3.a. Atlas–axis complex

The atlas–axis complex of SMU SMP 69120 (Fig. 2a–e) is very well preserved and shows complete co-ossification of its component elements without trace of sutures (consistent with osteological maturity, see Welles, Reference Welles1949; Brown, Reference Brown1981). The conjoined atlas intercentrum – axis centrum is cylindrical and distinctly longer than high (Fig. 2c, Table 1, an ontogenetic feature sensu Brown, Reference Brown1981), as illustrated in a number of other elasmosaurids (e.g. Welles, Reference Welles1943, p. 239, pl. 22, fig. a; Wiffen & Moisley, Reference Wiffen and Moisley1986, p. 212, fig. 4; Sachs, Reference Sachs2005a , p. 100, fig. 4A, B; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012, p. 561, fig. 4A, B, D; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014, p. 112, fig. 10A), and the atlantal cotyle is deeply concave (Fig. 2a). The cotylar rim is surrounded by a thin edge that is damaged along its right ventrolateral margin; its dorsal midline is incised by a tapered notch (Fig. 2a, d). Ventrally, the atlas intercentrum contributes to a prominent hypophyseal ridge (Fig. 2c, e) similar to that reported in Elasmosaurus platyurus (Sachs, Reference Sachs2005a ), Eromangasaurus australis (Kear, Reference Kear2005a ) and Albertonectes vanderveldei (Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012). Its proximal extremity forms a ‘lappet-like’ projection; distally this is expanded into a prominent flattened tubercle (15.7/8 mm in maximum length/width: Fig. 2e), which is elliptical in outline and situated beneath the atlas intercentrum (17 mm from the craniad edge of the atlantal contyle). Noticeable transverse expansion of the craniad hypophyseal ridge has been recorded in A. vanderveldei (Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012, p. 561, fig. 4D), and a hypophyseal tubercle is also known in Abyssosaurus nataliae Berezin, Reference Berezin2011 (Berezin, Reference Berezin2011, p. 651, fig. 1d). Caudally, the hypophyseal ridge widens and merges with the articular face of the axis centrum, whose ventral surfaces are deeply excavated adjacent to the ventrolaterally oriented axial rib heads. The atlas ribs are broken off but their elongate oval bases (20.8/8.4 mm in maximum length/height) are still visible and situated at the approximate mid-section of the atlas–axis complex. Only the left axial rib is complete (Fig. 2c, e) and directly abuts the atlas rib remnant. In lateral view, the shaft of the axial rib tapers caudoventrally but does not project beyond the articular face of the axis centrum (Fig. 2c); this is in marked contrast to the elongate and backswept atlas–axis ribs of Thalassomedon haningtoni (Welles, Reference Welles1943, p. 239, pl. 22a), Libonectes atlasense Buchy, Reference Buchy2006 (Buchy, Reference Buchy2006, p. 24, fig. 6b), A. vanderveldei (Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012, p. 561, fig. 4D) and Aristonectes quiriquinensis Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar & Quinzio-Sinn, Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014 (Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014, p. 112, fig. 10A). The dorsolateral side of the axial rib in SMU SMP 69120 is distinctly concave, and its craniad surface is rugose, presumably for overlapping contact from the atlas rib as in other elasmosaurids (see Chatterjee & Small, Reference Chatterjee, Small and Crame1989; Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003; Kear, Reference Kear2005a ; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014).

Table 1. Measurements (mm) of the atlas–axis complex in SMU SMP 69120, the holotype of Libonectes morgani

Figure 2. Atlas–axis complex of the Libonectes morgani holotype SMU SMP 69120 shown in (a) cranial, (b) caudal, (c) lateral, (d) dorsal and (e) ventral views (with labelled graphic restoration shown on right). Abbreviations: ac – atlantal cotyle; afa – articular facet of axis; ar – axis rib; atn – atlas neural arch; axc – axis centrum; axn – axis neural arch; ba – base of atlas rib; dn – dorsal notch in atlantal cup; hr – hypophyseal ridge; lf – lateral foramen; nc – neural canal; ns – neural spine; of – oval facet in hypophyseal ridge; poz – postzygapophysis; prz – prezygapophysis; ra – rugose area of neural spine; sl – sagittal lip. Scale bar equals 30 mm.

The articular face of the axis centrum is wider than high (Table 1) and surrounded by a thickened convex rim (Fig. 2b). Its surface is partly obscured by matrix but was clearly shallowly concave with a slight notch in its dorsal margin.

The neural canal is arched and wider cranially than caudally. The cranial end is sub-circular in outline but becomes triangular caudally (Fig. 2a, b). The floor of the neural canal is perforated by a single elongate foramen near its caudal end. The atlas neural arch bases are craniocaudally shorter than those of the axis (20.9 mm versus 26 mm), and their contact is perforated by a large foramen (similar to that depicted in Aristonectes parvidens Cabrera, Reference Cabrera1941, E. platyurus and L. atlasense, see Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003, p. 108, fig. 2C; Sachs, Reference Sachs2005a , p. 100, fig. 4A; Buchy, Reference Buchy2006, p. 24, fig. 6b; Fig. 2c this paper). A pronounced lateral ledge runs from the dorsocaudal edge of the atlas neural arch to the axis neural spine, and merges with the bone surface 22 mm from the top of the lateral foramen. The bone surface between this ledge and neural spine in gently dished.

The atlantal prezygapophyses (Fig. 2a) are incomplete. Likewise, the right axial postzygapophysis is damaged, but the preserved left postzygapophysis is gently dorsocaudally oriented in lateral view and has an elliptical outline with a weakly concave articular surface when observed in caudoventral aspect (Fig. 2b). The postzygapophyses protrude beyond the caudal edge of the axis centrum for their entire length (contrary to Buchy, Reference Buchy2006, p. 7; see Fig. 2c) and are separated by a deep medial excavation that is buttressed laterally against broad ledges converging towards the neural spine (Fig. 2b). The neural spine itself is dorsocaudally inclined at ~ 35° relative to the long axis of the centrum. The craniad side of the neural spine forms a sharp edge, and is transversely expanded dorsocaudally such that its apex expands into a rugose platform that is transversely broadest caudally (Fig. 2d).

3.b. Postaxial cervical vertebrae

Welles (Reference Welles1949, p. 15) reported an articulated series of 62 cervical vertebrae in SMU SMP 69120, with the 61^st situated in the clavicular region, and the disarticulated 62^nd lying on its side within the partly prepared block containing the pectoral girdle. Welles (Reference Welles1949) assumed that the 62^nd vertebra was the last cervical, but it could have been the first pectoral (Welles, Reference Welles1962). Welles (Reference Welles1949. pp. 15, 18) also calculated a maximum length for the neck at 5.618 m (‘18 feet 5 inches’). Presently, though, remnants of only 50 cervicals (including the atlas–axis complex) are preserved (this concurs with the count of G. W. Storrs, unpub. M.Sc. thesis, Univ. of Texas, 1981, p. 101). Some of these bear fused fragments of the neural spines and cervical ribs; unfortunately, though, most of these components were broken off and lost during the initial excavation (Welles, Reference Welles1949, p. 15). Many vertebrae, especially from the craniad section of the neck, are likewise incomplete. However, their proportions clearly increased towards the caudal end of the column (centrum measurements in Welles, Reference Welles1949, pp. 16–17, table 1), with the cranial-most cervicals being wider than long or high, and their length marginally greater than their height (Fig. 3 a–c; Welles, Reference Welles1949, p. 16, table 1). This proportional trend changes from cervicals 26–36, where the maximum length increases substantially and exceeds both the height and width; from cervical 36 onwards the centra once more become broader (Fig. 3i–t; Welles, Reference Welles1949 pp. 16–17, table 1). O’Keefe & Hiller (Reference O’Keefe and Hiller2006) established that cervical centrum proportions in elasmosaurids were subject to extreme variability, although a vague pattern of proportional length increase throughout the mid-section of the neck was noticeable in many taxa. They also extrapolated that the complete caudad cervical series of SMU SMP 69120 would have had dimensions markedly dissimilar to the extremely long-necked morphotypes Elasmosaurus platyurus and Styxosaurus snowii (Williston, Reference Williston1890) (the latter following synonymy of Carpenter, Reference Carpenter1999).

Figure 3. Selected cervical vertebrae of the Libonectes morgani holotype SMU SMP 69120. Labelled graphical restoration shown in (1) lateral, (2) articular, (3) ventral and (4) dorsal views. Photographs of cervical 4 in (a) lateral, (b) cranial, (c) ventral and (d) dorsal views; cervical 12 in (e) lateral, (f) cranial, (g) ventral and (h) dorsal views; cervical 20 in (i) lateral, (j) cranial, (k) ventral and (l) dorsal views; cervical 35 in (m) lateral, (n) cranial, (o) ventral and (p) dorsal views; and cervical 51 in (q) lateral, (r) cranial, (s) ventral and (t) dorsal views. Abbreviations: ck – central keel; cr – cervical rib; fs – foramen subcentrale; llr – lateral longitudinal ridge; nc – neural canal; ns – neural spine; poz – postzygapophysis; prz – prezygapophysis; tr – thickened rim of articular facet; vn – ventral notch in articular facet. Scale bars equal 30 mm.

The articular faces of the cervical centra in SMU SMP 69120 have thickened convex rims. These are pronounced in the cranial cervicals (Fig. 3b, f, j), but become less distinct caudally and are lost in the caudal-most cervicals. Brown (Reference Brown1981) considered this feature indicative of osteological maturity amongst plesiosaurians (e.g. ANSP 10081, the ‘adult’ type specimen of E. platyurus, see Sachs, Reference Sachs2005a ; Sachs, Kear & Everhart, Reference Sachs, Kear and Everhart2013). SMU SMP 69120 also displays distinct amphicoely throughout the cervical column, up until cervical 43 where it is reduced, and by cervical 50 the vertebrae become platycoelous. The presence of platycoelous craniad cervicals has long been considered a diagnostic trait of Elasmosauridae (e.g. Andrews, Reference Andrews1910; Welles, Reference Welles1943, Reference Welles1962; Persson, Reference Persson1963; Brown, Reference Brown1981, Reference Brown1993; Bardet, Godefroit & Sciau, Reference Bardet, Godefroit and Sciau1999; O’Keefe, Reference O’Keefe2001; T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Großmann, Reference Großmann2007; Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ; Ketchum & Benson, Reference Ketchum and Benson2010; Vincent et al. Reference Vincent, Bardet, Pereda Suberbiola, Bouya, Amaghzaz and Meslouh2011; Smith, Araújo & Mateus, Reference Smith, Araújo and Mateus2012; Otero, Soto-Acuña & Rubilar-Rogers, Reference Otero, Soto-Acuña and Rubilar-Rogers2012). Nevertheless, amphicoely is documented in the craniad centra of even the most classic elasmosaurid taxa including E. platyurus (Sachs, Reference Sachs2005a ), Hydrotherosaurus alexandrae and Callawayasaurus colombiensis (Welles, Reference Welles1962), as well as in Albertonectes vanderveldei (Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012), which is also known from a complete cervical series.

A shallow notch is evident in the preserved ventral margin of both the cranial and caudal articular faces of all the cervicals in SMU SMP 69120 (Fig. 3f, j, n, r). A more pronounced expression of this trait is often evident in Late Cretaceous elasmosaurids (e.g. Welles, Reference Welles1943; Cruickshank & Fordyce, Reference Cruickshank and Fordyce2002; Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003; Hiller et al. Reference Hiller, Mannering, Jones and Cruickshank2005; Sachs, Reference Sachs2005a ; Sato, Hasegawa & Manabe, Reference Sato, Hasegawa and Manabe2006; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012; Hiller, O’Gorman & Otero, Reference Hiller, O’Gorman and Otero2014; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014), but has also been observed in basal plesiosauroids (e.g. Occitanosaurus tournemirensis: see Bardet, Godefroit & Sciau, Reference Bardet, Godefroit and Sciau1999). In contrast, Early Cretaceous elasmosaurid taxa including C. colombiensis (Welles, Reference Welles1962), Eromangasaurus australis (Kear, Reference Kear2005a ; Sachs, Reference Sachs2005b ), as well as other coeval indeterminate specimens (e.g. Kear, Reference Kear2002, Reference Kear2005b , Reference Kear2006; Sachs, Reference Sachs2004) tend to lack this feature. A pronounced longitudinal ridge is developed in between, and in contact with, the articular surface rims in most of the craniad cervicals of SMU SMP 69120 (Fig. 3a, e, i, m). It is not visible in the incomplete third cervical, and is only weakly expressed in the fourth, but is otherwise expressed on every centrum up until cervical 50. In the more craniad cervicals the longitudinal ridge is situated slightly dorsal to the lateral midline. However, by cervical 43 it progressively migrates down the lateral side of the centrum towards the rib facet and is eventually placed immediately dorsal to the rib facet on cervical 49, where it also loses contact with the articular surface rims (Fig. 4c). The presence of a longitudinal ridge is a phylogenetically defined synapomorphy for Elasmosauridae (Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ; Ketchum & Benson, Reference Ketchum and Benson2010), although it is barely developed in some taxa (e.g. Aristonectes parvidens and A. quiriquinensis, see Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003; O’Gorman, Gasparini & Salgado, Reference O’Gorman, Gasparini and Salgado2013; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014; caudad cervicals of Futabasaurus suzukii Sato, Hasegawa & Manabe, Reference Sato, Hasegawa and Manabe2006), apparently absent in Tuarangisaurus keyesi (Wiffen & Moisley, Reference Wiffen and Moisley1986) and otherwise occurs widely in disparate long-necked plesiosaurians including Seeleyosaurus guilelmiimperatoris (Dames, Reference Dames1895) (Fraas, Reference Fraas1910), Muraenosaurus leedsii (Andrews, Reference Andrews1910; Brown, Reference Brown1981), O. tournemirensis (Bardet, Godefroit & Sciau, Reference Bardet, Godefroit and Sciau1999) and Spitrasaurus wensaasi Knutsen, Druckenmiller & Hurum, Reference Knutsen, Druckenmiller and Hurum2012 (Knutsen, Druckenmiller & Hurum, Reference Knutsen, Druckenmiller and Hurum2012), where it probably correlates with convergent increase in the number of cervical vertebrae (Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ). The adjacent lateral surfaces of the cervical centra are shallowly concave. The broken bases of the cervical ribs are ventrolaterally positioned and elliptical in basal outline throughout the column. They extend almost the entire length of the centrum in the cranial-most cervicals. The rib facets become longitudinally shorter towards the caudal end. In the mid-neck level, there is a gap between the rib facets and the craniad articular faces (e.g. in cervical 20). In the caudad cervicals (e.g. in cervical 54) the rib facets are placed about in the longitudinal mid-section of the centra (Fig. 3a, e, i, m).

Figure 4. Diagnostic structures of the cervical vertebrae in the Libonectes morgani holotype SMU SMP 69120. (a) Conjoined foramina subcentralia in cervical 27 viewed in ventral aspect. (b) Divergent laminae at the craniad base of the neural spine in cervical 52. (c) Articulated cervicals 45–49 in lateral view indicating progressive migration of the longitudinal ridge (white arrows) across the lateral centrum surface (black arrow indicates craniad direction). Scale bars equal 30 mm in (a) and (b); 10 mm in (c).

A prominent midline keel transects between the paired foramina subcentralia (Storrs, Reference Storrs1991) on the ventral surfaces of all centra (Fig. 3c, g, k, o, s). This keel is narrow and rounded in the craniad cervicals, but becomes transversally broader and more flattened caudally (see cervical 51, Fig. 3s). It is also transversely flared at its terminal ends and more caudally expanded in the craniad cervicals. The elliptical foramina subcentralia are positioned directly adjacent to the midline keel. In the craniad cervicals they are situated within inset grooves (lost caudally), and may be offset relative to each other in cervical 19 and 20, but are fully conjoined in cervical 27 (Fig. 4a); a similar polymorphism has been detected in the holotype skeleton of H. alexandrae (UCMP 33912, see Welles, Reference Welles1943) and indeterminate elasmosaurid specimens from the Maastrichtian of Morocco (Vincent et al. Reference Vincent, Bardet, Houssaye, Amaghzaz and Meslouh2013). In the caudal-most cervicals the foramina subcentralia are very prominent and occupy about one-fifth of the length of the centra.

The cervical neural arches enclose a neural canal whose outline shape varies from nearly circular (at the cranial end) and triangular (at the caudal end) in the craniad cervicals (e.g. in 4, 11 and 20), to sub-triangular (at the cranial end) and oval (at the caudal end) in the mid-neck (present in cervicals 27 to 35), and finally circular (at the cranial end) to oval (at the caudal end) in the caudal part of the neck (present in cervical 51). The prezygapophyses are not divided, but form a continuous concave articular surface (Fig. 3b, f, j, n, r: as described in many elasmosaurids, e.g. Welles, Reference Welles1943; Sato, Reference Sato2003; Sato, Hasegawa & Manabe, Reference Sato, Hasegawa and Manabe2006; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012; Vincent et al. Reference Vincent, Bardet, Houssaye, Amaghzaz and Meslouh2013) whose width is distinctly less than the maximum transverse dimension of the centrum. The projecting cranial edge of the prezygapophyseal articular surface is embayed in dorsal view (most pronounced in cervical 4, Fig. 3d), and the lateral margins curve dorsolaterally when viewed in cranial aspect. In lateral perspective they also project well beyond the articular faces of the centra (Fig. 3a, e, i, m, q); this extends up to one-third of their articular surface length in the craniad cervicals (e.g. cervical 12, Fig. 3e), or approximately half in the mid-section of the neck (e.g. cervicals 20 and 35; Fig. 3i, m), and at around two-thirds in the caudal-most cervical vertebrae (e.g. cervical 52). The postzygapophyses are elliptical in shape in ventrolateral aspect, and dorsolaterally oriented in caudal view. Their articular surfaces are flattened. Unlike the prezygapophyses, the postzygapophyses diverge along their midline, but a transverse sheet of interconnecting bone is still evident caudally (Fig. 3l, p). The combined width of the postzygapophyses is less than that of the accompanying centrum. They also project beyond the caudal articular face by about half their length (e.g. cervicals 4 and 12; Fig. 3a, e); this reduces to around a third in the middle and caudal parts of the cervical column (e.g. cervicals 20 and 35; Fig. 3i, m). A rounded ledge runs from the caudal edges of the postzygapophyses to the base of the neural spine, and continues to contact the prezygapophyses in cervical 4; conversely this is less obvious in cervical 12 and disappears by cervical 20.

Only broken remnants of the neural spine bases are preserved. These suggest a transversely compressed cross-section that widens towards the caudal part of the column, and has a sharply tapered cranial edge (a corresponding rounded ridge is evident caudally). In lateral view, the cranial edge is dorsocaudally curved in the craniad cervicals. In cervical 4 it projects forward along the midline of the prezygapophyseal platform (Fig. 3d). In cervical 52 (and probably 58) the basal craniad edge of the neural spine bifurcates into paired lamellae that run parallel, and eventually merge (Fig. 4b). Deeply excavated cavities border these lamellae and are floored by prominent foramina opening into the spongiosa. Sato, Hasegawa & Manabe, (Reference Sato, Hasegawa and Manabe2006) and Vincent et al. (Reference Vincent, Bardet, Houssaye, Amaghzaz and Meslouh2013) described similar structures on the neural spines of other elasmosaurids and speculated that they might have served as attachments for connective tissue.

3.c. Ribs and gastroliths

Some sections of both the dorsal and gastral ribs, together with some gastroliths, are still preserved in SMU SMP 69120. Welles (Reference Welles1949) did not discuss these components in detail, but Shuler (Reference Shuler1950) presented photographs of sectioned ribs (Shuler, Reference Shuler1950, pp. 10–11, figs 3, 4) and showed the gastralia in association with dispersed gastroliths (Shuler, Reference Shuler1950, pp. 19–21, figs 11–17). A few incompletely prepared gastralia are today evident in a large block of matrix and suggest an interlocking meshwork of dorsoventrally flattened elements similar to that found in other plesiosaurians (e.g. Sollas, Reference Sollas1881, pl. 23; Fraas, Reference Fraas1910, p. 114, fig. 3; Wegner, Reference Wegner1914, p. 281, fig. 9; Ketchum & Smith, Reference Ketchum and Smith2010, p. 1077, fig. 5).

Shuler (Reference Shuler1950, p. 18) mentioned that more than 70 gastroliths were originally present with the skeleton, and these ranged from around 10–80 mm (‘half an inch’ to ‘4 inches’) in maximum diameter. They were also apparently found enclosed within the ribcage, and lithologically composed of chert (‘flint’ sensu Shuler, Reference Shuler1950, p. 18).

3.d. Appendicular skeleton

Welles (Reference Welles1949, pp. 18–21) provided a brief description of the incompletely prepared pectoral girdle and partial right forelimb of SMU SMP 69120 (Fig. 5). The former incorporated the clavicle, possibly the interclavicle, both scapulae and the coracoids exposed in ventral view; these elements were articulated but somewhat distorted by crushing (Welles, Reference Welles1949, pp. 18–19, fig. 2). Welles (Reference Welles1949, Reference Welles1962) examined the humerus, radius and radiale in his discussion of the limbs; however, part of the ulna and intermedium were also shown in his accompanying drawings, and a distal carpal and metacarpal were additionally mentioned. Furthermore, Shuler (Reference Shuler1950, p. 9) stated that Welles (Reference Welles1949) measured the ‘pelvic area’, but this appears to be an error since there is no indication that the rear part of the skeleton was ever recovered.

Figure 5. Schematic drawings (reproduced from Welles, Reference Welles1949) of the now lost pectoral girdle and proximal forelimb elements of the Libonectes morgani holotype SMU SMP 69120. (a) Pectoral depicted in ventral view; (b) articulated proximal forelimb elements. Abbreviations: ce – caudal extension; cl – clavicle; co – coracoid; h – humerus; ice – intercoracoid embayment; in – intermedium; pb – pectoral bar; pf – pectoral fenestrum; r – radius; re – radiale; sc – scapula; u – ulna. The original restorations were not drawn to scale.

According to Welles (Reference Welles1949, p. 18), the clavicles obscured part of an element that could have been the displaced interclavicle, but the exact arrangement of dermal bones along the craniad margin of the pectoral girdle was unclear. Nevertheless, the clavicles were apparently ‘fused along their midline’ (with greatest anteroposterior dimension along the central margin) and did not enclose a median fenestra (sensu Druckenmiller & Russell, Reference Druckenmiller and Russell2008a , p. 19, characters 119–121, who scored these traits as present in the putative elasmosaurid Muraenosaurus leedsii). Welles (Reference Welles1949, p. 18) further reported that the scapulae met along the midline in a ‘long suture’ that excluded the clavicles from contact with the coracoids via a broad pectoral bar (‘ten centimeters wide and at least as deep’, but where these dimensions were measured was not stipulated). Development of the pectoral bar is known to have been ontogenetically variable in plesiosaurians (see Welles, Reference Welles1952; Brown, Reference Brown1981; Carpenter, Reference Carpenter1999), yet it has been disparately incorporated into many phylogenetic character descriptions (e.g. O’Keefe, Reference O’Keefe2001; T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ; Ketchum & Benson, Reference Ketchum and Benson2010; Benson & Druckenmiller, Reference Benson and Druckenmiller2014). Amongst elasmosaurids, formation of the pectoral bar by the scapulae and coracoids seemingly occurred in the osteologically mature type specimens of E. platyurus (ANSP 10081: based upon the reconstruction by Cope, Reference Cope1869, the original fossils are now lost) and Waputskanectes betsynichollsae (TMP 98.49.02, see Druckenmiller & Russell, Reference Druckenmiller and Russell2006), as well as in indeterminate remains from the Maastrichtian of Morocco (Vincent et al. Reference Vincent, Bardet, Houssaye, Amaghzaz and Meslouh2013). It has also been documented in M. leedsii and the reconstructed growth series of the Jurassic cryptoclidid Cryptocleidus eurymerus (Phillips, Reference Phillips1871) (Andrews, Reference Andrews1910; Brown, Reference Brown1981; Carpenter, Reference Carpenter1999).

According to the measurements published by Welles (Reference Welles1949, p. 21, table 2), the maximum parasagittal length of the coracoids exceeded that of the scapulae by a ratio of 1.5 (600 mm versus 400 mm). This does not comply with the proportions of the accompanying line drawing (yielding 1.36 for the same dimensional relationship, see Welles, Reference Welles1949, p. 19, fig. 2), which we regard as schematic. Likewise, the redrafted diagram presented by Welles (Reference Welles1962, p. 58, fig. 12a) depicts different, albeit closer, parameters again (coracoid/scapula ratio = 1.48), and implies that the vaunted pectoral bar of SMU SMP 69120 was actually incomplete. Interestingly, O’Keefe (Reference O’Keefe2001, Reference O’Keefe2002) proposed that ‘plesiosauromorphs’ generally trended towards shortening of the coracoid in comparison to the scapula, and especially elasmosaurids, of which SMU SMP 69120 purportedly possessed ‘subequal’ coracoid/scapula lengths. The doctoral thesis of Sato (T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002), as well as Druckenmiller & Russell (Reference Druckenmiller and Russell2008a ), Ketchum & Benson (Reference Ketchum and Benson2010) and other derivative analyses have subsequently followed this incongruous phylogenetic coding, although Benson & Druckenmiller (Reference Benson and Druckenmiller2014) alternatively introduced a subjective multistate subdivision that placed SMU SMP 69120 together with W. betsynichollsae (coracoid/scapula ratio = 1.38 based on Druckenmiller & Russell, Reference Druckenmiller and Russell2006) in a derived ratio range of ‘< or equal to 1.6’.

Welles (Reference Welles1949, p. 18) stated that ‘a heavy central thickening [was present] at the ventral centre of the midline suture’ of the coracoids in SMU SMP 69120. This corresponds with the distinctive ventral process encountered in many elasmosaurids (e.g. Welles, Reference Welles1943; T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Hiller et al. Reference Hiller, Mannering, Jones and Cruickshank2005; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012; Otero, Soto-Acuña & Rubilar-Rogers, Reference Otero, Soto-Acuña and Rubilar-Rogers2012; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014), some of which may elaborate the structure into a prominent midline projection (e.g. Hiller & Mannering, Reference Hiller and Mannering2005; Druckenmiller & Russell, Reference Druckenmiller and Russell2006). Traditional diagnoses of Elasmosauridae (e.g. Welles, Reference Welles1962; Persson, Reference Persson1963; O’Keefe, Reference O’Keefe2001) also usually list the presence of a substantial intercoracoid embayment that was ‘cordiform’ in outline in SMU SMP 69120 (Welles, Reference Welles1962, p. 56) and ‘almost closed [off]’ behind the coracoids (Welles, Reference Welles1949, p. 19). The intercoracoid embayment is often returned as an unequivocal synapomorphy for Cretaceous elasmosaurids (e.g. O’Keefe, Reference O’Keefe2001; T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Ketchum & Benson, Reference Ketchum and Benson2010), but has not yet been clearly differentiated from the coracoid perforations evident in polycotylids (see comments in T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ) or the proximally positioned coracoid fenestra occurring in the enigmatic ‘pliosauromorph’ Leptocleidus superstes Andrews, Reference Andrews1922 (Kear & Barrett, Reference Kear and Barrett2011). Furthermore, Sennikov & Arkhangelsky (Reference Sennikov and Arkhangelsky2010) ascribed fragmentary Late Triassic sauropterygian material to Elasmosauridae because its coracoid remnants were reconstructed as being much wider proximally than distally (inferring an intercoracoid vacuity), but we consider this to be tenuous given the absence of tangible evidence (the only other noted elasmosaurid similarity was an extrapolated ‘subtrapezoid’ outline of the pubis, see Sennikov & Arkhangelsky, Reference Sennikov and Arkhangelsky2010, p. 569).

The right humerus of SMU SMP 69120 seems to have been preserved in two sections: the ‘flat’ proximal articular head, which was still in place in the glenoid, and the broken distal extremity that Welles (Reference Welles1949, p. 20) reconstructed as having a ‘long postero-distal extension’, but was damaged in the area of its ‘anterior knee’. Little unambiguous information can be gleaned from these remarks, although Carpenter (Reference Carpenter1999) did list caudal expansion of the humerus as a distinguishing feature of Hydralmosaurus serpentinus (Cope, Reference Cope1877), and Welles (Reference Welles1949) also identified the discrete epipodial facets on SMU SMP 69120 as being concave (a convex ulnar facet is known in Terminonatator ponteixensis: Sato, Reference Sato2003).

The epipodials of SMU SMP 69120 were represented by part of the right ulna and both radii, which according to Welles (Reference Welles1949, p. 21, table 2) were noticeably broader than long (unspecified right or left radius craniocaudal width = 180 mm; proximodistal length = 150 mm). This conforms to the typical epipodial proportions found in most elasmosaurids, except for some aberrant taxa such as Callawayasaurus colombiensis and Aristonectes quiriquinensis in which the radii are longer than broad (left/right craniocaudal width = 180/175 mm; proximodistal length = 16.5/150 mm, see Welles, Reference Welles1962, p. 33, table 2; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014, pp. 115, 118, fig. 15A, C). Phylogenetic analyses of SMU SMP 69120 have also pointedly scored both the presence of an epipodial foramen (the development of which is intraspecifically variable in elasmosaurids, see T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002, character 208; Benson & Druckenmiller, Reference Benson and Druckenmiller2014, character 261) and a preaxial concavity on the radius (T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002, character 204; Benson & Druckenmiller, Reference Benson and Druckenmiller2014, character 256). We assume that these interpretations were made from Welles's (Reference Kennedy, Walaszczyk and Cobban1949, p. 20, fig. 3) diagram since his only written reference to such structures is a ‘very deep notch on the ulnar surface’ of the right radius. Critically, the drawing of SMU SMP 69120 produced by Welles (Reference Welles1962, p. 58, fig. 2b) showed different epipodial foramen proportions, with the same preaxial margin of the right radius depicted as convex (coded ‘0/1’ by T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002, p. 286, appendix D).

The only comment that Welles (Reference Welles1949, p. 21) made about the distal components of the limbs in SMU SMP 69120 was the occurrence of an ‘anteroproximal projection’ on the radiale, which he surmised to be a ‘completely fused supernumerary ossicle’. Postaxial accessory ossicles are certainly preserved in some elasmosaurids, but where identifiable, form discrete elements that often lack obvious synarthrotic facets (see T. ponteixensis, W. betsynichollsae, A. quiriquinensis Sato, Reference Sato2003; Druckenmiller & Russell, Reference Druckenmiller and Russell2006; Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014).

4.a. Atlas–axis complex

Welles (Reference Welles1949) listed complete fusion of the atlas and axis, and the presence of a hypophyseal ridge as distinguishing features of SMU SMP 69120. Carpenter (Reference Carpenter, Callaway and Nicholls1997, p. 214) also added a qualitatively determined ‘short and deep’ atlas–axis centrum, a low axis neural spine, and caudad projection of the atlas postzygapophyses beyond the centrum articular face. Relative external fusion of the various atlas–axis complex components is clearly ontogenetic (see Brown, Reference Brown1981), and a hypophyseal ridge is expressed in a variety of long-necked plesiosaurians (Andrews, Reference Andrews1910; Welles, Reference Welles1943; Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ), although its expansion into a transversely broad craniad tubercle does appear to be distinctive.

The length/height dimensions of the atlas–axis complex in SMU SMP 69120 yield a ratio of 1.76. Carpenter (Reference Carpenter, Callaway and Nicholls1997, Reference Carpenter1999) described this as ‘short and deep’ similar to Tuarangisaurus keyesi (1.89, see Wiffen & Moisley, Reference Wiffen and Moisley1986, p. 211, table 2) and Thalassomedon haningtoni (1.51, see Welles, Reference Welles1943, p. 162, table 4). Elasmosaurus platyurus was alternatively classified as ‘long and low’ (1.83, see Sachs, Reference Sachs2005a , p. 95, table 2), despite its lack of coherent proportional differentiation. Other elasmosaurids produce disparate ratio values, and are influenced by ontogeny (Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003): Aristonectes parvidens = 1.44 (Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003, p. 110); A. quiriquinensis = 1.43 (Otero et al. Reference Otero, Soto-Acuña, O’Keefe, O’Gorman, Stinnesbeck, Suárez, Rubilar-Rogers, Salazar and Quinzio-Sinn2014, p. 111); Eromangasaurus australis = 1.6 (Kear, Reference Kear2005a , p. 799); and Albertonectes vanderveldei = 2.2 (Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012, p. 560).

Buchy (Reference Buchy2006, p. 7) diagnosed the type specimen of Libonectes atlasense (SMNK-PAL 3978) by its axis neural arch being ‘1.5 times higher’ than the accompanying centrum. Our personal inspection of SMNK-PAL 3978, however, showed the axial neural spine to be highly distorted. Moreover, Buchy (Reference Buchy2006) did not provide measurements to support this proportional relationship, which we find misleading given that the equivalent ratio of SMU SMP 69120 is virtually identical (1.4).

Carpenter (Reference Carpenter1999) discriminated SMU SMP 69120 from T. keyesi and T. haningtoni by its ‘low’ neural spine. Conversely, Wiffen & Moisley (Reference Wiffen and Moisley1986, p. 211) described this structure as ‘long’ and ‘low’ in the holotype of T. keyesi (NZGS CD426), but no precise measurements were given. On the other hand Welles (Reference Welles1943, p. 162, table 4) explicitly recorded a craniocaudal length for the axial neural spine of T. haningtoni as 91 mm and a dorsoventral height of 61 mm. According to Carpenter (Reference Carpenter1999), this represented a ‘tall’ category relative to SMU SMP 69120 (preserved craniocaudal length = 59.84; dorsoventral height = 19.97 mm) and E. platyurus, despite the neural spine in the E. platyurus type specimen (ANSP 10081) being incomplete (Sachs, Reference Sachs2005a ). Furthermore, we found that the neural spine apex of SMU SMP 69120 was transversely expanded, forming a platform. This resembles the squat axial neural spine of A. parvidens (Chatterjee & Small, Reference Chatterjee, Small and Crame1989, p. 207, fig. 10A, B), but has not been documented in other elasmosaurids and appears to be diagnostic (e.g. compare with Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012).

Carpenter (Reference Carpenter, Callaway and Nicholls1997, Reference Carpenter1999) cited projection of the atlas postzygapophyses beyond the centrum articular face as identifying SMU SMP 69120 in comparison to T. haningtoni and E. platyurus. Incongruously, though, Welles (Reference Welles1943) did not mention the condition of the postzygapophyses in his scant report on the atlas–axis of T. haningtoni. Indeed, his only figure shows that the caudal half of the postzygapophyseal facets extends past the axis articular facet (see Welles, Reference Welles1943, p. 239, pl. 22a). Likewise, Sachs (Reference Sachs2005a , p. 97) reported that the postzygapophyses of E. platyurus ‘reach over the level of the centrum’ in the cervical series, but are missing from the atlas–axis complex (see Sachs, Reference Sachs2005a , p. 100, fig. 4A–C). Based on our observations, caudal projection of the axial postzygapophyses seems to be universal amongst elasmosaurids (e.g. Wiffen & Moisley, Reference Wiffen and Moisley1986; Chatterjee & Small, Reference Chatterjee, Small and Crame1989; Gasparini et al. Reference Gasparini, Bardet, Martin and Fernandez2003; Kubo, Mitchell & Henderson, Reference Kubo, Mitchell and Henderson2012), and also manifests in a variety of other plesiosaurians (e.g. Andrews, Reference Andrews1910; Wegner, Reference Wegner1914; Druckenmiller, Reference Druckenmiller2002; Druckenmiller & Russell, Reference Druckenmiller and Russell2008b ).

As an extension of this character, Buchy (Reference Buchy2006, p. 7) noted projection of the axial postzygapophyses to ‘entirely overhang the centrum of the third cervical’ in L. atlasense (SMNK-PAL 3978). However, our re-examination not only confirmed that the axial postzygapophyses of SMU SMP 69120 extend beyond the centrum for their entire length, but also that the supposed ‘overhang’ of the third cervical in SMNK-PAL 3978 only affects two-thirds of the length of the succeeding vertebra (see Buchy, Reference Buchy2006, p. 24, fig. 6a). We therefore find no discriminative veracity in this trait and do not consider it sufficient grounds (on its own) for separating the species of Libonectes (currently under study by BPK and SS).

An additional diagnostic character state of SMU SMP 69120 is the presence of a rounded foramen enclosed laterally between the atlas and axis neural arches. Gasparini et al. (Reference Gasparini, Bardet, Martin and Fernandez2003), Sachs (Reference Sachs2005a ) and Buchy (Reference Buchy2006) illustrated similar foramina in A. parvidens, E. platyurus and L. atlasense, respectively, but compatible examples also occur in some cryptoclidids (Andrews, Reference Andrews1910; Gasparini, Bardet & Iturralde-Vinent, Reference Gasparini, Bardet and Iturralde-Vinent2002).

4.b. Postaxial cervical vertebrae

Welles (Reference Welles1949), Carpenter (Reference Carpenter, Callaway and Nicholls1997, Reference Carpenter1999) and Buchy (Reference Buchy2006) all listed the occurrence of > 62 cervical vertebrae as a key discriminant of SMU SMP 69120 amongst elasmosaurids. The absence of a complete cervical series for this specimen, however, renders this count as little more than an estimate (Welles, Reference Welles1949, p. 7 even stated this fact). Limited understanding of intraspecific variability in elasmosaurid cervical vertebra number (O’Keefe & Hiller, Reference O’Keefe and Hiller2006), together with conflicting approaches towards weighting (Carpenter, Reference Carpenter1999), scoring (T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Druckenmiller & Russell, Reference Druckenmiller and Russell2008a ) and delimitation (Sachs, Kear & Everhart, Reference Sachs, Kear and Everhart2013), further renders its use problematic.

The merged foramina subcentralia and bifurcating lamellae at the craniad bases of the caudad cervical neural spines identified during our inspection of SMU SMP 69120 are obviously polymorphic and of uncertain diagnostic value. However, the number and relative size of the foramina subcentralia have been used elsewhere as independent discrete character states (Benson & Druckenmiller, Reference Benson and Druckenmiller2014, characters 156, 191); we alternatively consider them to be inconsistent and only applicable when a substantial number of cervical vertebrae are preserved in articulated sequence.

4.c. Appendicular bones

Welles (Reference Welles1949, p. 7) incorporated the presence of a median pectoral bar, transverse midline constriction and subsequent distal expansion of the coracoids, a pronounced caudal projection on the distal articular extremity of the humerus, and a ‘large’ epipodial foramen into his definition for Libonectes morgani. Unfortunately, none of these features can be confirmed from SMU SMP 69120 as it is preserved today. Indeed, the occurrence of a pectoral bar is currently substantiated in only one elasmosaurid taxon, Waputskanectes betsynichollsae (Druckenmiller & Russell, Reference Druckenmiller and Russell2006), and is otherwise an ontogenetically plastic trait (Brown, Reference Brown1981). The shape of the coracoid in SMU SMP 69120 is influenced by formation of the intercoracoid vacuity, which is a widely advocated synapomorphy for Elasmosauridae (Welles, Reference Welles1962; Persson, Reference Persson1963; O’Keefe, Reference O’Keefe2001; T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002; Ketchum & Benson, Reference Ketchum and Benson2010; Otero et al. Reference Otero, Soto-Acuña and Rubilar-Rogers2012). Sato (T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002, p. 43) also employed distocaudal expansion of the humerus to assign Campanian–Maastrichtian elasmosaurid remains (RSM P625.1) to ‘?Libonectes morgani’, yet this condition is present (albeit in an incrementally more extreme form, see T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002, p. 346, character 187) in the stratigraphically proximal taxon Hydralmosaurus serpentinus (Carpenter, Reference Carpenter1999). Unfortunately, the distocaudal extremity of the humerus was reported as missing in L. atlasense (Buchy, Reference Buchy2006).

Sato (T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002) additionally recorded the presence of an epipodial foramen (variable in elasmosaurids, see T. Sato, unpub. Ph.D. thesis, Univ. Calgary, 2002), gradational variation in the length/width ratio of the radius (uniformly broader than long in most elasmosaurids, see e.g. Ketchum & Benson, Reference Ketchum and Benson2010) and the outline of the caudal edge of the ulna (not figured by Welles, Reference Welles1949) as additional diagnostic features of L. morgani. However, we alternatively treat them as ambiguous, and note that minor proportional differences have also been advocated for the classification of other elasmosaurids from the Eagle Ford Group (e.g. TMM 42245–1: G. W. Storrs, unpub. M.Sc. thesis, Univ. of Texas, 1981, p. 102).

4.d. Accompanying cranial features

Welles (Reference Welles1949, p. 7) enigmatically noted ‘[t]eeth 9/16’ in SMU SMP 69120, which appears to refer to his maximum count of alveoli on the right maxilla versus dentary (see Welles, Reference Welles1949, p. 15). Carpenter (Reference Carpenter, Callaway and Nicholls1997, Reference Carpenter1999) alternatively concluded that these numbers were incorrect (we confirmed 14/18 teeth in the right maxilla/dentary, see Carpenter, Reference Carpenter1999, p. 157), and that the dental arrangement of SMU SMP 69120 was generally comparable to that of other elasmosaurids (e.g. Hydrotherosaurus alexandrae: Welles, Reference Welles1943). Furthermore, Carpenter (Reference Carpenter, Callaway and Nicholls1997, Reference Carpenter1999) attributed Welles's (Reference Welles1949) erroneous description of a pineal foramen to inadequate preparation, and additionally corrected his anomalous identification of the ‘nasal’ and ‘prefrontal’ (see Welles, Reference Welles1949, pl. I) as the prefrontal and frontal (corroborated by SS and BPK, pers. obs.). Finally, Carpenter (Reference Carpenter1999) introduced differential placement of the external bony nasal opening over the third–fourth alveolus in the maxillary tooth row to distinguish SMU SMP 69120 from Tuarangisaurus keyesi, in which the bony nasal opening was supposedly situated over the second–third maxillary tooth, and Thalassomedon haningtoni, which had a positioning over the fourth–fifth tooth, or proximally over the sixth tooth in Styxosaurus snowii. Our examination of the bony nasal openings in SMU SMP 69120 showed that their elongate margins actually commenced at the alveolar edge between the third–fourth maxillary tooth on the right side, or at the midline of the third alveolus on the left side, and extended proximally to a point parallel with either the fifth or fourth maxillary alveolus, respectively. Based on the undistorted left side of the holotype skull of T. keyesi (NZGS CD425), the bony nasal opening (as preserved) is delimited between the second and fourth maxillary alveolus (Wiffen & Moisley, Reference Wiffen and Moisley1986, p. 208, fig. 3). Similarly, the crania representing both T. haningtoni (Carpenter, Reference Carpenter1999) and S. snowii (Storrs, Reference Storrs1999; Everhart, Reference Everhart2006) are crushed but again accord with SMU SMP 69120 (see Carpenter, Reference Carpenter1999, p. 164, fig. 12A), or have inconsistent restorations of their damaged external nasal regions as is the case in S. snowii (see Williston, Reference Williston1890, p. 174; Welles & Bump, Reference Welles and Bump1949, p. 525, fig. 3A; Carpenter, Reference Carpenter1999, p. 161, fig. 10A, B).