5.a. Maxilla (Figs 1, 2)

Most of the body of the maxilla is preserved; the posterior portion and the dorsal portion of the ascending process (= ‘dorsal process’) are missing. The ventral margin is relatively straight in lateral view with small scallops marking the alveoli. Anteriorly, the face of the maxilla arcs medially, forming a rounded anterior surface. This is similar to the condition in Erpetosuchus granti (NHMUK R3139) in which the contact between the maxilla and the premaxilla is pinched medially in ventral view (Benton & Walker, Reference Benton and Walker2002, fig. 2C). There is a large foramen dorsal to the first alveolus that opens anteriorly onto the anterior surface. A similar foramen is also present in Euparkeria (SAM-K-6047), Proterosuchus (Modesto & Sues, Reference Modesto and Sues2004) and Prolacerta (Modesto & Sues, Reference Modesto and Sues2004), but is only rarely present within Archosauria (e.g. Lotosaurus, IVPP V 4880). A similar foramen appears to be absent in Erpetosuchus granti (NHMUK R3139). CT data indicates that this large anterior foramen opens into an elliptical sediment-infilled neurovascular canal that extends posteriorly for the entire length of the element, opening into the antorbital fossa (see below). The neurovascular canal arches dorsally above the first two alveoli and then extends lateral to the alveoli and immediately ventral to the ventral surface of the antorbital fossa. The main neurovascular canal is fully surrounded by bone in some places, but elsewhere it communicates medially with the alveoli.

Two tiny nutrient foramina are present immediately posterior to the large anterior foramen. A further intermediate sized foramen is present, dorsal to the second alveolus. A longitudinal row of several foramina is present just ventral to the ridge that forms the ventral margin of the antorbital fossa. All of the foramina communicate with the main neurovascular canal, either directly or via short canals.

The ridge forming the ventral margin of the antorbital fossa runs subparallel to the oral margin above alveoli four and five. Dorsal to alveolus three the ridge begins to arc dorsally and then posterodorsally, immediately before its termination (as a result of breakage of the ascending process). Anteriorly, the antorbital fossa is not recessed medial to this ridge, unlike the condition in Erpetosuchus granti (NHMUK R3139). Posteriorly, the fossa is recessed medial to the ridge, and some of the internal surface of the fossa is therefore hidden in lateral view. In addition, the ridge expands laterally as it extends posteriorly. As a result, the ridge progressively creates an increasingly well-developed overhang over the ventral portion of the maxilla, with the lateral surface of the posterior part of the maxilla facing ventrolaterally rather than strictly laterally. A similar condition is present in Erpetosuchus granti (NHMUK R3139). The internal surface of the antorbital fossa is corrugated, with several large dorsoventrally extending furrows; this corrugated appearance is probably an external reflection of the alveoli, and is similar to the condition in Riojasuchus (PVL 3829).

Medially, the maxilla bears a sharp longitudinal ridge that bisects the element and that expands transversely at its anterior end to form the palatal process. This process projects a short distance anterior to the main body of the maxilla and is horizontally oriented with a broken medial margin. Because of the breakage it is unclear if the palatal process met its antimere on the midline. The surface dorsal to the ridge and palatal process is corrugated (see above). The five alveoli are positioned ventrolateral to the longitudinal ridge. There are apparently only four teeth in Erpetosuchus granti (NHMUK R3139) and possibly nine in a specimen referred to Erpetosuchus sp. (AMNH 29300: Olsen, Sues & Norell, Reference Olsen, Sues and Norell2000). Each alveolus has a subcircular outline and is separated from adjacent alveoli by thin sheets of bone and unfused interdental plates. Small foramina lie at the dorsal apices of the spaces between the interdental plates. The first two alveoli are significantly smaller than the remaining three. The tooth row is very gently arched anteromedially towards its anterior end. A sharp longitudinal ridge extends from the posterior edge of the fifth alveolus and proceeds posteriorly, forming the ventral margin of the edentulous portion of the element.

In dorsal view, the maxilla is transversely expanded relative to maxillae of other archosaurs. A transversely concave, dorsally facing surface is located posterior to the anterior margin of the antorbital fenestra, and bears a posteriorly opening foramen that represents the exit of the neurovascular canal described above. This concave surface is separated by a weak U-shaped ridge from the rest of the dorsal surface of the element.

A single maxillary tooth was discovered in the fourth alveolus when the specimen was CT scanned. The small tooth appears to lack serrations, is round in cross-section, and slightly posteriorly recurved (Fig. 2). Similar teeth are present in both the holotype of Erpetosuchus granti (NHMUK R3139) and a referred specimen of Erpetosuchus sp. (AMNH 29300).

5.b. Scapula (Fig. 3)

The left scapula is nearly complete and is only missing the distal portion of the blade. The scapula is described with the blade oriented vertically. The element consists of an expanded ventral plate and a blade that bends slightly anteriorly along its length and expands in anteroposterior width at its distal end. The blade has a relatively straight posterior margin and a concave anterior margin, with the result that the distal expansion is asymmetrical, being developed most strongly anteriorly. At its base the blade is constricted anteroposteriorly; the distal expansion (as preserved) of the blade is more than twice the width of the basal constriction. At the proximal constriction the blade has a subcircular cross-section, but distally the blade is transversely compressed. The heavy waisting of the proximal scapula blade in Parringtonia gracilis is similar to that of Erpetosuchus granti and was noted by Krebs (Reference Krebs1965, Reference Krebs and Kuhn1976) and used as a character state to unite the two taxa (Krebs, Reference Krebs and Kuhn1976). The distal portion of the scapula blade is more broadly expanded in Parringtonia gracilis in comparison with Erpetosuchus granti (NHMUK R3139).

The lateral surface of the blade bears two small yet distinct tubera: the first is positioned just dorsal to the glenoid on the posterior margin of the base of the blade, and the second is positioned anteriorly on the lateral surface of the scapula blade at the point of its proximal constriction. The more proximally located tuber is possibly the attachment site of m. triceps brachii and is well developed, as in Erpetosuchus granti and other pseudosuchians such as the loricatan pseudosuchian Batrachotomus (Gower & Schoch, Reference Gower and Schoch2009). The second, more distally placed, tuber is the more pronounced of the two and is unique to Parringtonia gracilis. A sharp and narrow ridge extends from the anteroventral corner of this tuber and forms the anterior margin of the prominent acromion process.

The glenoid faces posterolaterally, similar to the plesiomorphic condition in archosauriforms (Nesbitt, Reference Nesbitt2011). The large acromion process is distinctly raised above the concave proximal surface of the scapula, a character that supports the position of Parringtonia gracilis as an archosaur (Nesbitt, Reference Nesbitt2011). A short, low, anteroposteriorly extending ridge is present on the lateral surface of the proximal end, ventral to the acromion.

In anterior view, the scapula is strongly arched laterally along its length, with the point of maximum curvature occurring at the junction between the proximal plate and the blade. The scapula blade itself is relatively straight in anterior view along most of its length. When the sutural surface for the coracoid is held horizontally, the blade of the scapula blade is directed slightly anterodorsally rather than strictly dorsally.

5.c. Ischium (Fig. 4)

Huene (Reference Huene1939) identified a bone fragment as the proximal portion of a left ischium, but this identification was considered doubtful by A. J. Charig (unpub. Ph.D. thesis, Univ. Cambridge, 1956). Here, we follow the identification of Huene (Reference Huene1939) because it appears the bone has an expanded proximal portion and possibly an acetabular rim. Unfortunately, all of the edges, with the exception of the dorsal edge, are broken. Therefore, it is unclear whether, and how, the ischium contacted the pubis and ilium (contra Huene, Reference Huene1939). The dorsal edge is the thickest portion of the preserved element and is rounded. The medial surface is gently concave and the proximal portion tapers posteriorly.

5.d. Dorsal vertebrae (Fig. 5)

The centra are spool shaped, with lateral and ventral surfaces that are gently concave anteroposteriorly; however, several vertebrae (interpreted as more anterior dorsals) bear a weak midline keel on the ventral surface. The lengths of the centra are typically about 150% of their dorsoventral heights. The weakly amphicoelous articular surfaces are subcircular in outline with gently rounded and slightly thickened rims. The lateral surfaces of the centra lack any kind of lateral fossa immediately ventral to the neurocentral suture, unlike the condition in Euparkeria (SAM-K-6047), suchians (e.g. Revueltosaurus, PEFO 34561) and dinosauromorphs (e.g. Marasuchus, PVL 3870). There are also no lateral fossae in presacral 9 of Erpetosuchus granti (NHMUK R3139).

The neural spines lie over the posterior half of the centrum. They are short and only slightly taller dorsoventrally than the centrum. They are transversely and anteriorly narrowest at mid-height, and they expand weakly anteroposteriorly and very strongly transversely at their apex. In dorsal view, this apical expansion has convex lateral margins and flattened anterior and posterior margins. The dorsal surface is roughly textured and transversely concave, forming a median pit that is also present in Erpetosuchus granti (NHMUK R3139). The dorsal expansions of the neural spines are similar to those present in Euparkeria (SAM-K-6049B) and Erpetosuchus granti (NHMUK R3139). The lateral expansions of the dorsal vertebrae of Erpetosuchus granti (NHMUK R3139) are more restricted anteroposteriorly than Parringtonia gracilis giving them a rectangular shape with a longer mediolateral side. The anterior and posterior surfaces of the spines are formed by sharp ridges.

The large prezygapophyses face dorsomedially at about 30° to the horizontal and have articular surfaces that are gently concave transversely. They do not extend beyond the anterior margin of the centrum. The postzygapophyses face ventrolaterally at a similar angle to the prezygapophyses and overhang the posterior end of the centrum. A deep median fossa located at the base of the neural spine separates the postzygapophyses. Hypantrum–hyposphene accessory vertebral articulations are absent.

The spacious neural canal is relatively large compared to that of the centrum size. Anteriorly, it is broader than deep, with a nearly rectangular outline. The two isolated dorsal centra demonstrate that the neural canal was deeply excavated into the dorsal surface of the centrum. The short parapophyses lie lateral to the mid-point of the neural canal, well above the neurocentral suture and broadly separated from the diapophyses. Their articular surfaces are flattened and face laterally and slightly ventrally. The diapophyses are broken in nearly all of the preserved vertebrae, with one exception. The diapophysis of this vertebra extends posterolaterally and slightly dorsally, and has subparallel anterior and posterior margins in dorsal view. Its distal end is slightly expanded dorsoventrally to form the convex articular surface. In some, but not all, of the dorsals there is a low ridge connecting the dorsal margin of the parapophysis to the anterior margin of the diapophysis. A shallow concavity is present beneath the diapophysis; however, deep fossae and well-developed vertebral laminae are absent.

5.e. Caudal vertebrae (Fig. 5)

Three caudal vertebrae, one anterior, one anterior–mid and one mid–posterior, are represented in the material. The caudal vertebrae have centra that are all approximately the same length. Neurocentral sutures are indistinguishably fused in all three vertebrae. The anterior caudal has dorsoventrally deep, laterally directed transverse processes that have a triangular outline in dorsal and anterior views. The centrum has subcircular and gently concave articular faces; in lateral view the centrum is longer than dorsoventrally deep with constricted lateral surfaces that are concave anteroposteriorly and dorsoventrally. A sharp break-in-slope (paramedian keel) separates the lateral and ventral surfaces of the centrum; the ventral surface is flattened with a weakly developed median groove. The ventral margin of the centrum is bevelled at anterior and posterior ends (this is most obvious in lateral view), with the posterior bevelled surface being the larger of the two. This surface, which articulated with the chevron, is subdivided by a subtle median break-in-slope.

The spacious neural canal of the anterior caudal has a subrectangular outline, being wider transversely than dorsoventrally deep. It is widest at its anterior end and decreases in transverse width posteriorly. The prezygapophyses extend only very slightly beyond the anterior margin of the centrum; they face dorsomedially at approximately 15° above the horizontal. Their articular faces are gently concave transversely. The postzygapophyses face ventrolaterally at a similar angle. A small median fossa is present between the postzygapophyses in posterior view. The neural spine lies over the posterior half of the centrum and its height is 1.5 times greater than the height of the centrum. It is narrowest in anteroposterior and transverse dimensions immediately above the postzygapophyses, and it expands anteroposteriorly and transversely towards its apex. At its apex, the transverse expansion of the spine is marked, but does not occur to the same degree as in the presacral vertebrae. Among early archosaurs, the presence of transverse expansions of the anterior caudal neural spines is rare and only occurs to our knowledge in some aetosaurs (Walker, Reference Walker1961, fig. 10a–d). The dorsal surface of the spine bears a deep anteroposteriorly extending pit surrounded by a distinct rim, as present in the dorsal vertebrae. The anterior and posterior margins of the spine are formed by sharp ridges that birfurcate dorsally to form the lateral margins of the dorsal expansion. The lateral surfaces of the spine are gently convex anteroposteriorly.

The neural spine of the anterior–mid caudal is incomplete. It differs from the anterior caudal primarily in possessing a much less well-developed transverse process. On the ventral surface of the centrum there is a sharp midline ridge or keel, in addition to the paramedian keels seen in the anterior caudal. The mid–posterior caudal (also lacking its neural spine, and with damaged zygapophyses) retains a short and dorsoventrally shallow transverse process. The centrum is nearly as elongated as the more anterior caudal vertebrae, but is substantially reduced in height.

5.f. Osteoderms (Fig. 6)

Five osteoderms, one of which is nearly complete, are preserved. The osteoderms are nearly square with rounded corners. This is in contrast to the osteoderms of Erpetosuchus granti, which are distinctly longer than wide (NHMUK R3139). There is no pointed process on the anterior edge, unlike the condition in paracrocodylomorphs (Nesbitt, Reference Nesbitt2011). The osteoderms are distinctly bent into an L-shape, with an anteroposteriorly extending rounded ridge (which is positioned at two-thirds of the transverse distance from the medial margin of the element marking the point of the bend). The anterior edge of each osteoderm does not have a smooth surface devoid of sculpturing for articulation with the preceding osteoderm like that of aetosaurs (e.g. Lucasuchus; TMM 31100-361), but instead has a sculptured anterior facet that indicates that the osteoderms were imbricated in life, as in Erpetosuchus granti (Benton & Walker, Reference Benton and Walker2002). Deep pits and grooves cover the dorsal surface of the osteoderm and radiate from the midpoint of the anteroposteriorly extending ridge as in Erpetosuchus granti (NHMUK R3139). The presence of deep sculpturing in small osteoderms is rare among early archosaurs, with most small early archosaur osteoderms having relatively smooth surfaces (e.g. Aetosaurus ferratus; SMNS 18554). The ventral surface is smooth and convex anteroposteriorly. The anterior, posterior and lateral edges taper, whereas the medial edge is slightly thickened, suggesting that it contacted its antimere at the midline. It is not clear if Parringtonia gracilis had any osteoderms lateral to the paramedian row, as occurs in one specimen referred to Erpetosuchus granti (Benton & Walker, Reference Benton and Walker2002)

It is not clear how the osteoderms were arranged. However, it is likely that there was only a single pair of paramedian rows (two rows) and there was only one set per vertebrae given the length of each osteoderm relative to each presacral vertebra. This pattern is present in Erpetosuchus granti (NMS 1966.4.3; Benton & Walker, Reference Benton and Walker2002).

6.b. Parringtonia gracilis and Erpetosuchus granti

Our results indicate that Parringtonia gracilis and Erpetosuchus granti form a clade, Erpetosuchidae (see definition and diagnosis in Section 4 above). Within the numerical phylogenetic analysis conducted, the two taxa are united by three unique unambiguous optimized synapomorphies: dentition present only in the anterior half of the maxilla (17-1), mediolateral length greater than dorsoventrally length of the posterior portion of the maxilla (21-1) and tooth serrations absent (168-1). These three characters states have a high consistency index in this analysis (CI = 0.7–1) indicating that they essentially only occur in Parringtonia gracilis and Erpetosuchus granti (see Nesbitt, Reference Nesbitt2011). Our analysis also identifies one other unambiguously optimized character state that unites Parringtonia gracilis and Erpetosuchus granti: paramedian osteoderms with a distinct longitudinal bend near the lateral edge (404-1). However, given the uncertainty of the phylogenetic position of Erpetosuchidae, and that this character state has a low consistency index (= 0.25), this optimization is most likely an artefact given that many early archosaur groups with osteoderms also have this character state (e.g. crocodylomorphs; Nesbitt, Reference Nesbitt2011).

Parringtonia gracilis and Erpetosuchus granti also share many character states not yet sampled in the cladistic analysis. The following shared character states appear to represent synapomorphies of the two taxa, but this must be considered tentative prior to their incorporation into future phylogenetic analyses. Parringtonia gracilis and Erpetosuchus granti share a sharp ridge on the lateral margin of the maxilla that marks the ventral extent of the antorbital fossa. Ventral to this ridge, the external surface of the maxilla is inclined medially towards the alveolar margin. This medially inclined external surface of the maxilla continues posteriorly onto the jugal in Erpetosuchus granti, effectively exposing much of the external surface of the jugal in ventral view. This morphology represents a character state that unites the North American and European specimens of Erpetosuchus with Parringtonia gracilis.

Both Parringtonia gracilis and Erpetosuchus granti bear a hypertrophied tuber hypothesized for the attachment of the m. triceps brachii on the posterolateral surface of the proximal scapula blade. Although other archosauriforms have a small tuber in the same location (Nesbitt, Reference Nesbitt2005, Reference Nesbitt2011), the size of the tuber in Parringtonia gracilis and Erpetosuchus granti is exceptionally large relative to the overall size of the scapula (Fig. 3). Similarly, the extreme waisting of the proximal portion of the scapula blade is rare among non-dinosaurian archosauromorphs. Krebs (Reference Krebs and Kuhn1976) hypothesized that the presence of an anteriorly inclined scapular blade unites Parringtonia gracilis and Erpetosuchus granti. The scapula blade does slightly arc anteriorly in both taxa; however, the amount of curvature in the element in Parringtonia gracilis and Erpetosuchus granti is not dissimilar to that of other archosaurs (e.g. Postosuchus kirkpatricki, TTU P9000).

The transversely expanded distal ends of the neural spines bear median pits on their dorsal surfaces in the presacral vertebrae (vertebrae 4–12) of Erpetosuchus granti, and similar median pits occur in the posterior presacral vertebrae and caudal vertebrae (exact positions unknown) of Parringtonia gracilis (Fig. 5). Many pseudosuchians (e.g. the aetosaur Desmatosuchus; Parker, Reference Parker2008) and taxa just outside Archosauria (e.g. Euparkeria, SAM-K-6047) have laterally expanded distal neural spines (= spine tables), but the dorsal surface is either convex or nearly flat. The deep pits in Parringtonia gracilis and Erpetosuchus granti are only present in these two taxa, as far as we have observed.

Benton & Walker (Reference Benton and Walker2002) challenged the assignment of Parringtonia gracilis to Erpetosuchidae by Walker (Reference Walker1970) and Krebs (Reference Krebs and Kuhn1976), suggesting that most of the similarities between Parringtonia gracilis and Erpetosuchus granti are plesiomorphic for Archosauria. Of the six autapomorphies of Erpetosuchus granti listed by Benton & Walker (Reference Benton and Walker2002), three of the character states (characters 1, 2 and 6) are present on the maxilla, an element that is present in both Parringtonia gracilis and Erpetosuchus granti. Parringtonia gracilis and Erpetosuchus granti share a reduced maxillary tooth row (4–5 teeth) that is restricted to the area anterior to the midpoint of the antorbital fenestra (autapomorphy 1 of Benton & Walker, Reference Benton and Walker2002). In the discussion of this character state, Benton & Walker (Reference Benton and Walker2002) stated that only the anterior part of the maxilla of Parringtonia gracilis is preserved and therefore that it was not clear if Parringtonia gracilis had an anteriorly restricted maxillary dentition. However, there is a clear absence of alveoli in the preserved posterior portion of maxilla (Figs 1, 2). Additionally, Parringtonia gracilis and Erpetosuchus granti share large antorbital fenestra set in a deep fossa whose margins are marked by distinct sharply angled ridges (autapomorphy 2 of Benton & Walker, Reference Benton and Walker2002) and teeth oval in cross-section, lacking distinct anterior and posterior carinae and lacking marginal serrations (autapomorphy 6 of Benton & Walker, Reference Benton and Walker2002), as discussed in Section 4 above.

6.c. The phylogenetic position of Erpetosuchidae

Parringtonia gracilis and Erpetosuchus granti form a clade historically referred to as Erpetosuchidae (Walker, Reference Walker1970; Krebs, Reference Krebs and Kuhn1976) and phylogenetically defined for the first time here. Unfortunately, the phylogenetic position of this clade within Archosauriformes is poorly resolved in the results of our early archosaur phylogenetic analysis. The high number of autapomorphic features in the skull of Erpetosuchus granti (see Benton & Walker, Reference Benton and Walker2002), the uncertainty in determining cranial homologies between Erpetosuchus granti and other early archosaurs (e.g. the ‘overhanging’ squamosal; see discussion in Nesbitt, Reference Nesbitt2011), genuine character conflict (see above) and the absence of information on the anatomy of the braincase, pelvis and hindlimb all contribute to the poor resolution of the relationships of the clade. Minimally, however, Erpetosuchidae can be assigned to the Archosauria based on the presence of palatal processes of the maxilla that are in contact along the midline (32-1), an antorbital fossa that is present on the posterodorsal portion of the maxilla (137-2) and a distinctly raised acromion process (220-1) of the scapula. The cranial and postcranial skeletons of Erpetosuchidae share no unambiguous synapomorphies with most of the clearly monophyletic groups of Triassic archosaurs (e.g. Aetosauria, Dinosauromorpha, Pterosauria).

Previously, Erpetosuchus granti was incorporated into a small phylogenetic analysis including an aetosaur, Gracilisuchus, Postosuchus and early crocodylomorphs by Olsen, Sues & Norell (Reference Olsen, Sues and Norell2000). The subsequent analyses of Benton & Walker (Reference Benton and Walker2002) and Brusatte et al. (Reference Brusatte, Nesbitt, Irmis, Butler, Benton and Norell2010a ) placed Erpetosuchus granti into more comprehensive early archosaur phylogenies with increased taxon sampling, but continued to recover Erpetosuchus granti as the sister taxon to Crocodylomorpha. In all three previous analyses (Olsen, Sues & Norell, Reference Olsen, Sues and Norell2000; Benton & Walker Reference Benton and Walker2002; Brusatte et al. Reference Brusatte, Benton, Desojo and Langer2010a ), the sister-taxon relationship between Erpetosuchus granti and Crocodylomorpha was supported by three character states: the absence of a postfrontal, the complete fusion of the parietal, and an anterodorsally sloped quadrate and quadratojugal. Nevertheless, accurate recognition of the absence of both a postfrontal and an interparietal suture relies upon the ability to observe clear sutures between skull elements. Unfortunately, the only available skull material of Erpetosuchus granti that includes the relevant parts of the skull roof is preserved as a natural mould, as is the case for most of the other Lossiemouth reptiles (Benton & Walker, Reference Benton and Walker1981, Reference Benton and Walker1985). Frequently, sutures cannot be clearly identified in fossils preserved in this manner, and this is particularly true for specimens from the Lossiemouth Sandstone Formation. Therefore, the presence or absence of a postfrontal and an interparietal suture cannot be confirmed with the available material, and these characters were scored as unknown for Erpetosuchus granti in our analysis. Furthermore, it is clear that some early crocodylomorphs (e.g. Dromicosuchus grallator, Sues et al. Reference Sues, Olsen, Carter and Scott2003) have a clear interparietal suture. The anterodorsally sloping quadrate and quadratojugal may represent a shared character of crocodylomorphs and Erpetosuchus granti; however, this character should be scored cautiously given that the accurate assessment of the orientation of the quadrate and quadratojugal in life requires exquisite preservation and the absence of crushing. Crocodyliforms clearly have a similarly sloping quadrate and quadratojugal, but the orientation of these elements in early crocodylomorphs such as Dibothrosuchus elaphros (Wu & Chatterjee, Reference Wu and Chatterjee1993) and Dromicosuchus grallator (Sues et al. Reference Sues, Olsen, Carter and Scott2003) remain uncertain.

Given these scoring differences, we examined the evidence for a possible sister-taxon relationship of Erpetosuchidae and Crocodylomorpha in our analysis. Even if Erpetosuchus granti is scored as lacking both a postfrontal and an interparietal suture, Erpetosuchidae is still recovered in a polytomy at the base of Archosauria, and is not recovered as the sister taxon of Crocodylomorpha. A sister-taxon relationship between Erpetosuchidae and Crocodylomorpha requires 13 extra steps in our analysis. If Erpetosuchus granti is solely used to represent Erpetosuchidae (i.e. if Parringtonia gracilis is a priori excluded), it is recovered as the sister taxon of aetosaurs + Revueltosaurus, but this position is poorly supported (Bremer support = 1).

Our results illustrate the difficulties involved in accurately reconstructing the phylogenetic position of Erpetosuchidae within Archosauria. Consequently, the discovery of new material of erpetosuchids is likely required to fully resolve the relationships of the enigmatic clade.

At the moment, we restrict Erpetosuchidae to only include Erpetosuchus and Parringtonia gracilis. However, the poorly understood Dyoplax arenaceus (SMNS 4760) from the Carnian Stuttgart Formation (Lucas, Wild & Hunt, Reference Lucas, Wild and Hunt1998; Kozer & Bachmann, Reference Kozer and Bachmann2010) has been hypothesized to be a member of the group (Walker, Reference Walker1961, Reference Walker1968, Reference Walker1970; Romer, Reference Romer1966; Carroll, Reference Carroll1988) but has also been considered to be an aetosaur (Zittel, Reference Zittel1890; Huene, Reference Huene1902; McGregor, Reference McGregor1906), a protosuchid crocodyliform (Benton, Reference Benton, Fraser and Sues1994) or a sphenosuchian (Lucas, Wild & Hunt, Reference Lucas, Wild and Hunt1998; Benton & Walker, Reference Benton and Walker2002). The broad array of taxonomic assignments is largely the result of poor preservation of the anatomical details of the specimen: the specimen is preserved as a natural cast in sandstone (Fraas, Reference Fraas1867). Clark, Sues & Berman (Reference Clark, Sues and Berman2000) questioned the recent assignment to a crocodylomorph by Lucas, Wild & Hunt (Reference Lucas, Wild and Hunt1998), noting that proposed cranial similarities between Dyoplax and crocodylomorphs are impossible to verify on the cast.

The assignment by Walker (Reference Walker1961) of Dyoplax arenaceus to Erpetosuchidae was based solely upon gross similarity. Although the skeleton of Dyoplax arenaceus is largely complete, the elements that can be compared between Erpetosuchus granti, Parringtonia gracilis and Dyoplax arenaceus are limited. The body of Dyoplax arenaceus appears to be an external cast but it is not clear if the skull is an internal or external cast; if the skull is an internal cast, comparisons with the external skull cast of Erpetosuchus granti is impossible. As preserved, the skull of Dyoplax arenaceus appears to lack the mediolaterally expanded posterior portion of the maxilla that occurs in Erpetosuchus granti and Parringtonia gracilis. Furthermore, Dyoplax arenaceus lacks the ventrally deflected jugal and narrow frontal and parietal of Erpetosuchus granti (Benton & Walker, Reference Benton and Walker2002). Postcranially, little is visible in Dyoplax arenaceus. The osteoderms of Dyoplax arenaceus are superficially similar to those of Erpetosuchus granti and Parringtonia gracilis, but the common character states have a wide distribution among pseudosuchians (Nesbitt, Reference Nesbitt2011).

Dyoplax arenaceus cannot be shown to pertain to Erpetosuchidae at this time, although we cannot exclude the possibility that future discoveries and future analyses may show it to be the sister taxon of Erpetosuchus granti and Parringtonia gracilis. We refrain from including it into a phylogenetic analysis at this time, but note that the taxon can be clearly assigned to Archosauria based on the presence of an antorbital fenestra with a fully circumscribing antorbital fossa. The presence and arrangement of the dorsally placed osteoderms is suggestive of a pseudosuchian assignment (see Nesbitt, Reference Nesbitt, Liu and Li2011).