.g. Turner & Buckley, 2008).

The genus Sunosuchus is here represented by three taxa, although more material has been described (see Fu et al., 2005; Schellhorn et al., 2009). The type species of the genus, Sunosuchus miaoiYoung, 1948 is from the Hokou Series at Kansu (China). It is the posterior half of a skull, with mandibles in occlusion. Sunosuchus thailandicusBuffetaut & Ingavat, 1980 comes from the Phu Kradung Formation (Khorat Group) of Thailand, and its material is limited to a fragmentary and poorly preserved (right) hemimandible, lacking the symphysis. The age of both is regarded as Late Jurassic to Early Cretaceous (Buffetaut & Suteethorn, 2007; Schellhorn et al., 2009). The best known taxon assigned to this genus is Sunosuchus junggarensisWu et al., 1996, represented by several well-preserved specimens from different ontogenetic stages, including skull, mandible, and postcranial elements. It comes from the Shishugou Formation (Junggar Basin) at Pingfengshan (Xinjiang, China), which are most likely to be Oxfordian (Upper Jurassic) in age. Although Sunosuchus is often included in phylogenetic analyses (e.g. Turner & Buckley, 2008), the different species have never been evaluated as separate terminals. This taxon comprises a wide variety of morphologies (compare Young, 1948; Wu et al., 1996; Fu et al., 2005), and is a potential basket-genus. Indeed, Lauprasert et al. (2007) did not find support for the monophyly of the genus. Therefore, terminals are treated separately.

Calsoyasuchus valliceps Tykoski et al. 2002 is a fragmentary but otherwise well-preserved skull from the Kayenta Formation of Arizona (USA). The material has been thoroughly described and information available includes tomographic data (Tykoski et al., 2002). This species is currently the oldest record of Goniopholididae (Sinemurian–Pliensbachian, Lower Jurassic), and has several peculiarities relative to other goniopholidid taxa (e.g. narial opening facing anterodorsally).

Vectisuchus leptognathus Buffetaut & Hutt, 1980 comes from the Vectis Formation (Wealden Group) of the Isle of Wight (England), late Early Cretaceous (Barremian–?early Aptian) in age. The single specimen (SMNS 50984) is a poorly preserved, but mostly complete, set of skull and mandibles, also including cervical vertebrae, dorsal/ventral armour, and forelimbs. Buffetaut & Hutt (1980) mention associated posterior postcranial elements (pelvic girdle, hind limbs, tail), but these were actually lost to a cliff collapse prior to collection. These authors classify Vectisuchus as Goniopholididae, which is supported by the phylogenetic analysis of Jouve (2009). However, the problem is still a matter of debate, as Vectisuchus appears to be more related to pholidosaurids (sensu lato) in other phylogenetic analyses (Jouve et al., 2006; Young & Andrade, 2009).

Goniopholishartti (Marsh, 1869) is a case of neglected taxonomic history. The specimen (symphyseal section of a mandible and a few postcranial remains) originates from the Aptian−Albian (Lower Cretaceous) Ilhas Formation (Recôncavo Basin). Erected as Crocodilus hartii, the species was moved to Goniopholis by Mawson & Woodward (1907) based on the existence of an anterior process (‘peg’) on the lateral end of a wide osteoderm (a character that is now known to be found also in clades other than Goniopholididae). The overlooked twist in the confusing history of crocodylomorph (and goniopholidid) taxonomy is that, later on, Buffetaut & Taquet (1979) recognized that G. hartii truly belongs to the genus SarcosuchusBroin & Taquet 1966, a genus often referred to Pholidosauridae (sensu lato). In fact, in the absence of convincing autapomorphic features between the type materials of both species, it is possible that Sarcosuchus hartti is the senior synonym of Sarcosuchus imperatorBroin & Taquet, 1966 (as in the analysis by Turner & Buckley, 2008). Nonetheless, the specimen is included in the analysis as a separate terminal to address formally the amendment proposed by Buffetaut & Taquet (1979).

Poorly known goniopholidid taxa used in additional exploratory runs

Denazinosuchus and ‘G.phuwiangensis constitute two poorly known taxa. The limited information available on the specimens only allows a preliminary scoring of characters, which limits their use in phylogenetic analyses and the evaluation on the position of these taxa. Each was added to a separate additional exploratory run.

Denazinosuchus kirtlandicus (Wiman, 1932) is a remarkably important taxon from North America, as it may be the earliest trustworthy record of Goniopholididae. The single specimen (PMU R232; a poorly preserved skull, without mandibles) comes from the De-na-zin Member, upper Campanian (Upper Cretaceous) of the Kirtland Formation (San Juan Basin; New Mexico). Denazinosuchus was originally ascribed to the genus Goniopholis (Wiman, 1932; see also Wolberg, 1980; Mateer, 1981), but morphological differences made evident by Lucas & Sullivan (2003) supported its placement in a new genus. As Denazinosuchus shows distinctive characteristics that easily differentiate this taxon from other European and North American goniopholidids (e.g. absence of maxillary depressions), Lucas & Sullivan (2003) preferred to consider Denazinosuchus as a ‘Mesosuchia incertae sedis’, rather than a Goniopholididae. However, the relationships of this taxon have never been tested through phylogenetic analysis. Unfortunately, previous studies commented mostly on general aspects and emphasized unique characteristics of the taxon, rather than common aspects with other goniopholidids.

‘Goniopholis’ phuwiangensis comes from the younger Sao Khua Formation (Khorat Group) of Thailand (Buffetaut & Ingavat, 1983). The single specimen (TF 1478) is thoroughly described, but limited to an incomplete anterior left dentary at the symphyseal region. Therefore, its assignment to Goniopholis is based on general features (e.g. alveoli 3–4 proportionally larger than surrounding alveoli, and located next to each other; proportionally short symphysis). As ‘G.phuwiangensis shares the same provenance and stratigraphical level as Si. phuphokensis, and there is no overlap between specimens, it is possible that both truly belong to a single species. This hypothesis has never been tested prior to the present analysis.

Fragmentary material not included in the analysis

Although taxa included in the analysis comprise fragmentary forms, most allow scoring of an appropriate number of characters that allow a preliminary identification. This is the case for Su. thailandicus, Sa. hartti, and even ‘G.’ phuwiangensis, as well as several other taxa not closely related to goniopholidids.

However, several reports exist on fragmentary material that is attributed to Goniopholis. Nondiagnostic remains were previously removed from the genus (Benton & Spencer, 1995; Salisbury et al., 1999; Salisbury, 2002). Nonetheless, it is important to review briefly the case of Goniopholis paulistanusRoxo 1936, which is actually described as a separate species. The material comes from a railroad cut between the cities of Jupiá and Araçatuba (São Paulo State, Brazil). These sediments are now known to be from the Upper Cretaceous Bauru Group, an intracratonic unit that includes a wide range of crocodylomorph reports, mostly notosuchians and is interpreted as a semi-arid palaeoenvironment (see Bertini et al., 1993; Kellner & Campos, 1999; Andrade & Bertini, 2008c).

The presence of Goniopholididae at São Paulo State was previously mentioned by von Ihering (1911; apudRoxo, 1936), who reported fragmentary material from the city of São José do Rio Preto (also Bauru Group), and attributed the teeth to Goniopholis or Machimosaurus. The specimens used by Roxo (1936) are as fragmentary as those mentioned by von Ihering (1911; apudRoxo, 1936), but limited to two tooth crowns and a fragment of a (right) tibia (SGM N2948, SGM N2949, SGM N2950). Roxo (1936) also mentions other teeth and fragmentary caudal vertebrae, but provides no origin, deposit number, or illustration of the specimens. Therefore, G. paulistanus has been understandably neglected by the key modern references on the genus. Indeed, the tooth material illustrated by Roxo (1936) resembles Goniopholis in the overall robustness and enamel ornamentation, but it is equally similar to other neosuchian taxa (e.g. Pholidosaurus, Machimosaurus), offering no diagnostic information (the same applies to the fragment of tibia). Based on the lack of evidence supporting this taxon, Price (1950) considered it impossible to place G. paulistanus either in Goniopholis or Machimosaurus, and the taxon is not included in a broad review of mesoeucrocodylian taxa in southern South America (Candeiro & Martinelli, 2006). Indeed, lack of diagnostic information prevents specific association with any other material or known species/genus, and provides no relevant character. This taxon was therefore not included in the present phylogenetic analysis.

As the type material (two tooth crowns, one fragmentary tibia) is too fragmentary to provide sufficient diagnosis, we propose to consider G. paulistanusRoxo, 1936 a nomen dubium. The specimens (SGM N2948, SGM N2949, SGM N2950) are here regarded as Neosuchia incertae sedis, pending the discovery of further material that can be confidently related to these, and provide appropriate diagnostic characters.

Methods

The phylogenetic analysis was carried out with PAUP* 4.0b10 (Swofford, 2002), using heuristic search (tree bisection-reconnection branch swapping; 1000 replicates), with all characters equally weighted, but 24 of those treated as ordered (for detailed list, see Supporting Information File S1). The collapse option for zero length branches was applied, to avoid the possible grouping of clades unsupported by actual data. Branch decay (Bremer, 1994) was calculated with TreeRot v.2 (Sorenson & Franzosa, 2007), in combination with PAUP. Bootstrap (Felsenstein, 1985) and jackknife (50% deletion) were calculated in PAUP, with 1000 replicates obtained through fast stepwise addition. Unless otherwise stated, consistency and homoplasy indexes presented herein exclude uninformative characters.

Additional tests

Complementary analyses allowed the evaluation of particular problems, tested through three separate runs. The separate runs identify the isolated effects of the inclusion/exclusion of each terminal, and followed identical procedures to the core analysis (but used 200 replicates). As these are additional analyses, nodal support was not evaluated.

The first test verified the impact of Nannosuchus in the topology. As the taxon is included in the core analysis, this test simply demanded the exclusion of Nannosuchus, via a subsequent new run. The resulting topology was compared with the one obtained from the core analysis to evaluate the stability of the terminal and overall impact in the result. This was not considered an exploratory run, but rather a simple additional test, because BMNH 48217 was examined directly and information on the specimen is broadly available in the literature (Owen, 1879; Kalin, 1933; Joffe, 1967; Clark, 1986). Therefore, scoring of characters was not biased by poor information, as in the other cases.

The second and third tests were exploratory runs that evaluated the positions of poorly known taxa, respectively, Denazinosuchus and ‘G.’ phuwiangensis. In each case, we evaluated the resolution of the topology, the phylogenetic relationships, and their putative placement in Goniopholis.

A fourth exploratory test aimed to verify the possibility that ‘G.’ phuwiangensis is congeneric with Si. phuphokensis. The analysis used a special alternative terminal that combined the data of both species, creating a composite terminal (‘ALTSiamosuchus’). The result of this run was compared to the result of the third test, to evaluate whether the placement of ‘G.’phuwiangensis in the genus Siamosuchus is appropriate (i.e. as parsimonious as to the result of the third test), or not. We stress that this terminal is to be considered as a chimera, built for the sole purpose of testing how congruent is the association of these taxa in a single genus and, therefore, should not be used for any other purpose. Until discovery of appropriate supporting data, these species are better seen as separate taxa.

Results

General results from core analysis

The analysis resulted in two equally parsimonious trees [length = 2186; consistency index (CI) = 0.2903; homoplasy index (HI) = 0.7097; retention index (RI) = 0.7626; rescaled consistency index (RC) = 0.2254], diverging only in the position of the notosuchian Araripesuchus buitreraensisPol & Apesteguia, 2005, and summarized by a strict consensus (Fig. 12). A full report of results is available in Supporting Information File S1 (Sp1-2.1, strict consensus and nodal support; Sp1-2.2, apomorphy list). Log files for all analyses are available in Supporting Information File S2.

Figure 12.

Phylogenetic analysis of Mesoeucrocodylia, with emphasis on Goniopholididae. The topology is the strict consensus of two trees, and major clades are summarized by thick lines (proportional to the number of specimens included in the analysis). The family Goniopholididae is indicated by the light grey box, whereas the genus Goniopholis is indicated by a dark grey box. Note that ‘pholidosaurids’ in their traditional sense are paraphyletic (dotted box), but Sarcosuchus, Elosuchus, and Vectisuchus constitute a monophyletic group (i.e. Elosuchidae). Nodal support is given for each node (Bremer decay/bootstrap/jackknife), but bootstrap and jackknife are only shown for values above 50%.

The resulting topology is overall consistent with previously published results (e.g. Clark, 1994; Ortega et al., 2000; Salisbury et al., 2006; Delfino et al., 2008; Turner & Buckley, 2008), with several aspects in common: (1) Thalattosuchia assumes a derived position within Metasuchia (as in Clark, 1994; Ortega et al., 2000; contraAndrade & Bertini, 2008a, b; Young & Andrade, 2009); (2) a monophyletic Sebecosuchia is not supported (as in Clark, 1994; contraOrtega et al., 2000), although the clade Sebecia is (as in Larsson & Sues, 2007); (3) both Notosuchia (sensuGasparini, 1971) and Ziphosuchia (sensuOrtega et al., 2000) are also supported.

However, a few peculiarities do exist: sebecians and a clade of ‘mahajangasuchids’ (Anatosuchus +Kaprosuchus + Mahajangasuchus) are successive sister groups of Notosuchia, contrary to the prevalent view that these groups are more closely related to (or part of) Neosuchia (e.g. Turner & Buckley, 2008; Sereno & Larsson, 2009). Susisuchus appears as the sister group of Isisfordia (contraSalisbury et al., 2006), a relation particularly well supported within Eusuchia (Bremer = 3; bootstrap/jackknife > 50%), and both are herein considered as Susisuchidae. Finally, all goniopholidids, pholidosaurids, dyrosaurids, and thalattosuchians form an exclusive monophyletic lineage within Neosuchia, and not including Eusuchia (only supported by Jouve et al., 2006).

Overall, nodal support can be considered weak: branches with both bootstrap/jackknife values of at least 50% comprise 51.6% of nodes, and only 15.8% of the nodes achieve indexes of 90% or better. Bremer decay shows slightly better robustness, as 66.3% of the nodes have at least a decay index of 2, and 24.2% have a nodal value of 5 or better. Nevertheless, several traditional clades have at least a reasonable Bremer decay (e.g. ≥ 2, Eusuchia, Ziphosuchia, Mesoeucrocodylia). Low support and indexes are somewhat expected in this result because of the extent of the matrix. As the numbers of taxa increase, the chance that reversions/parallelisms are found for each character will also increase, with impact on the overall support.

Phylogenetic relationships of Goniopholis kiplingi and other goniopholidids

Goniopholis kiplingi DORCM 12154 has well-established relationships, based on the analysis herein, where the new species appears as the sister taxon of G. simus. Together they form a well-supported node (Bremer = 3; bootstrap/jackknife > 80%), which is characterized by seven transformations, three of them unambiguous (indicated by ‘*’): heart-shaped narial opening (32-1); diamond-shaped naso-oral fossa (69-3); supratemporal fenestra subequal to the orbit (106-1); frontoparietal suture arched, with ‘concavity’ facing posteriorly (120-2*); postorbital with anterior and lateral edges, separated by a distinct anterolaterally facing edge (123-1); presence of a triangular tuberous projection on the medial line of frontal (139-1*); choanal opening wedges between bony lamina as an acute V-shape, internal nares assuming a lanceolate profile (238-1*). The morphology of the narial opening and naso-oral fossa were previously observed by Salisbury et al. (1999) in G. simus. These unfortunately are not preserved in the type of G. baryglyphaeus, and further comparison must await description of new specimens. However, distinction is immediate as the tuberous frontal projection has a triangular profile, shared only by G. kiplingi and G. simus.

Goniopholis baryglyphaeus Schwarz, 2002 (Fig. 5) is the immediate sister-taxon of (G. simus + G. kiplingi), a node that is also reasonably well supported (Bremer = 2; bootstrap/jackknife > 70%). It clearly differs from G. kiplingi and G. simus by the presence of a different type of frontal tuberosity (not limited medially, less conspicuous dorsally), a well-defined postnarial fossa, heavier ornamentation, proportionally smaller supratemporal fenestrae, and an almost straight frontoparietal suture, to cite but a few. These three species (G. baryglyphaeus, G. kiplingi, and G. simus) comprise a group that can be defined as Goniopholis sensu stricto, and is characterized by eight apomorphies (three unambiguous): anterior border of maxillary depressions deep, well defined (90-1*); square-shaped to subrectangular supratemporal fossae (111-1*); anterior border of the orbit composed by lachrymal and jugal (160-0); anterior jugal ramus with multiple (two to five) small foramina, ventrally orientated, near the contact with the maxilla (180-2*); anterior face of palatine process at palate is truncated, with suture transversely orientated (225-1); external mandibular fenestra is absent (312-0); symphyseal alveoli 1–2 are confluent, separated by a thin alveolar wall, and clearly apart from neighbouring alveoli (402-1); paramedial osteoderms not keeled (478-0).

The clade including Dollo's and Hooley's specimens is the best supported node within Goniopholididae (Bremer = 9; bootstrap/jackknife > 95%). This node has ten unambiguous apomorphies, and further, two ambiguous changes: circular supratemporal fossae (111-3*); prefrontals are very long, reaching the posteromedial borders of the orbit (126-2*); robust anterolateral postorbital process present and very long (152-2*), almost reaching the dorsal edge of the anterior jugal ramus, shielding the posterolateral section of the orbit (153-1*); orbits have a strong dorsal component (157-2*); jugal only forms a narrow band of bone below the orbit (175-0*); palpebrals are robust, but small (187-2*), and rectangular or very elongated (188-2*); postorbital bars are vertical in anterior view (195-0*); the nasopharyngeal duct has a deep sulcation on its ventral surface, at the medial contact of palatine (230-1*); symphyseal alveoli 1–4 are transversely aligned, so the fourth alveolus is lateral (or lateral and slightly posterior) to first alveolus, and following alveoli are posterior to them (398-2), and the third alveolus is medial to the fourth one (405-0). The analysis herein corroborates Salisbury (2002) in considering Hooley's and Dollo's specimens as very similar, and their distinction from other goniopholidids has been recognized (Hooley, 1907; Salisbury et al., 1999). Their common node with Hulke's specimen also has strong support (Bremer = 5; bootstrap/jackknife > 60%) and is characterized by ten transformations (six unambiguous): loss of the prefrontal−lachrymal crest, dorsal to orbit (99-0*), and of the transverse frontal crest (101-0*); the supratemporal fenestra is subequal to the orbit (106-1), whereas the supratemporal fossa is larger than the orbit (107-2); lateral processes of frontal arched laterodorsally, with palpebral and postorbital curved dorsally (115-1*); posterior ramus of jugal becomes subcircular to subpolygonal, in cross-section (173-0*); ventral margin of anterior jugal ramus level with the ventral margin of posterior ramus (178-0*); the postorbital bar is vertical also in lateral view (194-0); the palatal ramus of maxilla takes part in the anteromedial border of the suborbital fenestra (215-1); the anterior palatine process of palate is evidently longer than wider (223-1*). The solid placement of Hooley's, Dollo's, and Hulke's specimens in the phylogenetic analysis not only clarifies their relationships, but also demonstrates that they cannot be lumped into previously established names such as G. simus because they actually represent a diverging radiation of goniopholidids.

All the taxa above form a clade of European goniopholidids, alongside Nannosuchus, which is its basal-most branch. This node, however, lacks strong support (Bremer = 1; bootstrap/jackknife < 50%) and is defined by only seven apomorphies (only two unambiguous): presence of a pair of anterior narial notches (34-1); distal margin of frontal is posterior to the postorbital bars, clearly reaching into the intertemporal bar and mid skull roof (122-2*); prefrontals are long, composing the anteromedial and medial borders of the orbit (126-1), which becomes even longer in Hooley's and Dollo's specimens; the anterior process of frontal is narrow throughout (136-2*), and at least reaches the anterior-most tip of prefrontals (142-1); spina quadratojugalis present (167-1); and palpebrals are robust and large (187-1). Some of these characters may be biased by preservation, or their distribution may be affected by missing data, as in the case of the quadratojugal spine (which can easily be damaged or worn, hardly leaving a trace and leading to an apparent absence). However, the frontoparietal suture is clearly located in a posterior region relative to other more basal taxa (e.g. Amphicotylus, Sunosuchus) and provides an easily recognizable parameter for the identification of this clade.

Amphicotylus is clearly set outside the clade containing all European Goniopholis, which supports the taxonomic decision by Mook (1967). Siamosuchus is the basal-most member of this lineage of goniopholidids, which can be easily recognized by the ‘broad-snouted’ morphology, where festooning is evident and maxillae project in waves, both ventrally and laterally. Siamosuchus is also the only one to share a sagittal crest (102-1) with Sunosuchus, granting this taxon a more basal position within the node. The ‘broad-snouted’ clade is well supported (Bremer = 3.2; bootstrap/jackknife > 50%) and has 11 apomorphies (nine unambiguous): perinarial crests present (29-1*); the notch at the premaxillae−maxillae suture closely encase dentary tooth (58-1*); maxilla festooned, with two major waves clearly identifiable, separated by an evident concave area, and ventral maxillary margin strongly sinusoidal (82-1*); prefrontal−lachrymal crest present (99-1*); frontal is wide, relative to skull roof, roughly corresponding to 40–50% of the width of the skull table (134-1*); presence of an evident boss on lateral edge of paroccipital process (277-1); teeth robust, ‘inflated’, or bulbous apically (373-1*); maxillary and mandibular dentition set in-line or interlocked (384-0); last premaxillary tooth evidently anterolateral to first maxillary tooth (391-2*); dentition composed by acute caniniforms anteriorly, followed by blunter caniniform teeth (393-1*); maxillary/dentary teeth forming two waves of larger crowns/alveoli (397-2*).

The sister group of this ‘broad-snouted’ group is opposed by a ‘narrow-snouted’ group, containing all Sunosuchus and also Eutretauranosuchus, but not Calsoyasuchus. This lineage has a very low support (Bremer decay = 1; bootstrap/jackknife < 50%), and is secured by only four transformations (two unambiguous): main axis of supratemporal fossae parallel (110-1); maxillo-palatine fenestrae present and anteroposteriorly elongated (205-2); external mandibular fenestra small, approximately the same length as the orbit (313-1*); symphysis moderately elongated (332-1*). Within this node, Eutretauranosuchus appears as sister group of Su. junggarensis, a relationship that is only poorly supported (Bremer decay = 1; bootstrap/jackknife < 50%). In contrast, the clade composed of Su. miaoi and Su. thailandicus has better nodal support (Bremer decay = 3.25; but bootstrap/jackknife < 50%), and is defined by narrow teardrop-like mandibular fenestrae (315-4*), with anterodorsal and anteroventral margins meeting at an acute angle anteriorly (316-1*), and a long surangular (346-1). As Eutretauranosuchus nests within Sunosuchus, this last genus is recovered as paraphyletic.

Calsoyasuchus valiceps was previously considered as the sister group of Eutretauranosuchus (see Tykoski et al., 2002), but here it figures as the basal-most taxon within Goniopholididae. Calsoyasuchus clearly differentiates from the other groups by the combination of an elongated and somewhat narrow rostrum, shared with Sunosuchus/Eutretauranosuchus, with incipient lateral festooning of the maxillae, shared with the ‘broad-snouted’ clade (e.g. Siamosuchus, Goniopholis). It is also the only goniopholidid to have a truly steep narial opening, facing anterodorsally (31-1). Salisbury et al. (1999) considered that Goniopholis also has an anterodorsally orientated naris, but we interpret the structure as dorsally orientated (31-2), with a minor inclination induced by perinarial crests. Conversely, it is only in Calsoyasuchus that the vestibulum is visible in anterior view. In the present analysis, this is an important character that is present in the basal-most Neosuchia (Atoposauridae), is retained in Stolokrosuchus, Calsoyasuchus, teleosaurids, and Pelagosaurus, and possibly paralleled by Sarcosuchus. This character is certainly relevant in determining a basal position of Calsoyasuchus relative to the other goniopholidid lineages. Furthermore, Calsoyasuchus shares with Stolokrosuchus, teleosaurids, and Pelagosaurus the retention of the antorbital fenestra and its elongated fossa. This important element, translated in this analysis as a complex set of characters (43–52), is only present in basal lineages of Neosuchia.

Overall, the family Goniopholididae is a clade reasonably well supported (Bremer decay = 3; but bootstrap/jackknife < 50%). As a general rule, goniopholidids are easily recognized by the presence of lateral maxillary fossa/fossae next to the alveolar margin (86-1*), which are combined into a single pair of maxillary depressions, and located next to the maxilla−jugal suture (89-1*). Furthermore, five other characters (three unambiguous) also characterize the node: ventral-most neurovascular foramina at the posterior maxilla are located high on it lateral surface, distant from the alveoli (up to twice the distance from other foramina) (26-1*); maxillae have a distinct smooth ventral surface alongside the alveolar margin (79-1); the distal margin of frontal (frontoparietal suture) is medial to the (dorsal end of) postorbital bars, or slightly anterior to them (122-0*); presence of a vascular opening at the lateral edge of the postorbital bar (202-1); anterior process of palatines short, with length subequal to width (223-0*). It should be noticed that the presence of maxillary depressions actually constitutes a complex set of morphological features that also involves the distribution of neurovascular foramina in the maxillae and ornamentation (see Andrade, 2009).

Vectisuchus constitutes a controversial case in the systematics and taxonomy of neosuchians, as different works disagree in its placement, either as a goniopholidid (Jouve, 2009) or a clade more related to dyrosaurids and Pholidosaurus (Jouve et al., 2006; Young & Andrade, 2009). The analysis of Lauprasert et al. (2007) also shows Vectisuchus as a goniopholidid, but includes no tethysuchian and therefore provides no clues on the matter. In the analysis presented herein, Vectisuchus clusters with Elosuchus, rather than with goniopholidids, as previously indicated by Young & Andrade (2009). This node is reasonably well supported (Bremer decay = 2; although bootstrap/jackknife > 60%), and can be identified by eight apomorphies, five of them unambiguous: tooth row unaligned and ventrally displaced relative to occipital condyle (03-0*); rostrum proportions (height and width) subequal (6-1); lachrymal fossa present and small (53-1*); supratemporal fossae subtriangular, with axis converging medially (111-2); orbits strongly facing anterodorsally (158-1*); anterior border of the choanae near the end of the nasopharyngeal duct (235-2); anterolateral rami of pterygoid embrace most of the choanae, but do not meet medially anterior to it (242-2*); external mandibular fenestrae are large, evidently longer than the orbit (313-2*). Indeed, Vectisuchus has characters more typical of tethysuchians (Pholidosaurus, Elosuchus, Dyrosaurus) than goniopholidids, such as premaxillae that project laterally and are wider than maxillae (64-1), the proportionally large supratemporal fossae (106-2), and the temporal bars oblique and anteriorly convergent (148-1). The postorbital bars are verticalized in lateral view (194-0), although some of the European goniopholidids share this feature. Vectisuchus also has a rostrum that is narrower than even Su. junggarensis or Eutretauranosuchus, and more similar to the condition seen in most tethysuchians and thalattosuchians (although the rostrum of Su. miaoi and Su. thailandicus is not preserved). The absence of the maxillary depressions count against considering Vectisuchus as a goniopholidid, but this evidence alone is not definitive because other goniopholidids are known to lack the structure (e.g. Nannosuchus). Above all, the restroarticular process of Vectisuchus may be the best evidence of its relationship with tethysuchians, or at least of its distant relationship with goniopholidids. Goniopholidids typically have a short retroarticular process that is directed posteriorly, with a small dorsal component, and its surface for attachment of the m. depressor mandibulae faces posterodorsally. However, in Vectisuchus this process is long, posterodorsally orientated, and the surface for the m. depressor mandibulae faces dorsally, as in typical Tethysuchia, or in Thalattosuchia. At present, the evidence cannot support Vectisuchus as a goniopholidid, but rather as a tethysuchian. Furthermore, other characters (e.g. the subtriangular supratemporal fossae, the orientation of the orbits, and the morphology of frontal) provide a range of features that can be easily recognized, which are (in combination) exclusively shared by Vectisuchus and Elosuchus. Therefore, there is seemingly enough evidence to consider Vectisuchus as a sister clade to Elosuchus, and thus to place this taxon in the Family Elosuchidae.

Relationships of Goniopholididae and closely related groups

The family Goniopholididae is closely related to a group containing all Pholidosauridae, Dyrosauridae, and Thalattosuchia. The last two clades are recovered as monophyletic. However, Pholidosauridae is paraphyletic in its traditional sense (as, for example, in Buffetaut, 1982; a broad group, encompassing Sarcosuchus and Elosuchus). In the phylogeny herein, Pholidosaurus is more closely related to Dyrosauridae than to other ‘pholidosaurid’ taxa, which constitute a more basal, monophyletic group. In this sense, a more exclusive version of the family is favoured, and Pholidosauridae is used here only for the genus Pholidosaurus. The close relation of Pholidosaurus and dyrosaurids is weakly supported (Bremer decay = 1), characterized by eight transformations, four of which are unambiguous, and presented here: spina quadratojugalis present (167-1); jugal with multiple (two to five) small foramina, ventrally orientated (180-2); second symphyseal alveolus not in line with alveoli 3–4, and at a more lateral position (407-2); presacral vertebrae with fully projecting hypapophysis (421-1). Most characters occur in several other taxa (e.g. spina quadratojugalis, jugal foramina, presacral hypapophysis), and their occurrence at this node may simply reflect missing data in other taxa.

The group encompassing all ‘pholidosaurids’ and dyrosaurids corresponds to Tethysuchia Buffetaut 1982, which is used herein. This node is reasonably well supported (Bremer decay = 3; although bootstrap/jackknife < 50%), and is established by ten unambiguous transformations (13 in total; see Appendix and Supporting information S2): loss of antorbital cavity (43-0) and internal antorbital fenestra (44-0); notch at premaxillae−maxillae suture closely encase dentary tooth in occlusion (c. 50–60%) (57-1); premaxillae project laterally and are wider than maxillae, in dorsal view (64-1); maxillae are excluded from the naso-oral fossa (68-0); anterior process of frontal reach or surpass the anterior tip of prefrontals (142-1); temporal bars oblique and anteriorly convergent (148-1); presence of the anterolateral process of postorbital, short and robust (152-1); postorbital bar verticalized in lateral view (194-0); insertion area for m. pterygoideus posterior extends onto the lateral surface of angular (353-1). It must be noted that the analysis herein does not include Terminonaris robusta Mook, 1934, an upper Cretaceous (Turonian) taxon from North America that seems to be intimately related with Sarcosuchus (Wu et al., 2001; Sereno et al., 2001, 2003). Future inclusion of this taxon may change relationships and nodal support among tethysuchians.

Sarcosuchus appears as the sister clade of ‘(Elosuchus + Vectisuchus)’. Sarcosuchus hartti consistently and strongly relates to Sa. imperator (Bremer decay = 6; bootstrap/jackknife ≥ 60%), as in Buffetaut & Taquet (1979). The genus parallels goniopholidid taxa in different aspects: (1) the narial opening has a steep anterodorsal orientation, shared with Calsoyasuchus; (2) the teeth at the symphysis are arranged in a broad arch, as seen in Hooley's and Dollo's specimens; (3) the long postorbital process shields the orbit in lateral aspect, as in Hooley's and Dollo's specimens (also shared, to a certain extent, by Elosuchus and Vectisuchus). However, Hooley's and Dollo's specimens have a long thin process, rather than the laminar projection seen in Sarcosuchus, and the protection offered to the orbit can be easily distinguished.

The ‘other pholidosaurid monophyletic group’ therefore encompasses both species of Sarcosuchus, Vectisuchus, and Elosuchus, and a modified version of the family name Elosuchidae Broin, 2002 is used herein to accommodate these taxa. The family Elosuchidae was created to include Elosuchus cherifensis and Stolokrosuchus lapparenti, an arrangement that is not adequate for the present phylogeny, resulting in many unnecessary taxonomic amendments. Furthermore, the phylogenetic placement of Stolokrosuchus is still disputed (e.g. compare Larsson & Sues, 2007; Pol et al., 2009), and is uncertain amongst the Mesoeucrocodylia. Therefore, a reduced version of the family Elosuchidae is adopted here to represent this monophyletic basal clade of tethysuchians. The family Elosuchidae (sensu this paper) has very poor support (Bremer decay = 1; bootstrap/jackknife < 50%). However, it is indicated by five unambiguous transformations (11 in total; see Appendix and Supporting information S2): lateral processes of frontal arched laterodorsally, with palpebral and postorbital curved dorsally (115-1), a characteristic also seen in the undescribed goniopholidids by Hulke (1878), Dollo (1883), and Hooley (1907); dorsal surface of frontal is concave (135-1); surangular is forked (348-1); the last premaxillary tooth is evidently anterolateral to first maxillary tooth (391-2); dentary tooth opposite to premaxilla−maxilla contact is subequal to other neighbouring teeth (408-0).

The group containing all goniopholidids, pholidosaurids, dyrosaurids, and thalattosuchians has low branch support (Bremer decay = 1.83; bootstrap/jackknife < 50%), but is associated with 16 transformations. Five of those are unambiguous changes: internarial bar absent and external nares confluent (character 37-0); jugal takes part in the external antorbital fenestra (50-1); premaxillae are paddle-shaped in dorsal view and expanded laterally (65-3); frontal is ±33% (or more) of skull width (133-1); postorbital is ±50% (or more) of the upper temporal bar (151-1);

Similarities that occur both in Pholidosaurus and Goniopholis include characters that are present in more basal neosuchian taxa. These include: the overall shape of the frontal; the particular way the palpebral attaches to the orbit (Andrade & Hornung, 2011); the short paraoccipital process, with a robust thickened lateral edge (‘boss’); the short and robust anterolateral process of the postorbital; imbricated dermal armour with paired paravertebral rows; similar dentition. Most of these features are poorly sampled within Mesoeucrocodylia, but often occur in other taxa. Tethysuchians, for instance, retain at least some of these traits (e.g. short paroccipital process, postorbital process). However, such characters underwent further transformation in the group (e.g. robust dentition in Sarcosuchus; accessory rows of osteoderms present in dyrosaurids).

Results from additional analysis

Relationships of Nannosuchus: is it a Goniopholis?: The first additional run focused on the phylogenetic relationships of Nannosuchus. After a single run (heuristic search, 200 replicates), only six topologies were recovered, with scores (length = 2174; CI = 0.2919; HI = 0.7081; RI = 0.7629; RC = 0.2267) similar to the core analysis. The different topologies vary in the position of Pholidosaurus and A. buitreraensis. A strict consensus (Fig. 13A; Supporting Information File S2) shows a tree that mirrors the topology from the core analysis, except for: (1) the position of Pholidosaurus, which collapses from its position as sister clade to dyrosaurids, and assumes a basal position in Tethysuchia, creating a polytomy; and (2) the absence of Nannosuchus itself.

Figure 13.

Miscellaneous problems in ‘goniopholidid’ evolution, summarized by strict consensus. In the first additional analysis (A), the exclusion of Nannosuchus (= ‘Goniopholisgracilidens) leads to a less resolved topology, affecting the putative relationships of tethysuchians, but not goniopholidids. In the second additional (exploratory) analysis (B), the relationships of Denazinosuchus (= ‘Goniopholis’ kirtlandicus) are poorly defined and its presence also induces collapse of the tethysuchian taxa; although Denazinosuchus nests well inside Goniopholididae, it never clusters with Goniopholis. The third additional analysis (C) shows that the relationships of ‘Goniopholisphuwiangensis are also poorly defined, but the taxon nests inside a SunosuchusEutretauranosuchus clade, rather than with any Goniopholis. In the fourth additional analysis (D), the putative assignment of ‘G.’phuwiangensis is tested for the genus Siamosuchus (composite terminal with Siamosuchus phuphokensis, both from Phu Phok Formation, Thailand), which results in a poorly resolved topology, showing that there is no support for this combination. Terminals added in each topology are in bold; nodes affected are indicated by a grey arrow. Data on Thalattosuchia include 14 terminals, and are represented by a thick line.

Interestingly, the inclusion/exclusion of Nannosuchus does not impact relationships amongst goniopholidids. When included in the analysis, Nannosuchus does not exhibit the ‘rogue’ behaviour typical of a ‘wildcard’ taxon. It rather nests as a basal node in a group formed exclusively of European goniopholidids, and not as sister clade of G. simus or any other species. The inclusion of Nannosuchus marginally impacts the indexes and results in a relatively higher tree length (12 steps). Such differences are not profound, and an increased tree length is already expected for a taxon that has such a peculiar morphology, at least partially biased by ontogeny. Finally, Nannosuchus improves the resolution of the consensus, reducing the number of topologies obtained. Results herein show that Nannosuchus is neither incongruent as an independent taxon, nor is it more closely related to any particular known goniopholidid.

Therefore, this analysis cannot find support for the use of ‘Goniopholisgracilidens, as proposed by Salisbury et al. (1999). It rather shows that, if conceived as an independent taxon, Nannosuchus will not cluster with or within other species of Goniopholis. It also shows that, although BMNH 48217 is indeed an immature specimen, there is no reason to consider Nannosuchus specifically as a young specimen of G. simus (or of any other species), as previously proposed by Joffe (1967). This means that, for the time being, the best taxonomic approach for BMNH 48217 is to refer to this material in its original form, as N. gracilidens.

Relationships of Denazinosuchus: The second additional run focused on the phylogenetic relationships of Denazinosuchus. After a single exploratory run (heuristic search, 200 replicates), 26 topologies were recovered, with scores (length = 2189; CI = 0.2902; HI = 0.7098; RI = 0.7628; RC = 0.2251) similar to the core analysis. The inclusion of this taxon produces a major impact on the ‘broad-snouted clade’ of goniopholidids, and on Tethysuchia. The strict consensus (Fig. 13B; Supporting Information File S2) mirrors the topology from the core analysis, but: (1) the clade containing the European goniopholidids, Denazinosuchus, and Amphicotylus is collapsed to the same node; and (2) collapse of most clades of nondyrosaurid tethysuchians. The inclusion of Denazinosuchus marginally impacts the indexes and results in a relatively higher tree length (three steps). Although these differences are not profound, Denazinosuchus displays limited ‘wildcard’ behaviour, flipping to different nodes inside the broad-snouted goniopholidid clade. Nonetheless, it seems clear that: (1) Denazinosuchus is part of this broad-snouted group; (2) it does not cluster with G. simus, G, kiplingi, or G. baryglyphaeus, (3) it is not particularly related to any of the unnamed European taxa.

Denazinosuchus then is a goniopholidid, more closely related to Goniopholis, Nannosuchus, and Amphicotylus than to Sunosuchus, Eutretauranosuchus or Siamosuchus. This analysis cannot find support for the use of Goniopholis kirtlandicus, as originally proposed by Wiman (1932). It rather shows that Denazinosuchus does not cluster with or within other species of Goniopholis. Other than that, the incompleteness of the specimen and poor understanding of the taxon does not allow a more precise statement about relationships. This means that, for the time being, the best taxonomic approach for PMU R232 is to abide by the proposal of Lucas & Sullivan (2003) to refer to the species as Denazinosuchus kirtlandicus.

Testing the relationships of ‘Goniopholis’ phuwiangensis amongst goniopholidids: The third additional analysis focused on the phylogenetic relationships of ‘G.phuwiangensis. After a single exploratory run (heuristic search, 200 replicates), ten topologies were recovered, with scores (length = 2186; CI = 0.2903; HI = 0.7097; RI = 0.7627; RC = 0.2254) almost identical to the core analysis. The different topologies vary in the position of ‘G.’phuwiangensis and A. buitreraensis. The strict consensus tree (Fig. 13C; Supporting Information File S2) mirrors the topology from the core analysis, except for the collapse of relationships within the Sunosuchus−Eutretauranosuchus clade, where ‘G.’phuwiangensis nests. The inclusion of ‘G.’phuwiangensis does not impact the final indexes, but increases the number of topologies obtained and results in a less resolved consensus. However, this is expected for such an incomplete taxon. These results indicate that ‘G.’phuwiangensis is more related to Sunosuchus or Eutretauranosuchus than to Goniopholis, and therefore its referral to this genus is in error. Unfortunately, many goniopholidid taxa do not preserve the symphysis (i.e. Siamosuchus, Su. miaoi, Su. thailandicus, Calsoyasuchus), and their relation with ‘G.’phuwiangensis cannot be fully tested.

Nonetheless, in order to evaluate the possibility that ‘G.phuwiangensis is at least congeneric with Si. phuphokensis, a fourth additional analysis was carried out, using a single terminal to combine the non-overlapping data on both taxa. After a single exploratory run (heuristic search, 200 replicates), 12 topologies were recovered, with scores (length = 2191; CI = 0.2897; HI = 0.7103; RI = 0.7619; RC = 0.2247) slightly worse than for the second exploratory run (separate terminals for ‘G.phuwiangensis and Si. phuphokensis). The different topologies vary in the position of: (1) the species of Sunosuchus and Eutretauranosuchus; (2) the nondyrosaurid tethysuchians; (3) A. buitreraensis. The strict consensus (Fig. 13D; Sp2) shows a tree that follows the topology from the core analysis, except for the collapse of the SunosuchusEutretauranosuchus clade (although its internal nodes remain unaffected), and the poorer resolution of relationships within Tethysuchia. The chimaeric Siamosuchus remains as the basal-most branch of the ‘broad-snouted’ clade of goniopholidids. It also has a feeble impact on the final indexes, when compared to the second supplementary run, and increases the final number of topologies obtained, resulting in a less resolved consensus. These results indicate that there is no support for the idea that ‘G.’phuwiangensis is more related to Siamosuchus than to Sunosuchus or Eutretauranosuchus, and therefore ‘G.’phuwiangensis cannot be, at this point, associated with the genus Siamosuchus.

Based on the information on the provenance of specimens provided by Buffetaut & Ingavat (1983) and Lauprasert et al. (2007), results herein support the occurrence of two different goniopholidid taxa in the pre-Aptian of Sao Khua Formation of Phu Phok (Thailand). The results of the second and third analysis also indicate that ‘G.’phuwiangensis is likely to pertain to a Sunosuchus-like lineage, and its referral to this last genus may be more appropriate than to Goniopholis or Siamosuchus.

DISCUSSION

As phylogenetic relationships are explored, it becomes easier to place the diversity of goniopholidids into context. This allows clarification of questions on taxonomy and evolution. The results indicated the overall relationships of European, North American, and Asiatic goniopholidids, and tested the support of particular hypotheses of relationships and classification. In summary, it is possible to recognize that: G. kiplingi, G. simus, and G. baryglyphaeus are closely related; the three unnamed European taxa form a distinct lineage of the European radiation; Nannosuchus is a basal European goniopholidid, and there is no support to consider it as part of the genus Goniopholis (contraSalisbury, 2002), nor to relate it to G. simus (contraJoffe, 1967); Denazinosuchus is a goniopholidid (as in Wiman, 1932; contraLucas & Sullivan, 2003), but its relationships are not fully understood; ‘G.phuwiangensis is more closely related to Sunosuchus than to Goniopholis (contraBuffetaut & Ingavat, 1983), and there is no support for a close relationship with Siamosuchus; Su. thailandicus is the sister-taxon of Su. miaoi, and is undoubtedly part of the genus Sunosuchus (as in Buffetaut & Ingavat, 1980); the morphology of Vectisuchus is not congruent with its position as part of the Goniopholididae (contraBuffetaut & Hutt, 1980), but as the sister group of Elosuchus (as in Young & Andrade, 2009), and placed in Tethysuchia; ‘G.hartti (= Sarcosuchus hartti) is the sister group of Sa. imperator, confirming the taxonomic amendment by Buffetaut & Taquet (1979).

Our analysis also identifies a major split in goniopholidid evolution that led to a Goniopholis-like lineage, with broad rostrum and conspicuous festooning, and a Sunosuchus-like lineage, with narrower unfestooned rostrum. On the one hand, taxa commonly referred to Goniopholis nest in both lineages, creating a confusing usage for the genus. On the other, characters that previously defined Goniopholis and were common in the group have been found not to be diagnostic, and to be more widespread in their occurrence than originally thought. These factors have all made it possible to revise our understanding of Goniopholis (definition, range of taxa), and the distribution of goniopholidids.

Definition of the genusGoniopholis

Amongst the taxa included in the present analysis, the genus name Goniopholis was originally applied to the European G. simus, G. baryglyphaeus, Nannosuchus, and three unnamed taxa (Hulke, 1878; Dollo, 1883; Hooley, 1907), and to the Asian ‘G.’phuwiangensis, as well as to Amphicotylus, to Denazinosuchus, and to several other putative species from North America. In this traditional sense (e.g. Hulke, 1878; Owen, 1878; Buffetaut, 1982), the use of the name Goniopholis is linked to characters of tooth (crown) morphology, osteoderms that are wider than long (bearing an anterior process at the anterolateral corner), and confluence of alveoli 3–4 (see Buffetaut, 1982; Buffetaut & Ingavat, 1983). As seen here, these characters are present in other clades (e.g. Thalattosuchia, Tethysuchia) and cannot be further used to identify Goniopholis, or even Goniopholididae. Furthermore, the spread of taxa previously referred to ‘Goniopholis’ is almost as wide as Goniopholididae itself. Following this broad interpretation of Goniopholis, the genus corresponds to the well-supported (3.2/64/58) clade of ‘broad-snouted’ goniopholidids (Fig. 12). The use of this node to represent Goniopholis is relatively simple, but would imply the inclusion of Si. phuphokensis in this genus, and yet does not allow the inclusion of ‘G.phuwiangensis (Fig. 13C, D). Furthermore, the relationships amongst the multiple broad-snouted goniopholidids have not been fully explored, and future analyses may demand further taxonomic amendments. Until additional work expands the present set of data, we find it more convenient to preliminarily refer to this group as ‘derived’ goniopholidids.

As seen in the comparative description and phylogenetic analysis herein, the ‘derived’ goniopholidids comprise a range of morphological patterns (see Fig. 3), most of which can be distinguished easily from G. simus. This is the case for Hooley's and Dollo's specimens, which have no transfrontal crest, but bear long postorbital ‘spines’ that shield the orbit laterally; or for Hulke's specimen, which also lacks the transfrontal crest, but actually has a proportionally narrower and longer rostrum than all its relatives. A more manageable interpretation of the genus Goniopholis can be restricted to the node comprising G. baryglyphaeus, G. kiplingi, and G. simus, offering several advantages: the group is easily distinguishable from others (e.g. anterior border of maxillary depression well defined, anterior face of palatines truncated, presence of transfrontal crest with medial intumescence) (2) the close relationship between G. baryglyphaeus and G. simus has already been recognized in the literature (Schwarz, 2002; Lauprasert et al., 2007); and (3) the node has reasonable support (2/80/74), clearly composing a monophyletic group. Furthermore, the change does not require unnecessary additional names or nomenclatural changes, because the unnamed European forms arguably demand at least new specific names, and ‘G.phuwiangensis needs reassignment by any definition adopted for Goniopholis. More important than these, the restricted use of Goniopholis appropriately reflects the morphological patterns seen amongst ‘derived’ goniopholidids, and also our growing knowledge of the group's diversity. It also offers the advantage of providing a more consensual use of names in a phylogenetic framework. A new revised definition of the genus Goniopholis is therefore proposed:

SYSTEMATIC PALAEONTOLOGY

Suborder Mesoeucrocodylia Whetstone& Whybrow, 1983

Family Goniopholididae Cope, 1875

GenusGoniopholisOwen, 1841

Type-species: Goniopholis crassidens Owen, 1841

1841 Goniopholis crassidens Owen

1878 Goniopholis simus Owen

1999 Goniopholis crassidens (Owen) Salisbury et al.

1999 Goniopholis simus (Owen) Salisbury et al.

2002 Goniopholis simus (Owen) Salisbury et al.

2002 Goniopholis baryglyphaeus Schwarz.

Etymology: After Gonius, meaning ‘angled’; and Pholis, meaning ‘scale’.

Geographical range: Continental Europe and England.

Stratigraphical range: Upper Jurassic (Kimmeridgian) to Lower Cretaceous (Berriasian).

Diagnosis for the genus: Neosuchian crocodylomorphs with the following combination of characteristics: platyrostral mesorostrine−sublongirostrine skull, laterally expanded premaxillae, shaped like an axe blade in dorsal view; naris dorsally orientated; perinarial crests present, robust and larger lateral to the naris; naso-oral fossa longer than wider, diamond-shaped; maxilla festooned, with well-defined anterior wave, projecting laterally and ventrally; deep maxillary depressions facing laterally, located next to the jugal suture, with abrupt margins around its entire perimeter and complex internal structure (i.e. divided internally by dorsoventrally orientated ridges); periorbital crests present and robust in structure; dorsal periorbital crest extending on lachrymal−prefrontal surfaces, next to the anterodorsal−dorsal border of the orbit, and presenting a notch at the prefrontal−lachrymal suture; ventral orbital crest formed by projecting jugal margin, which is deflected anterior to the orbit and delimits the ventral edge of a lachrymal fossa; anterior border of the orbit composed by lachrymal and jugal; extensive scar for the attachment of palpebral, reaching at least the lachrymal and prefrontal; frontal and postorbital in contact with palpebral; single palpebral with an overall triangular shape, firmly attached to the primary orbital border, and creating a secondary orbital border; frontal anterior process in a lower level relative to its main body; frontal anterior process separated from main body by a strong transfrontal crest, transversally orientated and bearing a medial buttress; frontal and prefrontal participating in the primary orbital border, but excluded from secondary orbital border (shared with Protosuchidae, Peirosauridae, and other Goniopoholididae) by the palpebral; postorbital with anterolateral process present, robust and short; supratemporal fossae medium to large, and square-shaped to subrectangular in shape; anterior face of palatine process at palate is truncated, with suture transversely orientated.

Remarks on diagnosis: Goniopholis is further characterized by the following features, frequently found in other Goniopholididae, as well as other crocodylomorphs: skull ornamentation dominated by subcircular pits, when present, with grooves absent or only poorly represented at maturity, not reaching the area next to the maxillary alveolar margin; internarial bar absent; premaxillary margin high anterior to naris; nasals contact premaxilla, excluded from naris; nasals contact lachrymals at medial margin only; postnarial fossa present; antorbital fenestra/fossa absent; lachrymal fossa greatly reduced, delimited dorsally and ventrally by crests; laterodorsally orientated orbits; prefrontals composing the anterior to medial borders of the orbits; frontal wide, flat, T-shaped; frontal-parietal contact broadly inside the intertemporal bar and supratemporal fossae, with parietal−postorbital contact not exposed at skull table; supratemporal fenestrae lacking an evident main axis; laterotemporal fenestrae longer than higher, laterodorsally orientated; jugal anterior ramus wide; postorbital bar inclined posteromedially; lateral surface of jugal−quadratojugal fully ornamented; cranioquadrate canal open laterally; otoccipital not enclosing the cranioquadrate passage; post-temporal foramen present, wider than higher; foramen squamoso-otoccipitalis present; choanal border composed by palatines and pterygoids; choanae lanceolate, ample, longer than wider, as narrow as the nasopharyngeal duct; ventral face of the nasopharyngeal duct not trenched; posterolateral palatine processes absent; palatine−ectopterygoid contact absent; pterygoids reach suborbital fenestra; dorsal surface of retroarticular process faces posterodorsally, with medial wing high and facing medially; five premaxillary, 18–20 maxillary, and 20–24 dentary alveoli; teeth crowns anisometric, subisomorphic (single cusped, caniniform), keeled, nonziphodont or false-ziphodont; poorly compressed (subcircular cross-section), well ornamented with basi-apical, poorly anastomosed, well-defined enamel ridges; maxillary teeth with oblique (paradistal) implantation; dorsal and ventral osteoderms present, forming an extensive dermal armour; dorsal armour composed of paired dermal scutes, set in two paramedial rows; paramedial osteoderms are subrectangular, wider than longer, not keeled, with a flat but strongly ornamented dorsal surface.

Taxonomic amendments

Considering this new definition, currently only three species may be considered as members of the genus Goniopholis: G. baryglyphaeus, G. kiplingi, and G. simus. All other references to the genus are in error. The impact on the taxonomy of Goniopholididae is not profound though, as most taxa have already been attributed or reverted to other genera (e.g. Amphicotylus, Denazinosuchus, Nannosuchus).

Most North American specimens referred to Goniopholis have not been evaluated through phylogenetic analysis. ‘Goniopholisstovalli is included in the phylogenetic analysis of Turner & Buckley (2008), but the resulting topology only places the taxon in a polytomy containing G. simus, Calsoyasuchus, Sunosuchus, and Eutretauranosuchus. In the present analysis, two taxa (Amphicotylus and Denazinosuchus) are found to be closely related to the genus Goniopholis (sensu this paper). The inclusion of these North American taxa/specimens in the genus Goniopholis is misleading. They should be referred provisionally to other previously used names (e.g. Amphicotylus), or to new generic names. In any case, a definitive statement on their relationships and generic assignment should await detailed revision and description of the material.

Asian genera, according to their affinities, may be preliminarily considered as Sunosuchus or Siamosuchus, but not Goniopolis. From the small range of taxa analysed, current data support the referral of Su. thailandicus to the genus Sunosuchus (as in Buffetaut & Ingavat, 1980), as it has a close and more exclusive relationship with Su. miaoi. The results however, point toward a close relationship between Su. junggarensis and Eutretauranosuchus, which is supported by other phylogenetic results (e.g. Allen, 2007; Lauprasert et al., 2007). Furthermore, Sunosuchus shunanensisFu et al., 2005 has important morphological differences that may indicate a much more basal position in the clade, a hypothesis yet to be tested through phylogenetic analysis. The taxonomy of the Sunosuchus group appears to be complex, and the internal relationships of the node containing all Sunosuchus and Eutretauranosuchus are yet to be fully resolved. Therefore it seems premature to propose taxonomic amendments for this group. Nonetheless, considering the results of the phylogenetic analysis herein, it seems appropriate to introduce a single nomenclatural change, for ‘G.phuwiangensisBuffetaut & Ingavat, 1983. This is important because the taxon does not fit any possible definition of Goniopholis, being more closely related to the Sunosuchus group. Considering the fragmentary nature of the specimen, we prefer to avoid creating a new monospecific genus to accommodate the taxon. Instead, we propose that ‘GoniopholisphuwiangensisBuffetaut & Ingavat, 1983 is provisionally moved to the genus Sunosuchus until further studies can accurately evaluate the internal relationships of the Sunosuchus group. This taxon is therefore changed to Sunosuchus phuwiangensis (Buffetaut & Ingavat, 1983) comb. nov., and should not be further referred to the genus Goniopholis.

Critical review of the distribution ofGoniopholisand Goniopholididae

Worldwide reports of Goniopholis have been based mostly on teeth or fragmentary material, and a wide sample of teeth and osteoderms can be found in most palaeontological collections in Europe, North America, Brazil, and Asia. Indeed, the morphologies of teeth and dermal scutes contrast with the pattern seen in basal crocodylomorphs, notosuchians, eusuchians, and most thalattosuchians. However, in the case of tooth morphology, overall the same pattern (i.e. robust crowns, keeled, with intense well-defined ornamentation that becomes anastomozed near the apex, nonziphodont to false-ziphodont) can be found in many goniopholidids, pholidosaurids, and dyrosaurids. Among thalattosuchians, at least the teleosaurid Machimosaurus has remarkably similar tooth crown morphology. Dermal scutes, on the other hand, seem to allow taxonomic assignment within certain groups (D. Schwarz-Wings, pers. com. 2010), but isolated material still provides only limited clues for identification. Indeed, not much is known of the intraspecific (or intrageneric) variability of teeth and components of dermal armour, a problem that hinders comparison between taxa, and identification of truly meaningful characters. The use of tooth morphology is particularly important, because it is the sole element supporting the Goniopholis paulistanus (Upper Cretaceous, Brazil). The few cases where the crown morphology provides highly distinctive characters that allow specific/generic identification are usually restricted to notosuchian taxa with remarkable heterodonty (e.g., Candidodon, Chimaerasuchus). Within neosuchia, perhaps the best exemple of autapomorphic teeth is Iharkutosuchus, but it must be noted that the taxon was described based on plentiful skull material (see Osi & Weishampel, 2009). It was also previously pointed out that tri-faceted laminar ziphodont crowns – so far – only occur in the metriorhynchid genus Geosaurus (Young & Andrade, 2009; Andrade et al., 2010). The identification of tribodontic teeth from the Purbeck beds to Bernissartia is another example, but it was also regarded as extremely tentative (see Salisbury, 2002; and references within). Salisbury et al. (1999) have already recognized that the variability present in tooth morphology in goniopholidids prevents the use of such characters to support identification of species. Apart from the examples above, descriptions that state presence of ornamentation, rhomboid apex, or presence of ‘bicarinate teeth’ in most cases cannot consubstantiate discrete characters that can – based on crown morphology alone – secure a definitive specific/generic assignment, let alone the erection/maintenance of a new taxon. Still, crown morphology remains an overall useful tool for gross recognition of morphotypes (see Andrade et al., 2010), but any taxonomic assignment or species definition must also be supported by a phylogenetic analysis, or other meaningful associated remains. Based on the above, tooth morphology is seen here as neither sufficiently precise to provide a secure assignment to Goniopholis/Goniopholididae, nor to provide a diagnostic combination of features. Therefore, its usage alone is disregarded here. In the absence of meaningful information that allows the placement of this type of material (i.e., teeth and other isolated material) in a phylogenetic analysis, the material itself cannot be used to support taxonomic assignment. In these cases, the phylogenetic analysis cannot effectively provide clues to the possible relationships, simply because of scarceness of details. The removal of these specimens from the genus is important, for they offer feeble support for the wide geographical and stratigraphical distribution of Goniopholis (previously global, ranging from Upper Jurassic to Upper Cretaceous).

Trustworthy accounts of Goniopholis are limited to Laurasian sediments (see a reappraisal of brief Gondwanan reports below). Considering the new definition for the genus presented herein, the genus itself is limited to Europe, and ranges from the Kimmeridgian (G. baryglyphaeus) to the Berriasian (G. simus, G. kiplingi). The presence of Goniopholis in Upper Cretaceous sediments was mostly based on Denazinosuchus (Kirtland Formation, USA), a relation that is rejected by our analysis. The putative occurrence of Goniopholis in South-East Asia (Thailand) was based on Su. phuwiangensis, and is also rejected.

Two cases of co-occurrence of ‘broad’ and ‘narrow-snouted’ goniopholidids in the same formation are now known. The first case is Amphicotylus and Eutretauranosuchus, both from the Morrison Formation (Morrison Basin), Kimmeridgian of the USA. The second case becomes evident with the recognition that Su. phuwiangensis is a ‘Sunosuchus-like’ taxon, and comes from the Sao Khua Formation of Thailand (Khorat Group), pre-Aptian of Thailand. Although further research is needed to explore the stratigraphical distribution and spatial range of these taxa, both cases may indicate niche partitioning in the group.

Overall, goniopholidids represent an important radiation from the Early Jurassic (Calsoyasuchus) to the Late Cretaceous (Denazinosuchus) of the USA (∼126 Mya), which has a particularly diverse record through the Late Jurassic−Early Cretaceous boundary (Fig. 14), spreading through Asia and Europe. The clade of European goniopholidids remains a poorly understood group, but is here recovered as monophyletic, with Dollo's specimen indicating presence up to late pre-Aptian times. Interestingly, Sunosuchus has no record in Europe, and the genus is limited to Asia. The earliest record of Sunosuchus is from the Middle Jurassic of China (Su. shunensis), and the latest are reports from the ?late Early Cretaceous of Mongolia/Russia (see Schellhorn et al., 2009 for a detailed review). In general, records of Sunosuchus-related taxa are older than the records from Goniopholis-related taxa. However, the genus Sunosuchus remains poorly explored and demands appropriate revision, which may reveal as much diversity as seen in Goniopholis sensu lato.

Figure 14.

Phylogenetic relationships of Goniopholididae and other main groups of neosuchians, calibrated through time. The goniopholidid fossil record begins in the Early Jurassic, but most members of this group occur from the Late Jurassic to the Early Cretaceous (pre-Aptian). Calsoyasuchus is the oldest taxon, and appears as the basal-most goniopholidid. Denazinosuchus is the most recent record (Late Cretaceous), but its relationships are poorly understood. Dashed ranges represent variation in the age attributed to the material; grey/red ranges are uncertain records; dotted lines represent taxa of uncertain relationships. Abbreviations: Ng, Neogene; Q, Quaternary.

The Gondwanan record of Goniopholis and Goniopholididae

Gondwanan reports of Goniopholis are much rarer than Laurasian ones. However, a couple of reports originate from Brazil. ‘Goniopholishartti (Marsh, 1869) from the Lower Cretaceous is in fact a member of the genus Sarcosuchus. The results herein confirm for the first time, on the basis of a phylogenetic analysis, the taxonomic amendment proposed by Buffetaut & Taquet (1979). Sarcosuchus hartti is either the sister group of Sa. imperator, or the first is a senior synonym of the second, and not part of the genus Goniopholis. The second report is G. paulistanusRoxo 1936, from the Upper Cretaceous Bauru Group (São Paulo State, Brazil). The material is limited to two tooth crowns and a fragment of a (right) tibia, and lacks diagnostic characters that can provide a reliable link to the genus Goniopholis, or even to the family Goniopholididae (see ‘Remarks on diagnosis’ of Goniopholis), regardless of its taxonomic status. These specimens of G. paulistanus are here treated as Neosuchia incertae sedis.

These Gondwanan accounts for Goniopholis (either sensu lato or stricto) are therefore based on information that cannot be truly linked to this genus. By solving these problematic reports, it is confirmed that Goniopholis is not known from Gondwana. Furthermore, the Bauru specimens assigned to the genus erroneously indicated the presence of the genus in the Upper Cretaceous of South America. Although possible, it seems unlikely that Goniopholis existed in the post-Barremian of Gondwana, and so far no evidence can corroborate this idea. Furthermore, there is plenty of evidence for the presence of tethysuchian taxa in Gondwana (e.g. Sarcosuchus, Meridiosaurus, Elosuchus, Dyrosaurus, Guarinisuchus), and their presence can arguably be used to explain at least some of the Gondwanan accounts of ‘Goniopholis’, based on fragmentary material. However, further research is needed to understand fully the palaeobiogeographical distribution of these clades.

The family Goniopholididae may have been present in Gondwana, although most notes refer to old reports on taxa with similar tooth morphology. This is the case for Itasuchus jesuinoiPrice, 1955, which was originally included in the family but is more often considered as a trematochampsid (e.g. Buffetaut, 1991). At least in South America, there is no further reference to the family Goniopholididae (see Candeiro & Martinelli, 2006). However, goniopholidids may have been present in Gondwanan territory, as preliminarily reported by Sereno (2009). The material corresponds to an undescribed, fairly complete ‘Sunosuchus’-like taxon, and will certainly bring important new information on the evolution and distribution of Goniopholididae.

Major aspects in neosuchian evolution and the paraphyly of‘Pholidosauridae

The majority of published works present a monophyletic Goniopholididae (e.g. Tykoski et al., 2002; Turner & Buckley, 2008; Pol et al., 2009; Young & Andrade, 2009), endorsed by characters such as the characteristic maxillary depressions. This is not unexpected, as goniopholidids have characters that contrast with eusuchians and atoposaurids (e.g. presence of anterolateral process of the postorbital, exclusion of supraoccipital from skull roof). Lauprasert et al. (2007) – who included one of the largest samples of goniopholidids – also found a monophyletic group, but the analysis does not include any Pholidosaurus-like taxon, or dyrosaurids. Currently, only the analysis by Jouve et al. (2006) has recovered a paraphyletic Goniopholididae. However, the topology provided by Jouve (2009) introduces key changes to the same data set, and recovers a monophyletic Goniopholididae. The phylogenetic analysis herein recovers the Goniopholididae as a monophyletic clade, but only with the modest support of a Bremer decay index of 3. It is possible that future work will change this picture but, at present, the monophyly of Goniopholididae can be accepted. This demonstrates the importance of using data from a wide range of taxa within the group.

The monophyly of Pholidosauridae, as traditionally indicated (e.g. Buffetaut, 1982), is also controversial. The group would include several semi-aquatic tubular-snouted longirostrine taxa (e.g. Pholidosaurus, Sarcosuchus, Elosuchus, Terminonaris), but the phylogenetic support for such a group is weak at best. Indeed, most analyses retrieve this assemblage as a paraphyletic group (e.g. Turner & Buckley, 2008; Pol et al., 2009; Young & Andrade, 2009), where Sarcosuchus and Terminonaris are more closely related to dyrosaurids than to Pholidosaurus. Our analysis agrees with a paraphyletic Pholidosauridae (sensu lato), but others disagree, and show Pholidosaurus as more closely related to dyrosaurids than to Sarcosuchus. Jouve et al. (2006) and Jouve (2009) are exceptions, and retrieve pholidosaurids as a monophyletic group. These last two analyses are the only ones to consider Elosuchus and Terminonaris together, which may account for the differences seen in the topologies (but not from the presence of Elosuchus alone, as the taxon is included in this data set). Therefore, no consensus has been achieved on the relationships of the so-called pholidosaurid taxa. Overall, it is only possible to state that dyrosaurids are more closely related to either a Pholidosaurus or a Sarcosuchus lineage, but not to a branch that encompasses both lineages. This favours a more exclusive use of Pholidosauridae, only including Pholidosaurus. Although this leads to potential redundancy (i.e. Pholidosauridae = Pholidosaurus), the arrangement is more consensual and can be easily applicable in the literature (Jouve et al., 2006; Turner & Buckley, 2008; Jouve, 2009; Pol et al., 2009; Young & Andrade, 2009). However, this leaves part of the ‘old pholidosaurids’ without a family. From our analysis, a modified version of the family Elosuchidae can then be used to encompass Elosuchus, Sarcosuchus, and Vectisuchus, which also conforms with the results of most authors (e.g. Turner & Buckley, 2008; Pol et al., 2009; Young & Andrade, 2009). Nonetheless, the theme needs further attention and detail, before definitive statements are made. In particular, a thorough redescription of Pholidosaurus is required, complementing the preliminary review by Salisbury (2002).

Despite disagreements on the relationships of Pholidosaurus, elosuchids, and dyrosaurids, most works do agree that all these forms constitute a true radiation of neosuchians (e.g. Jouve et al., 2006; Turner & Buckley, 2008; Jouve, 2009; Pol et al., 2009; Young & Andrade, 2009). This clade corresponds to the Tethysuchia, a name seldom used, but here applied to represent the node. The major disagreement in the literature is not the relationships of this taxon, but the position of Thalattosuchia relative to Neosuchia and to other tethysuchians. Although the problem of thalattosuchian relationships extends beyond the objectives of this paper, a summary view is discussed here. In some of these works, Tethysuchia is an exclusive group and Thalattosuchia appears as its sister group (Turner & Buckley, 2008; Pol et al., 2009; this paper), or even as a more basal Mesoeucrocodylia lineage (Young & Andrade, 2009). In contrast, a few analyses show Thalattosuchia nested within Tethysuchia (Jouve et al., 2006; Jouve, 2009). Thalattosuchian affinities were extensively explored by Pol & Gasparini (2009), who revised evidence for the close relationship between longirostrine groups. This paper also recovers the thalattosuchians as sister group of a clade of ‘pholidosaurs−dyrosaurids’ (= Tethysuchia sensu this paper). Although the relationships of thalattosuchians are still under discussion, the node containing Tethysuchia is often recovered when a wide sample of neosuchians is included. In the analysis herein the tethysuchian node has support equivalent to the Goniopholididae (Bremer decay = 3), which is actually stronger than the support for Thalattosuchia (Bremer decay = 1.83; bootstrap/jackknife < 50%), a widely recognized clade. To avoid misinterpretation on its content and extent, we find it convenient to update the definition of the group, as follows: Tethysuchia is the clade composed of Pholidosaurus purbeckensis (Mansell-Pleydell, 1888) and Dyrosaurus phosphaticus (Thomas, 1893), their common ancestor and all its descendants. Considering this definition, the presence of Thalattosuchia within the group become irrelevant for taxonomic purposes, as it may be contained in the group, or not. We consider that the inclusion of Tethysuchia within Neosuchia presents no nomenclatural problem because the former is defined as an unranked clade. Indeed, Eusuchia and Thalattosuchia are included in Neosuchia, which is included in Metasuchia, alongside Notosuchia (see Benton & Clark, 1988). As in the case for the restricted use of the family Pholidosauridae, the use of Tethysuchia (as proposed here) is advantageous and can be applied to most references in the literature (Jouve et al., 2006; Turner & Buckley, 2008; Jouve, 2009; Pol et al., 2009; Young & Andrade, 2009).

It is important to recognize that the analysis herein presents a relationship not often retrieved in phylogenetic analyses of crocodylomorphs, i.e. the existence of a basal neosuchian lineage that diverges from the lineage that led to Eusuchia. This lineage comprises an unnamed clade that is sister group of Stolokrosuchus, and includes goniopholidids, tethysuchians, and also thalattosuchians. Within the current literature, only Jouve et al. (2006) has a similar arrangement. In most other works, Goniopholididae is more closely related to Eusuchia than to Tethysuchia (Turner & Buckley, 2008; Jouve, 2009; Pol et al., 2009), although Young & Andrade (2009) retrieved a topology where tethysuchians are more related to eusuchians instead. The single origin for goniopholidids, tethysuchians, and thalattosuchians implies that a flourishing wide range of narrow-snouted longirostrine taxa (including part of Goniopholididae) gave rise to platyrostral mesorostrine forms (‘broad-snouted’ goniopholidids) some time during the Late Jurassic. This scenario is interesting and is contrary to the pattern seen in the radiation of crown crocodylians, where the tubular-snouted gavialoids arose from a range of mesorostrine platyrostral forms (although much later).

Present and future of goniopholidid/pholidosaurid systematics

The recent publication of new information on goniopholidids and pholidosaurids, particularly the wider sampling of taxa in phylogenetic works and the redescription of neglected and forgotten taxa, represent true advances in neosuchian systematics, tendencies that must be pursued. Recent works by Jouve et al. (2006), Lauprasert et al. (2007), Turner & Buckley (2008), Jouve (2009), and Pol et al. (2009) renew the available information and allow wider access to key data. In this sense, the present work contributes particularly with the description of a new species, and the introduction of three widely neglected forms, leading to a new definition of the genus Goniopholis.

However, not all aspects of the present analysis are definitive, and new data may substantially improve the results. Further information on goniopholidid taxa should be added, following direct study of specimens, particularly for Amphicotylus, Denazinosuchus, Eutretauranosuchus, Stolokrosuchus, and Sunosuchus. Goniopholis stovalli, Su. shunensis, and the Sunosuchus-like taxon reported by Sereno (2009) represent important unsampled diversity within Goniopholididae. Considering the range of taxa, basal forms such as Zaraasuchus and Sichuanosuchus were not included, and amongst metasuchians, the addition of Shamosuchus, Khoratosuchus, Pachycheilosuchus, and the Glen Rose form, would indeed contribute relevant data. Among tethysuchians, the inclusion of Terminonaris and Meridiosaurus are quite important to address questions on the evolution and distribution of the group, as well as the monophyly/paraphyly of pholidosaurids. Furthermore, many thalattosuchian taxa may be included in the analysis (e.g., Teleosaurus, Teleidosaurus, Suchodus). As crocodylomorph diversity is better reported by means of thorough descriptions, so too will our knowledge of their relationships, evolution, and distribution improve.

CONCLUSIONS

The present work introduces a new species, G. kiplingi, and by means of an extensive and well-documented analysis, explores the phylogenetic relationships of goniopholidids. Goniopholis has been traditionally misused as a waste-basket name for highly fragmentary and nondiagnostic material, hindering our understanding of its distribution in space and time. The phylogenetic analysis shows that only three species can be confidently assigned to the genus Goniopholis (G. simus, G. baryglyphaeus, G. kiplingi), which leads to a new definition proposed here. Following this definition, Goniopholis is a highly exclusive clade, restricted to the Kimmeridgian−Berriasian of Europe, with no evidence of its presence in South America, Africa, Australia, India, or Madagascar.

Our analysis confirms the current generic placement of Sa. hartti and Su. thailandicus, as proposed by Buffetaut & Taquet (1979) and Buffetaut & Ingavat (1980), respectively. We however reject the placement of ‘G.phuwiangensis in Goniopholis (contraBuffetaut & Ingavat, 1983) or Siamosuchus. The referral of highly fragmentary material to Goniopholis, frequently based on non-apomorphic characters, is dismissed. Currently, the evidence for the presence of Goniopholis in Gondwana is feeble, and the material still needs proper description and documentation. The same applies to other gonipholidid genera such as Sunosuchus, Calsoyasuchus, and Eutretauranosuchus. Tethysuchians, in contrast, clearly had a wide distribution in Gondwana, based on complete and thoroughly reported specimens (Sarcosuchus, Elosuchus, Dyrosaurus, Congosaurus, Guarinisuchus).

Vectisuchus, considered as a goniopholidid by Buffetaut & Hutt (1980), is found to share a more exclusive relationship with tethysuchians than other goniopholidids (as in Jouve et al., 2006; Young & Andrade, 2009; contraJouve, 2009), particularly with Elosuchus.

Nannosuchus gracilidens, previously assigned to Goniopholis, is here reverted to its original form (contraSalisbury, 2002), based on our phylogenetic analysis. Although a revision of the status of Nannosuchus is beyond the scope of this paper, our phylogenetic results show that its inclusion in the phylogenetic analysis does not lead to rogue behaviour or substantial impact on three statistics, but rather improves the resolution of the final topology.

More importantly, the present work finds support for a monophyletic Goniopholididae, but not for Pholidosauridae (in its traditional sense). Extensive and detailed revision of these groups is still needed, particularly the genus Pholidosaurus and the North American goniopholidids, which are often poorly documented. Our phylogenetic analysis shows that goniopholidids, tethysuchians, and most likely thalattosuchians, constitute an early split in neosuchian evolution, which radiated to a highly diverse lineage, hitherto unnamed, and not closely related to Eusuchia.

NOTE ADDED IN PROOF

After the submission of the manuscript, a revised description of Eutretauranosuchus became available, based on new specimens (Smith et al., 2010). It should be noted that both the description and the phylogenetic analysis herein do not include information from this paper. This is important, since the new specimens bring substantial new information to the picture, although periorbital morphology remains poorly explored in Eutretauranosuchus.

ACKNOWLEDGEMENTS

The authors are indebted to Chris Moore and Steve Etches, who assisted in the collection of DORCM 12154, and to Chris and Alex Moore, for early preparation of the specimen and advice on additional preparation. Christopher Lamb directed the CT scan at the Royal Veterinary College London. Paul Ensom gave invaluable help in interpreting the stratigraphy of the outcrop and determining the stratigraphical origin of the specimen. The specimen was recovered with the kind permission of Swanage Town Council and Natural England. For access of specimens under their care or study, we thank S. Chapman, L. Steel, P. Barrett, and C. McCarthy (BMNH), J. Cooper (BMNHB), L. Loeffler (BRSUG), O. Rauhut (BSPG), R. C. T. Cassab, and D. G. Campos (DNPM), J. Liston (GLAHM), A. Folie, T. Smith, and E. Sternbaum (IRSNB), R. J. Bertini (IGCE-UNESP), I. S. Carvalho, F. Vasconcellos, and T. Marinho (DG-UFRJ), J. F. Bonaparte, A. Kramarz, J. B. Desojo, and F. Novas (MACN), D. Schwarz (MB), D. Pol (MEF), C. E. M. Oliveira (FEF), M. Ramalho, and J. Sequeira (MG), Z. B. Gasparini, M. Reguero, and S. Bargo (MLP), A. W. A. Kellner, S. A. K. Azevedo, L. Carvalho, and D. D. R. Henriques (MN-UFRJ), B. Battail, and S. Bouetel (MNHN), A. C. Arruda Campos (MPMA), C. Howells (NMW), J. O. Calvo, J. Porfiri, and L. E. Fiorelli (Proyecto Dino), C. C. Guerra (PUC-MG), E. Frey (SMNK), R. Schoch (SMNS), A. Buscalioni, F. Ortega, and J. L. Sanz (UAM), and G. Dyke (UCD). Of particular importance to this study, the authors are truly indebted to J. Cooper (BMNHB), for the loan of Hulke's specimen (‘Mr Willett's crocodile’), S. Chapman (BMNH) for study of Hooley's specimen, and to A. Folie and T. Smith (IRSNB), for access to Dollo's specimen. M.B.A. is specially indebted to Jahn J. Hornung (Universität Göttingen) for sharing his expertise in goniopholidid crocodylians. For further information on geology/specimens and access to photos and papers, the authors thank L. E. Fiorelli (CONICET), D. Fortier (UFRGS), C. C. Guerra (PUC), J. J. Hornung (Universität Göttingen), S. G. Lucas (NMMNHS), A. E. P. Pinheiro (DG-UFRJ), D. Pol (MEF), D. Riff (INBIO-UFU), D. Schwarz-Wings (MB), W. A. P. Wimbledon (BRSUG), and M. Young (BRSUG). S. Powell (BRSUG) produced key photographs of G. kiplingi (Figs 6, 8), and also gave valuable direction on DSLR macrophotography and image treatment. Daniela Schwarz-Wings (MB), Emily Rayfield and Marcello Ruta (BRSUG) kindly revised an early version of the text, and J. Clark (George Washington University) and D. Pol (MEF) provided detailed critiques that improved the original manuscript. Eric Franzosa (Department of Biology, Boston University) provided support for the use of TreeRot software to M. B. A. (Sorenson & Franzosa, 2007). Translation of work(s) in foreign languages (i.e. Gao, 2001) was by Will Downs, and obtained as courtesy from the webpage ‘The Polyglot Paleontologist’ (Carrano, 2005), freely available from <http://ravenel.si.edu/paleo/paleoglot/>. The specimen was prepared and scanned with financial support from the Jurassic Coast Trust, England. This paper was originally produced as part of M.B.A.'s PhD Thesis, which was supported by a scholarship from Conselho Nacional de Desenvolvimento Cientifico e Tecnológico (CNPq – Proc. n° 200381/2006-7), Brazil. Additional funds to M. B. A., which supported short-term visits to collections, were provided by SYNTHESYS (FR-TAF-4858; BE-TAF-5357), the Bob Savage Memorial Fund (BRSUG), a Sylvester Bradley Award (Palaeontological Association), and TST Funds (BRSUG). The Synthesys project http://www.synthesys.info/) is financed by the European Community Research Infrastructure Action under the FP6 ‘Structuring the European Research Area Programme’.

Appendix

APPENDIX 1

Apomorphy list and nodal support (Bremer/bootstrap/jackknife) for key clades of Neosuchia discussed in this paper, with emphasis in Goniopholididae and related groups (full apomorphy/character report in Supporting Information Files S1 and S2). Double arrows (‘’) = unambiguous transformations; simple arrows (‘→’) = ambiguous transformations; * = clades with limited representation, considered the current known diversity. Bootstrap/jackknife frequencies shown for values above 50% (otherwise = ‘–’), and expressed in percentages.

((Thalattosuchia + Tethysuchia) + Goniopholididae)) + Stolokrosuchus (1.83/–/–):

3 (0  1); 5 (1 → 3); 46 (0 → 1); 69 (4  1); 72 (2  1); 80 (1  0); 216 (0  1); 217 (1 → 2); 224 (0 → 1); 309 (0  1); 368 (0  1)

(Thalattosuchia + Tethysuchia) + Goniopholididae (1.83/–/–):

37 (2  0); 45 (0 → 1); 50 (0  1); 65 (0  3); 110 (0 → 2); 133 (0  1); 151 (0  1); 333 (0 → 1); 369 (0 → 1); 375 (0 → 2); 383 (1 → 0); 403 (0 → 1); 405 (1 → 2); 416 (0 → 1); 477 (0 → 1); 481 (1 → 0)

Thalattosuchia + Tethysuchia (1.83/–/–):

4 (1  0); 5 (3 → 2); 8 (1  2); 70 (2  0); 81 (1  0); 106 (0  2); 107 (0  2); 192 (0  1); 195 (1  0); 240 (2  1); 277 (0 → 1); 282 (1  0); 332 (0  2); 338 (2 → 0); 343 (1  2); 384 (1 → 0); 392 (1  0)

Tethysuchia (3/–/–):

2 (2 → 0); 43 (1  0); 44 (1  0); 57 (2  1); 64 (0  1); 68 (1  0); 130 (1 → 0); 142 (0  1); 148 (0  1); 152 (0  1); 194 (1  0); 353 (0  1); 469 (0 → 1)

Dyrosauridae + Pholidosaurus (1/–/–):

31 (1 → 2); 145 (0 → 1); 167 (0  1); 180 (1  2); 360 (1 → 2); 363 (0 → 1); 407 (0  2); 421 (0  1)

Elosuchidae (including Sarcosuchus; 1/–/–):

109 (1 → 0); 115 (0  1); 124 (0 → 1); 135 (0  1); 153 (0 → 1); 157 (1 → 2); 178 (0 → 1); 348 (0  1); 373 (0 → 1); 391 (1  2); 408 (1  0)

Elosuchus + Vectisuchus (2/69/68):

110 (2 → 1); 205 (0 → 2); 313 (0  1); 332 (0  1)

Goniopholididae (3/–/–):

26 (0  1); 79 (0 → 1); 86 (0  1); 89 (0  1); 122 (1  0); 202 (0 → 1); 223 (1  0)

Sunosuchus + Eutretauranosuchus (1/–/–):

110 (2 → 1); 205 (0 → 2); 313 (0  1); 332 (0  1)

Sunosuchus miaoi + Sunosuchus thailandicus (3.25/–/–):

315 (0  4); 316 (0  1); 346 (0 → 1)

‘Derived’ Goniopholididae (3.2/64/58):

[Obs. derived goniopholodids are here considered Siamosuchus + Amphicotylus + all European goniopholidids]

29 (0  1); 58 (0  1); 82 (0  1); 99 (0  1); 134 (0  1); 277 (0 → 1); 373 (0  1); 384 (1 → 0); 391 (1  2); 393 (0  1); 397 (1  2)

European Goniopholididae (1/–/–):

34 (0 → 1); 122 (0  2); 126 (0 → 1); 136 (1  2); 142 (0 → 1); 167 (0 → 1); 187 (0 → 1)

Goniopholis sensu stricto + (Hulke + Dollo + Hooley forms) (2/–/–):

90 (0  1); 111 (0  1); 160 (1 → 0); 180 (1  2); 225 (0 → 1); 312 (2 → 0); 402 (0 → 1); 478 (1 → 0)

Goniopholis sensu stricto (2/80/74):

107 (0 → 1); 137 (1  0); 141 (0  1); 161 (0 → 1); 348 (0  1); 398 (0 → 1)

Goniopholis simus + Goniopholis kiplingi (3/93/86):

32 (0 → 1); 69 (1 → 3); 106 (0 → 1); 120 (1  2); 123 (0 → 1); 139 (0  1); 238 (0  1)

Hulke + Dollo + Hooley forms (5/67/61):

99 (1  0); 101 (1  0); 106 (0 → 1); 107 (0 → 2); 115 (0  1); 173 (1  0); 178 (1  0); 194 (1 → 0); 215 (0 → 1); 223 (0  1)

Dollo + Hooley forms (9/97/97):

111 (1  3); 126 (1  2); 152 (1  2); 153 (0  1); 157 (1  2); 175 (1  0); 187 (1  2); 188 (1  2); 195 (1  0); 230 (0  1); 398 (0 → 2); 405 (2 → 0)

Ancillary