The axial skeleton of Poposaurus langstoni (Pseudosuchia: Poposauroidea) and its implications for accessory intervertebral articulation evolution in pseudosuchian archosaurs

Vertebrae (general)

The axial column of P. langstoni consists of eight presacral vertebrae, and all are well preserved and nearly complete although most are missing parts of the diapophyses, parapophyses, zygapophyses, and some are crushed slightly (Figs. 2 and 3). Each vertebra can be assigned to a presacral position within a range of two or three positions, depending on its general position among the other vertebrae preserved and based on comparisons with other archosaurs with more complete vertebral columns. For this specimen of P. langstoni, all the vertebrae were assigned general positions in the vertebral column mainly based on the position of the diapophyses, as they appear higher on the neural arch further posteriorly in the series as with P. gracilis (TTU-P 10419), Parringtonia gracilis (NMT RB426), and a number of specimens of Alligator mississippiensis (e.g., TMM M-12606).

In P. langstoni, cervical vertebrae were identified by the presence of the parapophyses on the anterior rim of the centrum located anywhere from the base of the centrum to the base of the neural arch just ventral to the location of the neurocentral suture, whereas the parapophyses of trunk vertebrae are only present on the neural arch. The parapophyses migrate dorsally from the base of the centrum posteriorly along the tetrapod vertebral column (such as in P. gracilis and A. mississippiensis). Because the diapophyses are located on the neural arch in vertebrae posterior to presacral 7 in other paracrocodylomophs (e.g., P. gracilis, P. gracilis, A. mississippiensis) we can constrain the position of the anteriormost vertebra of P. langstoni (TMM 31025-1261.5) at presacral ∼7 or 8. This vertebra also has a very slight ventral keel, which further constrains its position to presacral ∼7 or 8 because this is characteristic of the most anteriorly located cervical vertebrae in other paracrocodylomorphs (e.g., P. gracilis).

Using these criteria for assignment of position within the axial column, the known vertebrae of P. langstoni include four cervical vertebrae and four trunk vertebrae (Fig. 1). All centra are amphicoelous and slightly mediolaterally compressed. All vertebrae have clearly defined vertebral laminae, and these laminae are identified using the nomenclature of Wilson (1999) and more broadly used for early archosaur taxa (Nesbitt, 2005, 2007, 2011; Nesbitt, Liu & Li, 2010; Nesbitt et al., 2014; Weinbaum, 2013; Lautenschlager & Desojo, 2011; Parker, 2008, 2016a, 2016b). All of the trunk vertebrae of this specimen preserve one or both of the accessory articulation structures (i.e., hyposphene, hypantrum) that form the hyposphene–hypantrum articulation. None of the cervical vertebrae however have the hyposphene–hypantrum present, and this is consistent with our observations that the articulation is only present in the trunk region (presacral position 10 to the last presacral before the sacrum) of archosaur vertebral columns.

Vertebrae (cervical)

The cervical vertebra of individual “A” (TMM 31025-177) and the three cervical vertebrae of individual “B” (TMM 31025-1261.5, TMM 31025-1262, TMM 31025-1261.3) (Fig. 2) are amphicoelous, but both the posterior and anterior articular surfaces of the centrum are only slightly concave. The lateral and ventral portions of the anterior and posterior articular faces of the centra of the cervical vertebrae of individual “B” are thickened and rugose, a morphology that is more similar to P. gracilis than to previously described material from P. langstoni (Weinbaum & Hungerbühler, 2007) and the centrum of the cervical vertebra of individual “A.” The centra of these vertebrae are shorter anteroposteriorly in comparison to their dorsoventral height.

Both articular facets of the centrum of TMM 31025-1261.5 are slightly elliptical, with the long axis oriented dorsoventrally. The anterior articular facets of the centra of TMM 31025-1262, TMM 31025-1261.3, and TMM 31025-177 are circular, and the posterior articular facets of these centra are slightly elliptical with the elongated axis oriented dorsoventrally (Fig. 2). These posterior centrum facets are very similar in shape to those of the anterior cervical vertebrae of P. kirkpatricki (UCMP A269/124557, Long & Murry, 1995; TTU-P 9002, Weinbaum, 2013) in that they appear elliptical and elongated dorsoventrally. All of the cervical vertebrae possess deep laterally opening fossae on the centrum on both their left and right sides just dorsal to the parapophyses and about 1 cm ventral to the neurocentral sutures.

The two more posterior cervical vertebrae of individual “B” (TMM 31025-1262 and TMM 31025-1261.3) and the posterior cervical vertebra of individual “A” (TMM 31025-177) possess a well-preserved accessory rib articulation between the diapophysis and the parapophysis. The presence of these accessory articulations provides evidence that these vertebrae had double-headed ribs associated with them, an uncommon occurrence in pseudosuchians, reported only in this specimen, P. gracilis (TTU P-10419), and A. babbitti (MSM 4590). Each of the accessory rib articulations are dorsoventrally long and thin, projecting laterally from their respective centra. The accessory rib articulation of TMM 31025-1262 is broken at its base on the left lateral side, but it is intact on the right lateral side and protrudes 0.5 cm from where it attaches to the rest of the centrum. It is located 1 cm dorsal to the parapophyses. The accessory rib articulations of TMM 31025-1261.3 are intact on both sides, protrude 1.5 cm laterally, are located 3.2 cm from the diapophyses, and are located just dorsal to and touching the parapophyses. On the right lateral side of TMM 31025-177, the accessory rib articulation is just dorsal to and separated by 1 cm from the parapophyses. The left lateral side does not preserve this articulation; however, the paradiapophyseal lamina is clearly broken in the place where the articulation would be present. The accessory rib articulations on each cervical vertebra connect to the diapophyses through the paradiapophyseal laminae. This accessory rib articulation is also present in the posterior cervical vertebrae of P. gracilis (TTU P-10419) and A. babbitti (MSM 4590), and it is a synapomorphy of the clade Poposauroidea (Nesbitt, 2005; Weinbaum & Hungerbühler, 2007).

Vertebrae (cervical; individual “A”)

Vertebrae (cervical; individual “B”)

TMM 31025-1262

TMM 31025-1262 is a nearly complete posterior cervical vertebra, with slight breakage and minor post-depositional compression in some areas, including the prezygapophyses being pushed toward the neural spine (Figs. 2B and 2D). This specimen can be attributed to presacral 8 or 9; the parapophyses are low on the centrum, but slightly higher than those of TMM 31025-1261.5, and well separated from the diapophyses, which are located on the neural arch. TMM 31025-1262 also differs from TMM 31025-1261.5 in that it has club-like diapophyses that are more expanded and flare out at the distal end where the articular facets are located; these articular facets are circular rather than elliptical like those of TMM 31025-1261.5. The neural arch is fused to the centrum and there is some evidence of where a neurocentral suture was present on the right lateral side in the form of a small, raised ridge (Brochu, 1996). There is no keel on the ventral surface of the centrum. In anterior view, the neural canal is circular and filled with matrix, and the postzygapophyses nearly come together at the dorsal portion of the neural canal.

The neural spine is tall and mediolaterally compressed. A long depression extends dorsoventrally along the posterior surfaces of the neural spine. The neural spine is nearly identical to that of TMM 31025-1265.5. In posterior view, the spinopostzygapophyseal laminae are paired and form distinct ridges from the base to nearly the distal end of the neural spine. There are no spinoprezygapophyseal laminae present on the neural spine in anterior view. The neural spine is much wider at its base, at 3 cm dorsal to the postzygapophyses in posterior view, and tapers to about 0.5 cm mediolaterally in anterior and posterior views at its most dorsal point. It does not expand into a spine table as in the mid-trunk vertebrae of N. songeaensis (NMT RB48), the anterior and posterior trunk vertebrae of F. tenax (PVL 3850, PVL 3851), the trunk vertebrae of P. chiniquensis (UFRGS-PV-0156-T), and anterior trunk vertebrae of B. kupferzellensis (SMNS 80294). From lateral views, the neural spine remains relatively consistent in thickness mediolaterally at about 2.7 cm. The distal end of the neural spine is also rounded slightly in lateral view and is slightly taller posteriorly.

TMM 30125-1262 has distinct laminae forming thin, pronounced ridges on its neural arch; however, these laminae are more laterally expanded than the laminae of the posterior cervical of individual “B” (TMM 31025-177) and more similar to the anterior cervical vertebrae of individual “A” (TMM 31025-1261.3 and TMM 31025-1261.5). On the lateral sides of the vertebra, there are clear posterior and anterior centrodiapophyseal laminae, which connect the diapophyses to the posterior and anterior portions of the neurocentral junction, respectively (Figs. 4A and 4C). There is a clear paradiapophyseal lamina connecting the parapophyses and diapophyses, and the accessory rib articulation is located on this lamina (Figs. 4A and 4C). This paradiapophyseal lamina extends nearly straight down from the diapophysis, ∼5° from the vertical. The postzygapophyseal laminae are well defined and connect the parapophyses and the lateral aspect of the postzygapophyses at about a 45° angle dorsal to the anteroposterior horizontal. This lamina is compressed on the left lateral side, because the specimen is crushed so that the left postzygapophysis is pushed toward the diapophysis (Figs. 2B and 2D). The prezygadiapophyseal laminae are thick in lateral view and the prezygapophyses and diapophyses are ∼3 cm apart. These laminae extend dorsally from the diapophyses at right angles to the anteroposterior horizontal plane, and the prezygapophyses are dorsal to the diapophyses. The diapophyses are intact where they connect to the neural arch, and they are more robust than in TMM 31025-1261.5 and end in circular articular facets. Centroprezygapophyseal laminae extend from the ventral side of the prezygapophyses ventrally along the neural canal to the dorsal edge of the centrum. There are epipophyses on the dorsal portion of the postzygapophyses that form a rugose structure about 0.2 cm from the end of the articular surface.

Vertebrae (trunk)

The trunk vertebra of individual “A” (TMM 31025-257) and all three trunk vertebrae of individual “B” (TMM 31025-1261.4, TMM 31025-1261.1, TMM 31025-1261.2) (Fig. 3) are waisted between the articular facets of the centra. The anterior and posterior articular facets of the two anteriormost trunk vertebrae (TMM 31025-1261.4 and TMM 31025-1261.1) and the trunk vertebra of individual “A” (TMM 31025-257) appear narrow mediolaterally in lateral and ventral views. The general shape of these centra in posterior and anterior view are elliptical, with a long dorsoventrally oriented axis, which is similar to that seen in B. kupferzellensis (SMNS 80296, Gower & Schoch, 2009), P. gracilis (TTU P-10419, Weinbaum & Hungerbühler, 2007), Effigia okeeffeae (AMNH FR 30587, Nesbitt, 2007), and P. kirkpatricki (TTU P-9002, Weinbaum, 2013). The articular facets of the centrum of the most posterior trunk vertebra of individual “B” (TMM 31025-1261.2) are round and not compressed in any direction, and the anterior and posterior articular facets of the centrum appear thick (∼1 cm) in the anteroposterior direction. There are deep fossae on the centra on both the left and right lateral sides about 1 cm ventral to the position of the neurocentral sutures. There is no ridge on the midline on the ventral surfaces of any of the three trunk vertebrae of individual “B;” however there is a recognizable, but poorly developed ridge along the midline of the ventral surface of the centrum of the trunk vertebra of individual “A,” which suggests that the trunk vertebra of individual “A” is located more anteriorly in the column than the other trunk vertebrae.

All four of the trunk vertebrae preserve one or both of the accessory articulation structures (i.e., hyposphene, hypantrum) that form the hyposphene–hypantrum articulation. TMM 31025-1261.4 (Figs. 3A and 3C) is weathered so that a hyposphene could not be recognized, but a clear and deep hypantrum is visible on the anterior aspect of the neural arch. TMM 31025-1261.1 (Figs. 3B and 3D) is weathered so that a hypantrum could not be recognized, but on the posterior aspect of the neural arch there is a hyposphene preserved. TMM 31025-1261.2 (Figs. 3E and 3G) and TMM 31025-257 (Figs. 3F and 3H) preserve both hyposphenes and hypantra on their neural arches.

Vertebrae (trunk; individual “A”)

Vertebrae (trunk; individual “B”)

Rib (TMM 31025-2160)

This specimen is broken about 12 cm distally from the capitulum. The capitulum, tuberculum, and accessory rib facet are all intact. A thin ridge extends from the dorsal edge of the capitulum along the rib (Fig. 5). This rib fragment has an accessory articular facet that is located just dorsal to the tuberculum (Fig. 5). We assign it to the posterior cervical region of the axial skeleton because of the relative locations of the capitulum and tuberculum and the presence of an accessory rib facet, as a corresponding accessory rib articulation is seen on the posterior cervical vertebrae described in this paper (TMM 31025-177, TMM 31025-1262, TMM 31025-1261.3). An accessory rib facet between the capitulum and tuberculum is also present in P. gracilis (Weinbaum & Hungerbühler, 2007) and A. babbitti (Nesbitt, 2005), and it matches in position, just dorsal to the parapophysis to the accessory rib articulation structures in the posterior cervical vertebrae described herein. Weinbaum & Hungerbühler (2007) cited this accessory articulation as a synapomorphy of the clade Poposauroidea, which includes P. gracilis, P. langstoni, E. okeeffeae, and A. babbitti.

Variation in the hyposphene–hypantrum articulation in P. langstoni

The hyposphene–hypantrum articulation of the trunk vertebrae of P. langstoni presents an important opportunity to examine within-vertebral column variation because these vertebrae are preserved in three dimensions so that articulation surfaces and hyposphene–hypantrum are visible and easy to discern. Moving posteriorly along the vertebral column, the hyposphene–hypantrum articulation is present in the anteriormost trunk vertebrae of the individuals represented by presacral 11 in individual “A” (TMM 31025-1261.4) and presacral 12 of individual “B” (TMM 31025-257). The articulation is also present in presacral 12 in individual “A” (TMM 31025-1261.2). All three of the preserved hyposphenes are triangular in posterior view, and the ventralmost edges of the triangular hyposphenes are oriented horizontally with their apices pointed dorsally along the midline. The ventrolateral corners of the triangular hyposphenes curve slightly ventrally around the neural canal. The hyposphenes (TMM 31025-1261.1, TMM 31025-1261.2, TMM 31025-257) appear to be more dorsoventrally compressed (i.e., smaller height to width ratio) in individual “B” than in individual “A.” This material described herein includes the twelfth presacral from both individuals (TMM 31025-1261.1, TMM 31025-257), so it is likely that this variation is not based on location in the vertebral column. The hypantra (TMM 31025-1261.4, TMM 31025-1261.2, TMM 31025-257) reflect a complementary shape of the hyposphene structures on their respective vertebrae. The horizontal distance between the prezygapophyses is narrower across the structure at the dorsal most edge than at the ventralmost extent, creating a triangular space for articulation. In summary, the hyposphene–hypantrum articulation appears triangular in all vertebrae described herein and therefore does not vary in general shape along the vertebral column of one individual of P. langstoni, but the two recognized individuals of P. langstoni have hyposphene structures that are markedly different in terms of height to width ratio (Table 1).

Definition of the hyposphene–hypantrum articulation

The hyposphene–hypantrum articulation appears in many clades within Archosauria and has been defined simply as a vertical wall of bone ventral to the postzygapophyses and a notch between the prezygapophyses (Rauhut, 2003; Apesteguia, 2005; Hibbard & Williston, 1971). Those previous studies included definitions of the hyposphene–hypantrum articulation that were based on the vertebral morphology of saurischian dinosaurs, but they are not comprehensive in incorporating the variation in shape in dinosaurs and pseudosuchians. Furthermore, a number of pseudosuchians also have been reported to bear the hyposphene–hypantrum articulation without an explanation of why it is homologous with those structures in saurischian dinosaurs (Bonaparte, 1981; Weinbaum & Hungerbühler, 2007; Peyer et al., 2008; Gower & Schoch, 2009; Lautenschlager & Desojo, 2011; Weinbaum, 2013). This is important to clarify because the hyposphene–hypantrum has been cited as a synapomorphy of Saurischia (Gauthier, 1986) and of at least one clade of pseudosuchians (Ticinosuchus + Paracrocodylomorpha, Nesbitt, 2011). Here we apply an explicit definition of this articulation to facilitate identification and future studies of intervertebral articulation in archosaurs.

We define the hyposphene–hypantrum articulation as a bony projection, the hyposphene, on the posterior portion of the vertebra that fits into a complementary space, the hypantrum, on the anterior portion of the subsequent vertebra within a vertebral series (Fig. 6). Specifically, the hyposphene is located ventral to the articular surfaces of the postzygapophyses and is connected to these articular surfaces where they converge. It is located dorsal to the neural canal and in posterior view is symmetrical or nearly symmetrical across the midline. There is a distinct angle change (typically between ∼45° and 90°) between the articular surfaces of the postzygapophyses and the lateral surfaces of the hyposphene. The hyposphene projection must be a comparable shape and size to that of the hypantrum space of the subsequent vertebrae because these structures articulate precisely. The shapes in lateral view of the hyposphenes may appear as circular, square, dorsoventrally elongate rectangular, triangular, diamond, and quadrilateral, and the most common shapes found in pseudosuchian archosaurs are triangles and dorsoventrally elongate rectangles (Fig. 7). Though these shapes can vary between taxa, they also can vary within an individual. P. langstoni does not show drastic variation in hyposphene shape, but the proportions do vary along the column and between our two individuals (i.e., ratio of height to width increasing posteriorly in individual “B” and greater ratio of height to width in individual “A” than in individual “B”) (Table 1). As long as these projections extend posteriorly from the neural arch, their lateral surfaces are confluent with the articular surfaces of the postzygapophyses, and there is a distinct angle change between the articular surfaces of the postzygapophyses and the projection, any of these aforementioned shapes may be considered hyposphenes. Additionally, it is important to note that these structures are currently only known from trunk vertebrae posterior to the first nine presacral vertebrae.

Descriptions of a vertebra with the hyposphene–hypantrum should describe the shape that the hyposphene appears to be in posterior view. In vertebrae without the hyposphene, the articular surfaces of the postzygapophyses may converge but not form a ventrally elongated bony projection, as in R. callenderi (PEFO 34561) (Fig. 7K), Erythrosuchus africanus (SAM 905) (Fig. 7J), P. gracilis (NMT RB426), or N. songeaensis (NMT RB48), or they may not converge at all and appear well separated, as in Deinosuchus riograndensis (TMM 43632-1) (Fig. 7I) or A. mississippiensis (Romer, 1956, Fig. 130).

The lateral surfaces of the hyposphene articulate with the medial surfaces of the hypantrum, which is located between and ventral to the prezygapophyses and dorsal to the neural canal (Fig. 6). The articular surfaces of the prezygapophyses continue ventrally from their medial surfaces to form the articular surfaces of the hyposphene. As such, there must be a distinct angle change (typically between ∼45° and 90°) between these articular surfaces. In dorsal view, a hypantrum appears as a gap framed by parallel to sub-parallel medial surfaces of the prezygapophyses, which contact the neural arch just dorsal to the neural canal. In vertebrae without a hypantrum articulation (e.g., R. callenderi, PEFO 34561; P. gracilis, NMT RB426; E. okeeffeae, AMNH FR 30587; D. riograndensis, TMM 43632-1; A. mississippiensis, Romer, 1956), the medial edges of the prezygapophyses will converge in a “v” shape and will not appear parallel in dorsal view.

Accessory intervertebral articulations within Pseudosuchia

The presence of the hyposphene–hypantrum articulation in P. langstoni allows for comparisons with the presacral vertebrae of other pseudosuchian archosaurs. Other closely related members of Poposauroidea, P. gracilis—the sister taxon to P. langstoni—and the smaller E. okeeffeae, have also been reported to have a hyposphene–hypantrum between trunk vertebrae (Weinbaum & Hungerbühler, 2007; Nesbitt, 2007).

The presacral vertebrae of the poposauroid E. okeeffeae (AMNH FR 30587, Nesbitt, 2007, Fig. 30) are only known from four semi-articulated vertebrae, and the anterior aspect of the neural arch is only visible on the anteriormost vertebra of the articulated series. There is not a clearly defined space between the prezygapophyses in this vertebra to satisfy our definition of a hypantrum. There is a slight gap between the medial aspects of the prezygapophyses, the articular surfaces of which are oriented horizontally, however these articular surfaces do not extend ventrally to form a hypantrum in accordance with our definition. No posterior surface of any vertebra of AMNH FR 30587 is clearly visible and intact, so we cannot conclusively state whether a hyposphene was present in the trunk vertebrae of known material of E. okeeffeae. Additionally, because we do not see a hypantrum present in this specimen of E. okeeffeae, we conclude that this taxon probably did not possess the hyposphene–hypantrum. Furthermore, there are no specimens of Shuvosaurus inexpectatus (e.g., TTU P-9001), the current sister taxon of E. okeeffeae (Nesbitt & Norell, 2006; Nesbitt, 2007, 2011), that preserve a neural arch where we could evaluate whether the hyposphene–hypantrum articulation was present or absent. The other known shuvosaurid, S. longicervix (PVSJ 85), is poorly preserved (Alcober & Parrish, 1997), and therefore it is difficult to determine definitively whether a hyposphene–hypantrum was present in that taxon as well. The sail-backed poposauroid Lotosaurus adentus (IVPP V 4880, IVPP V 48013) has not been reported to have the hyposphene–hypantrum articulation.

Arizonasaurus babbitti is one of the oldest members of Poposauroidea, outside of L. adentus + Shuvosauridae, and it has been reported to have the hyposphene–hypantrum articulation (MSM 4590, Nesbitt, 2005, 2007). The hyposphene structure of the trunk vertebra of A. babbitti (MSM 4590) is rectangular with the long axis oriented dorsoventrally in posterior view and located ventral to the postzygapophyses and dorsal to the neural canal (Nesbitt, 2005, Fig. 19). It does not have the triangular shape as in the trunk vertebrae of P. langstoni. The sister taxon of A. babbitti is the earliest diverging poposauroid, Xilousuchus sapingensis (IVPP V6026), and it has been cited as having the hyposphene–hypantrum as well (Nesbitt, Liu & Li, 2010). The ninth presacral vertebra figured by Nesbitt, Liu & Li (2010 Fig. 8) has a clearly defined hyposphene, and it is square in posterior view. A hypantrum between the prezygapophyses is clearly seen in anterior view and is of comparable size and shape to the hyposphene of the same vertebra (Nesbitt, Liu & Li, 2010, Fig. 8). The identification of this vertebra as the ninth presacral (Nesbitt, Liu & Li, 2010) may be incorrect given that there are no other taxa with a hyposphene in the first nine presacrals. Alternatively, the presence of a hyposphene in this position may be autapomorphic for this taxon.

Several Triassic loricatan paracrocodylomorphs and their closest relatives possess the hyposphene–hypantrum articulation, and this structure varies in shape and structure between taxa and even within an individual. Stagonosuchus nyassicus (GPIT/RE/3831, von Huene, 1938; Gebauer, 2004; Lautenschlager & Desojo, 2011) is reported to have the hyposphene–hypantrum in the trunk vertebrae. The anterior trunk vertebra of S. nyassicus (GPIT/RE/3831-9, von Huene, 1938; Gebauer, 2004; Lautenschlager & Desojo, 2011, Fig. 7) has a clearly defined hyposphene that is rectangular, but it is only slightly elongated dorsoventrally. The posterior trunk vertebra of S. nyassicus (GPIT/RE/3831-14, von Huene, 1938; Gebauer, 2004; Lautenschlager & Desojo, 2011, Fig. 7) has a hyposphene that is narrower mediolaterally and more elongated dorsoventrally than the hyposphene of its anterior trunk vertebra. This shape is distinctly different from the triangular hyposphenes of P. langstoni (Fig. 3). In anterior view, there is a clear space between the prezygapophyses to form a hypantrum in S. nyassicus. T. ferox is a close relative to S. nyassicus (Lautenschlager & Desojo, 2011), but the presence of the hyposphene–hypantrum articulation in that taxon is ambiguous because the known material (PIZ T 2817) is preserved in a flattened slab, and the vertebral column is mostly articulated so that the anterior and posterior aspects of the vertebrae cannot clearly be seen.

Fasolasuchus tenax (PVL 3850, Bonaparte, 1981, Figs. 10 and 11), B. kupferzellensis (SMNS 80296, Gower & Schoch, 2009, Fig. 2), P. chiniquensis (UFRGS-PV-0156-T, Azevedo, 1991, Fig. 11), and Saurosuchus galilei (PVSJ 32, Trotteyn, Desojo & Alcober, 2011, Fig. 7) have hyposphene–hypantrum articulations present in their trunk vertebrae. F. tenax has a hyposphene structure that appears rectangular in posterior view in the anterior trunk vertebra figured by Bonaparte (1981, Fig. 10). In the posterior trunk vertebra of F. tenax (Fig. 7B), the hyposphene structure is triangular in posterior view with the midline apex directed dorsally and the ventrolateral corners curved slightly ventrally around the neural canal, similar to that of P. langstoni and P. kirkpatricki (TTU P-9002). The trunk vertebra of B. kupferzellensis figured in lateral view by Gower & Schoch (2009, Fig. 2) (SMNS 80296) has a clearly defined hyposphene that is rectangular and elongated dorsoventrally in posterior view (Fig. 7C); in anterior view there is a clear space between the prezygapophyses to form a hypantrum in this specimen. The posterior trunk vertebra of P. chiniquensis that was figured in posterior view by Azevedo (1991, Fig. 11) has a well-defined hyposphene that is rectangular with the long axis oriented dorsoventrally in posterior view. The shape of this hyposphene most closely resembles the hyposphene shapes in B. kupferzellensis (SMNS 80296) (Fig. 7C) and P. kirkpatricki (TTU P-9002). S. galilei (PVSJ 32) has a hyposphene–hypantrum articulation present on at least one of its posterior trunk vertebrae (Trotteyn, Desojo & Alcober, 2011), but it is difficult to tell the shape because of poor preservation of the specimen.

A large member of the paracrocodylomorph clade Rauisuchidae, P. kirkpatricki has been reported to have the hyposphene–hypantrum articulation present on its middle and posterior trunk vertebrae (TTU P-9002, Weinbaum, 2013). The figured anterior trunk vertebra and posterior trunk vertebra of TTU P-9002 have clearly defined hyposphene structures that appear triangular in posterior view, with the ventralmost aspect as a horizontal surface and the dorsalmost aspect as a midline apex (TTU P-9002, Weinbaum, 2013, Figs. 4 and 5). The ventrolateral corners of the triangular hyposphene also slightly curve around the neural canal at its dorsal margin. This shape is nearly identical to those hyposphenes of P. langstoni in the anterior trunk vertebrae TMM-31025-1261.1 and TMM-31025-1261.2. The hypantrum of the anterior portion of the neural arch of the anterior (Weinbaum, 2013, Fig. 4) and posterior (Weinbaum, 2013, Fig. 5) trunk vertebrae of P. kirkpatricki (TTU P-9002) also appear triangular in anterior view with a shape and size complementary to the hyposphene of the same vertebra. The hyposphene structure of the mid-trunk vertebra of P. kirkpatricki (TTU P-9002, Weinbaum, 2013, Fig. 4) appears rectangular in posterior view with the long axis oriented dorsoventrally. The one known trunk vertebra of P. alisonae (UNC 15575) has clear hyposphene and hypantrum structures (Fig. 7A). The hyposphene appears triangular in posterior view, similar to those of the anterior and posterior trunk vertebrae of P. kirkpatricki (TTU P-9002), with the midline apex directed dorsally and the ventrolateral corners curved slightly around the dorsal edge of the neural arch (Peyer et al., 2008, Fig. 4). The hypantrum of P. alisonae is of complementary shape and size to the hyposphene.

Just outside of Paracrocodylomorpha, N. songeaensis (NMT RB48, Nesbitt et al., 2014) was reported to possess the hyposphene–hypantrum articulation on its mid-trunk vertebrae. Evidence for a hypantrum was cited as a small gap between the anterior portion of the prezygapophyses, and the hyposphene was cited as a thin, ventrally directed lamina of bone between the postzygapophyses. However, this proposed hyposphene was not clearly defined, and appears simply as the postzygapophyses converging at the midline, without an extension of a bony process ventral to them (NMT RB48, Nesbitt et al., 2014, Fig. 4). There is a slight gap between the medial aspects of the prezygapophyses; however, articular surfaces do not extend ventrally from the prezygapophyses to form a hypantrum in accordance with our refined definition. This morphology in N. songeaensis is similar to that in E. okeeffeae, which we have concluded does not have the hyposphene–hypantrum articulation.

Hyposphene–hypantrum articulations have been reported in some members of Aetosauria, a group of quadrupeds with an extensive osteoderm carapace (Parker, 2008; Desojo, Ezcurra & Kischlat, 2012; Parker, 2016b). The largest known aetosaur, D. spurensis (MNA V9300, Parker, 2008), has hyposphene–hypantrum articulations between its trunk vertebrae. The hyposphene structures of the anterior and mid-trunk vertebrae are rectangular in posterior view and more elongated dorsoventrally relative to the hyposphene structures of the posterior trunk vertebrae (Fig. 7D). The rectangular hyposphenes of the anterior and mid-trunk vertebrae of D. spurensis appear more similar to those structures of B. kupferzellensis, F. tenax, and P. chiniquensis than to taxa with triangular hyposphenes (e.g., P. langstoni). The posterior trunk vertebrae of D. spurensis have hyposphenes that have a width and height nearly equal to each other. These hyposphenes also curve around the neural canal and appear triangular in posterior view (MNA V9300, Parker, 2008, Fig. 9). These triangular hyposphenes of the posterior trunk vertebrae appear more similar to the structures of the anterior trunk vertebrae of P. langstoni and P. kirkpatricki than to taxa with rectangular hyposphenes.

Several other aetosaurs besides D. spurensis have been reported to have the hyposphene–hypantrum articulation including S. deltatylus (PEFO 34045, Parker, 2016b), possibly Longosuchus meadei (TMM 31100-448, TMM 31100-452), and Aetobarbakinoides brasiliensis (CPE 2 168, Desojo, Ezcurra & Kischlat, 2012). However, a hyposphene structure is difficult to infer on most of the vertebrae of A. brasiliensis (CPE 2186, Desojo, Ezcurra & Kischlat, 2012, Fig. 5) (Fig. 7F) because much of the holotype is articulated or broken on the neural arch. Desojo, Ezcurra & Kischlat (2012) described A. brasiliensis as having hyposphenes that are “Y-shaped” where the elongated hyposphene structure forms the “trunk” of the “Y” and the postzygapophyses connected and dorsal to it form the top two dorsolaterally projecting “branches” of the “Y” in posterior view. On our close inspection of the well-preserved and non-articulated vertebrae of A. brasiliensis, there does not appear to be a distinct angle change between the articular surfaces of the postzygapophyses to the bony projection ventral to and between them (Fig. 7F). Therefore we conclude that A. brasiliensis does not have a hyposphene–hypantrum according to our definition. Parker (2016b) did not consider the aetosaur S. deltatylus to have a true hyposphene–hypantrum, however the “ventral bar” structure on the posterior aspect of its vertebrae (Parker, 2016b, Fig. 12) does fit our definition of a true hyposphene. S. deltatylus has two different shape varieties of hyposphene structures present in its trunk vertebrae: one that is circular in posterior view and tapers to a point distally from the body of the vertebra (PEFO 34045 FF-22, Parker, 2016b, Fig. 13) (Fig. 7E) and a second that is triangular with a dorsally oriented point extending along the dorsoventral margin of the neural arch (PEFO 34045 FF-51, Parker, 2016b, Fig. 12). S. deltatylus also preserves hypantrum structures of complementary size and shapes on the anterior aspects of its neural arches. Scutarx robertsoni (Walker, 1961, Fig. 7) has a “ventral bar” shaped hyposphene in an anterior dorsal vertebra (R 4799) that appears very similar to that of S. deltatylus. S. robertsoni also preserves complementary hypantrum structures on the anterior aspects of the anterior dorsal (R 4799) and posterior cervical (E.M. 30R) vertebrae (Walker, 1961, Fig. 7).

The aetosaur L. meadei (TMM 31100-448, TMM 31100-452) may have a hyposphene–hypantrum articulation in its trunk vertebrae, but this is ambiguous because of the poor preservation of known material. The trunk vertebra of L. meadei, TMM 31100-448 appears to have a hyposphene that is square in posterior view; however, the postzygapophyses and surrounding portions of the neural arch are broken and compressed, making it difficult to definitively discern a hyposphene structure according to our definition (Fig. 7G). Neither Typothorax coccinarium (NMMNH P-12964, Heckert et al., 2010, Fig. 4A) nor Paratypothorax andressorum (PEFO 3004, Lucas, Heckert & Rinehart, 2006; Hunt & Lucas, 1992, Fig. 4) was reported to have the hyposphene–hypantrum, and we concur with this through our observations of the published figures of those materials and through direct observation. The small aetosaur Coahomasuchus kahleorum (TMM 31100-437, Parker, 2016a) does not appear to have the hyposphene–hypantrum in its trunk vertebrae; however, the known vertebral material is mostly articulated, poorly preserved, and portions of the neural arches are broken or obscured (TMM 31100-437).

Although it is present in a few aetosaurs, the hyposphene–hypantrum articulation is not present in the sister taxon to Aetosauria, R. callenderi (PEFO 34561) (Fig. 7K). It is also not present in P. gracilis (NMT RB426), the sister taxon of R. callenderi (PEFO 34561) + Aetosauria (Nesbitt et al., in press). The postzygapophyses of the vertebrae of P. gracilis converge at the midline but they do not extend into a projection ventrally. The articular surfaces of the postzygapophyses of the few known anterior trunk vertebrae of R. callenderi that are intact and not obscured by sediment (PEFO 34561-DVf, PEFO 34561-DVh) do converge; however, they do not have a bony and dorsoventrally elongated process that could definitively be a hyposphene. The cervical vertebrae (PEFO 34561-CV) of this specimen definitively do not have the hyposphene–hypantrum articulation. Therefore, both R. callenderi and P. gracilis are small pseudosuchian archosaurs that lack the hyposphene–hypantrum.

Relationship of the hyposphene–hypantrum articulation with body size

The articulation structures of the hyposphene–hypantrum are widespread in the clade Pseudosuchia, but are noticeably absent in the group Crocodylomorpha (e.g., Sphenosuchus acutus, UCMP 129740; D. riograndensis, TMM 43632-1; A. mississippiensis, TMM M-12606) (Fig. 8), which includes the only living pseudosuchians, the crocodylians. Most early members of Crocodylomorpha are markedly smaller (i.e., shorter femoral lengths) (Turner & Nesbitt, 2013) than their paracrocodylomorph relatives (Fig. 9), such as the poposaurids (e.g., P. langstoni; P. gracilis, Weinbaum & Hungerbühler, 2007, TTU P-10419), the rauisuchid P. kirkpatricki (TTU-P 9002, Weinbaum, 2013), other pseudosuchians (i.e., P. chiniquensis, UFRGS-PV-0156-T, Azevedo, 1991; F. tenax, PVL 3850, Bonaparte, 1981; S. nyassicus, GPIT/RE/3831, Lautenschlager & Desojo, 2011; Mandasuchus tanyauchen, NHMUK PV R6792, Butler et al., in press), and the largest-known aetosaur D. spurensis (MNA V9300, Parker, 2008). Because the hyposphene–hypantrum articulation is in many large-bodied taxa but not in smaller ones (Fig. 9), we explored this potential correlation with body size by assigning each taxon we examined (i.e., each taxon included in our phylogenetic tree, Fig. 8) to either presence or absence of the hyposphene–hypantrum (Table 2), and then plotted these two groups by their femoral length as an estimate for body size (see methods in Fariña, Vizcaíno & Bargo, 1998; Christiansen & Fariña, 2004; Farlow et al., 2005; Carrano, 2006; Turner & Nesbitt, 2013). With these data, we found a relationship between larger body sizes and the presence of the hyposphene–hypantrum articulation: taxa with femora length >300 mm typically have this extra articulation (Fig. 9). While our data support this correlation, the relationship is solely visual at present and deserves rigorous statistical testing, which will be done in future studies.

An exception to this body size relationship is phytosaurs, which are either the earliest diverging pseudosuchians (Gauthier, 1986; Benton & Clark, 1988; Sereno, 1991; Benton, 1999; Ezcurra, 2016) or the sister group of Archosauria (Nesbitt, 2011). All taxa in this clade that we observed lack the hyposphene–hypantrum articulation but some have femoral lengths greater than many pseudosuchians that possess the articulation (e.g., Smilosuchus gregorii, 545 mm; Machaeroprosopus pristinus, 444 mm) (Fig. 9). The absence of the hyposphene–hypantrum in phytosaurs may be related to their inferred semi-aquatic ecology (Parrish & Padian, 1986), which is divergent from the terrestrial ecology of other early pseudosuchians. This could mean that femoral length cannot explain the presence or absence of the hyposphene–hypantrum in aquatic forms, and this may also explain the absence of the structure in the transition to a more aquatic ecology within Crocodylomorpha.

In summary, our observations of pseudosuchian archosaurs both with and without the hyposphene–hypantrum articulation showed a clear relationship between larger body sizes and the presence of the hyposphene–hypantrum articulation in trunk vertebrae across pseudosuchian archosaurs. Because of this close fit, this feature may be controlled by increases or decreases in body size and not strictly by inheritance. If so, the presence or absence of hyposphene–hypantrum would have limited use as a character in phylogenetic analyses of archosaurs (Gauthier, 1986; Nesbitt, 2011) and the character should be applied with caution when used to assign isolated vertebrae to pseudosuchian clades.