Osteology of Galeamopus pabsti sp. nov. (Sauropoda: Diplodocidae), with implications for neurocentral closure timing, and the cervico-dorsal transition in diplodocids

Skull (Figs. 4–16; Table 1)

Preservation. The skull of Galeamopus pabsti SMA 0011 is nearly complete. The only bones lacking are the left maxilla and quadrate, the right lacrimal, and the bones from the palate with the exception of the right pterygoid. The skull has a typically diplodocid shape. It is elongate, with the external nares retracted and dorsally facing, and has slender, peg-like teeth (Figs. 4–7). Given the completeness of the skull, a reconstruction was created in cooperation with the Portuguese illustrator Simão Mateus (Fig. 7; Mateus & Tschopp, 2017). When compared with recent reconstructions of the skull of Diplodocus (Wilson & Sereno, 1998; Whitlock, 2011b), Galeamopus has a more triangular skull outline in lateral view, and more sinuous ventral maxillary edges in dorsal or ventral view (Fig. 7).

Premaxilla. The premaxillae are completely preserved. They are anteroposteriorly long and transversely narrow elements (Table 1) that contact each other medially and the maxillae laterally (Figs. 4–7). The posterior end of the premaxillae delimits the nasal opening anteriorly. In dorsal view, the elements are narrow in their central part and widen anteriorly and posteriorly. The anterior edge is straight to slightly convex, whereas the posterior margin is deeply concave, such that the two premaxillae together form a triangular process that enters the nasal opening anteromedially. The medial margin is straight, and the lateral one concave due to the central narrowing of the element. Some nutrient foramina are present on the anterior-most portion of the dorsal surface, as is a groove originating at the premaxillary-maxillary contact, and extending obliquely anteromedially (Figs. 5 and 7). The groove is faint and relatively short, not reaching either the anterior or the medial margin. Such a groove was usually interpreted as typical for dicraeosaurids (Remes, 2009; Whitlock, 2011a), but is also present in other diplodocids (Tschopp, Mateus & Benson, 2015). However, a fading out of this feature is uncommon in dicraeosaurids, where the groove is distinct (Janensch, 1935; Remes, 2009). Ventrally, the anterior portion of the premaxillae thickens slightly dorsoventrally in order to bear the replacement teeth, but not to the extent seen in the referred specimen USNM 2673 (Tschopp, Mateus & Benson, 2015). Five teeth are included in the mounted skull, but only four alveoli occur in the left element, whereas the right premaxilla appears to show five. The alveoli of the articulated premaxillae do not contact each other medially, such that there would be space for two more teeth in between, or a gap. The number of replacement teeth could not be discerned without a CT-scan. At the border with the maxilla, where the premaxilla narrows from the broader anterior part to the narrow central part, the two bones form an elongated fossa, which bears the subnarial and the anterior maxillary foramen. Both foramina lie on the medial edge of the maxilla, very close together.

Maxilla. Only the right maxilla is preserved, and it is complete. The broad anterior portion bears a posterior process, which contacts the jugal and quadratojugal, and a posterodorsal process, which contacts the lacrimal, nasal, and the prefrontal (Figs. 4, 5, 7). The maxilla forms the dorsal, anterior, and anteroventral margins of the antorbital fenestra, and completely encloses the preantorbital fossa. Unlike Kaatedocus and Dicraeosaurus, the preantorbital fossa is pierced by a large fenestra. The fenestra is dorsally capped by a distinct ridge similar to Diplodocus, but unlike Apatosaurus. This distinct dorsal edge was previously thought to represent an autapomorphy of Diplodocus, but was shown to occur in other taxa as well (Tschopp & Mateus, 2013b). The preantorbital fenestra does not fill the entire preantorbital fossa (Table 1): the anterior-most area remains closed by a thin bony wall. The fossa is anterodorsally accompanied by a short, narrow groove more or less following the curvature of the anterior end of the dorsal rim of the fossa. The posterior end of the fossa is interconnected with the central portion of the antorbital fenestra by a distinct groove that extends posterodorsally to the dorsal corner of the posterior process (Fig. 8). This groove otherwise only occurs in the specimen USNM 2673 (Tschopp, Mateus & Benson, 2015). Remaining parts of the external surface of the maxilla do not bear other distinctive morphological features, with the exception of the anterior-most portion, where a few nutrient foramina can be seen, similar to those on the premaxilla. The number of maxillary teeth is difficult to discern in the mounted skull, but is approximately 12.

Nasal. The right nasal is complete. It lies anterior to the frontal, and medial to the prefrontal (Figs. 4–7). A slender, anterior process connects to the maxilla. The nasal is a subtriangular element with a slightly concave anteromedial edge forming a part of the external naris, and posterior and lateral edges that include an angle of about 120°. The anteromedial edge is dorsoventrally thin, but the nasal suddenly gains thickness from there backwards and outwards. The medial corner does not reach the skull midline, such that the two nasals do not touch each other medially. The external naris thus extends posteriorly between the nasal bones into a notch between the frontals. A similar condition might be present in Kaatedocus, which has an anterior notch between the frontals as well, but no nasal is preserved in the holotypic skull, which would confirm the posterior extension of the naris (Tschopp & Mateus, 2013b).

Prefrontal. Both prefrontals are complete. They contact the frontals posteriorly, the nasals medially, the lacrimals posterolaterally, and the maxillae anterolaterally (Figs. 4–7; Table 1). The prefrontals are short, and have an anteroposteriorly convex dorsal surface. Their lateral margin is straight, the medial one is anteriorly and posteriorly concave for the articulation with the nasal and the frontal, respectively. A sharply pointed, medially projecting process separates the two concavities. The posterior edge is anterolaterally-posteromedially oriented, forming a hook-like posteromedial process as is typical for Diplodocidae (Wilson, 2002; Whitlock, 2011a). The process almost reaches the frontal midlength, as is the case in the diplodocine skulls CM 3452 and 11161 (Tschopp, Mateus & Benson, 2015). Anteriorly, the prefrontal tapers to a narrow tip, which is slightly dorsoventrally expanded. The left element bears a small nutrient foramen on the dorsal surface of the anterior part. The ventromedial edge is very distinct.

Frontal. Both frontals are completely preserved. They contact the prefrontal anterolaterally, the nasal anteriorly, each other medially, the parietal posteromedially, and the postorbital posterolaterally (Figs. 4–7). Ventrally, the frontal makes contact with the braincase, articulating with the orbitosphenoid. The frontals have a smooth dorsal surface, which is slightly convex posterolaterally-anteromedially. Their medial border is generally straight, but curves laterally at its posterior and anterior ends, leaving an opening between each other. However, the posterior curvature shows broken edges, so that it is uncertain how much of this opening is due to taphonomic breakage. Therefore, we did not indicate the presence of a pineal foramen in the reconstruction drawing (Fig. 7; Mateus & Tschopp, 2017), also because this is usually interpreted to be absent in diplodocids (Whitlock, 2011a; Whitlock, 2011b).

The anterior curvature forms an anterior notch between the frontals (length 18 mm), similar to the condition in Kaatedocus. The anterior notch is wider than in Spinophorosaurus (Knoll et al., 2012), and different from Kaatedocus in being V-shaped rather than U-shaped (Tschopp & Mateus, 2013b). This differs from the anterior midline projection formed by the frontals of Galeamopus hayi HMNS 175. The anterior margin of the frontal of G. pabsti SMA 0011 is strongly convex transversely in order to accommodate the posterior, hook-like process of the prefrontal anterolaterally. From the posterior-most point of the posterior process of the prefrontal, the frontal has a straight edge extending obliquely anterolaterally, until it reaches the lateral, orbital edge, with which it forms a very acute angle. The lateral border is distinctly concave in dorsal view, smooth in its anterior part, but becoming highly rugose posteriorly, close to where it articulates with the postorbital. Here, the lateral and posterior edges form an acute angle. The lateral portion of the posterior margin is slightly displaced anteriorly, compared to the medial portion, resulting in a somewhat sinuous posterior edge. Ventrally, the frontals are marked by a distinct ridge, extending obliquely from the anterolateral corner, below the posterior process of the prefrontal, to an elevated, broad area for the attachment of the braincase.

Postorbital. Both elements are complete. The postorbital is a triradiate bone with an anterior process articulating with the jugal, a posterior process overlapping the squamosal laterally, and a dorsomedial process covering the posterior edge of the frontal in posterior view and connecting to the anterolateral process of the parietal medially, thereby excluding the frontal from the margin of the supratemporal fenestra (Figs. 4–7). Anteromedially, the dorsomedial process abuts the antotic process of the braincase (Fig. 9). The anterior process has a subtriangular cross section, transversely elongate, with a narrow lateral and a very thin medial margin (Table 1). The dorsal surface of the anterior process is slightly concave transversely. Towards the anterior end, the process tapers to a point. The posterior process is short and triangular. At its base, one (on the right postorbital) or two (on the left element) nutrient foramina occur. The process is compressed transversely. The dorsomedial process is curved dorsoventrally, with the concave surface facing anteriorly. It is relatively high dorsoventrally, but narrow anteroposteriorly. It is anteroposteriorly broader laterally than medially. The anterior face of the dorsomedial process is marked by a horizontal ridge at its base. The ridge supports the posterior edge of the frontal.

Jugal. Both jugals are preserved and complete. The jugal is a flat, relatively large bone with a posterior process contacting the postorbital and a dorsal process articulating with the lacrimal (Figs. 4–7). The main portion connects to the quadratojugal ventrally and the maxilla anteriorly. The jugal forms the anteroventral rim of the orbit, the posteroventral border of the antorbital fenestra, and the anterodorsal edge of the laterotemporal fenestra. The bases of the dorsal and posterior processes are relatively broad, before they taper dorsally and posteriorly, respectively (Table 1). The dorsal process is bifid (Fig. 4). The anterior edge of the jugal is slightly concave, as is the anteroventral margin. These two edges form a rounded anteroventral corner.

Quadratojugal. The quadratojugals are both complete. They are transversely thin bones with a posterodorsal process overlying the quadrate laterally, and a long anterior ramus (Table 1) contacting the jugal dorsally and the maxilla anteriorly (Figs. 4–7). The quadratojugals form the anteroventral margins of the laterotemporal fenestrae, and the ventral borders of the skull. The anterior ramus of the quadratojugal is narrow at its base but expands dorsoventrally towards its anterior end. The ventral edge is almost straight; it is thus the concave dorsal margin of the anterior ramus that accounts mostly for this dorsoventral expansion. The shape of the anterior margin is not discernible in the mounted skull, but based on the photo taken before the mount, it bears a small dorsal projection that connects to the jugal and excludes the maxilla from the margin of the laterotemporal fenestra, and slightly tapers anteriorly towards the articulation with the maxilla. The posterodorsal process is less than half the length of the anterior process. It is inclined posterodorsally, as in all diplodocids (Upchurch, 1998; Wilson, 2002; Whitlock, 2011a). It is anteroposteriorly convex externally, relatively broad at its base, and tapers to a point dorsally, reaching about midlength of the quadrate shaft.

Lacrimal. Only the dorsal half of the left lacrimal is preserved. It is a narrow element expanding towards its dorsal end (Table 1), where it underlies the posterodorsal process of the maxilla anteriorly, and contacts the prefrontal dorsally, and possibly the nasal medially (Figs. 4, 6, 7). Ventrally, the lacrimal would contact the jugal, if this part of the bone were preserved. The lacrimal separates the orbit from the antorbital fenestra. It is anteroposteriorly narrow in its ventral half, with a triangular cross section, flat externally but bearing a distinct dorsoventral ridge internally. The anterior edge has a short, but dorsoventrally high, anterior process at its dorsal end. The posterior margin is generally straight, with only a weak bulge on its dorsal portion. The dorsal-most end curves backwards, below the prefrontal. The internal ridge becomes slightly more pronounced dorsally, posteriorly enclosing the lacrimal foramen, which is small and shallow in SMA 0011.

Quadrate. Only the right quadrate is preserved, but it is complete. It has a complex anatomy, with a quadrate shaft articulating with the squamosal and the paroccipital process posterodorsally and posteroventrally, respectively; a pterygoid flange interconnecting the outer skull with the pterygoid medially; and a ventral ramus overlapped by the quadratojugal externally and bearing the articulating surface for the lower jaw ventrally (Figs. 4, 5, 7). The quadrate shaft is elongate posteriorly (Table 1), and has concave dorsal and ventrolateral surfaces. The lateral edge is a thin crest, where it is not capped by the squamosal or the quadratojugal. The posterior surface of the quadrate shaft and the ventral ramus is shallowly concave, forming the quadrate fossa. The pterygoid flange originates on the medial half of the quadrate shaft. It is very thin mediolaterally, but anteroposteriorly long, and curves medially at its dorsal tip. The dorsal edge of the flange is straight and more or less horizontally oriented. The medial side of the pterygoid flange is concave, but does not form such a distinct fossa like that present in Kaatedocus SMA 0004 (Tschopp & Mateus, 2013b). The ventral ramus of the quadrate of Galeamopus pabsti SMA 0011 is subtriangular in cross-section, with concave anterior and posterolateral surfaces. It has a thinner lateral than medial margin. The articular surface is subtriangular, with a concave anterior border, and a pointed posterior corner. The entire ventral ramus of the quadrate of SMA 0011 is posterodorsally inclined, as in all diplodocids (Upchurch, 1998; Wilson, 2002; Whitlock, 2011a).

Squamosal. Both squamosals are preserved, but lack a part of their anterior process (the right one more so than the left). The squamosals form the posterolateral corner of the skull. They have a complicated morphology, accommodating a variety of elements from the braincase and outer skull (Figs. 4–7). The anterior process overlies the posterior end of the quadrate. Dorsally, the squamosal is laterally covered by the posterior process of the postorbital and forms the external margin of the supratemporal fenestra. Posteriorly the squamosal contacts the paroccipital process and dorsoposteriorly the posterolateral process of the parietal. The squamosal is strongly curved posterolaterally. The anterior process appears to be the longest of all squamosal processes (Table 1), even though it is not preserved in its entire length. The ventral edge of the squamosal bears a short ventral projection at its posterior end, similar to, but much less distinct than the ventral prong present in advanced dicraeosaurids (Salgado & Calvo, 1992; Whitlock, 2011a). A concave area on the dorsolateral surface accommodates the posterior process of the postorbital. Other morphological features are difficult to observe in the articulated, reconstructed skull of SMA 0011.

Parietal. Both parietals are complete but slightly distorted. They are tightly sutured with the frontals anteriorly and have a short anterolateral process to contact the dorsomedial process of the postorbital, with which they form the anterior margin of the supratemporal fenestra (Figs. 4–7). The posterior face of the parietal contacts the exoccipital and the supraoccipital medioventrally. The posterolateral process of the parietal forms the posterior margin of the supratemporal fenestra and reaches the squamosal laterally. The dorsal portion of the parietal in SMA 0011 is very narrow anteroposteriorly (Table 1). The two elements do not touch each other medially, but this appears to be due to postmortem breakage of the extremely thin bone behind the parietal fenestra, which the parietals form together with the frontals. The dorsal portion is flat and not well separated from the posterior surface by a transverse nuchal crest (sensu Knoll et al., 2015) like that in Kaatedocus (Tschopp, Mateus & Benson, 2015). The parietal of Galeamopus pabsti SMA 0011 widens anteroposteriorly at its lateral end, where it develops a short anterolateral and a long and dorsoventrally deep posteroventral process. The parietal thus contributes most to the margin of the supratemporal fenestra. The posterior surface has an oblique ventromedial border, which has a very sinuous suture together with the supraoccipital. The dorsal margin of the posterolateral process is straight and does not cover the anterior border of the supratemporal fenestra in posterior view. The ventral edges are excluded from the posttemporal fenestra by the squamosal and a laterally projecting spur of the exoccipital.

Supraoccipital. The supraoccipital is complete and fused with the parietals and the exoccipital-opisthotic complex. The supraoccipital is a somewhat hexagonal bone, which contacts the parietals dorsolaterally, the exoccipital-opisthotic complex ventrolaterally, and borders the foramen magnum ventrally (Figs. 5, 6, 10). The suture with the exoccipital-opisthotic is barely visible. The dorsolateral edges of the supraoccipital are slightly concave. The ventrolateral edges are visible only laterally; further medially, the suture becomes obliterated up to the foramen magnum, but probably extended below the two distinct tubercles located dorsolateral to the foramen magnum. These tubercles served for the attachment of the proatlases. The tubercles are ellipsoid, oriented with their long axes extending dorsomedially-ventrolaterally. The elevation is much more distinct ventrally than dorsally. The dorsal portion of the supraoccipital bears a complex arrangement of ridges and concavities (Fig. 10). This complex structure is symmetrical and well-defined, arguing against a taphonomic or pathological origin. No distinct sagittal ridge occurs. In fact, the elevated area is marked by a vertical midline groove, which is otherwise only present in the skull USNM 2673 among diplodocids. A similar, but wider, sagittal groove occurs convergently in the titanosaurs Rapetosaurus krausei, Muyelensaurus pecheni, and Bonatitan reigi (Curry Rogers & Forster, 2004; Calvo, González Riga & Porfiri, 2007; Salgado Gallina, & Paulina-Carabajal, 2015). Given that the supraoccipital of Galeamopus hayi HMNS 175 does appear to bear a distinct sagittal nuchal crest, the groove is here interpreted to be an autapomorphy of the species Galeamopus pabsti. The supraoccipital has its greatest width slightly below midheight. No distinct foramina occur close to the border with the parietal, unlike in Kaatedocus (Tschopp & Mateus, 2013b). The dorsal-most portion of the supraoccipital of SMA 0011 tapers, not forming a distinct dorsal elevation as in Apatosaurus CM 11162 (Berman & McIntosh, 1978), or the indeterminate flagellicaudatan MB.R.2388 (Remes, 2009).

Exoccipital-opisthotic complex. The outer portion of the braincase is completely preserved. No sutures can be seen between the exoccipital and the opisthotic (the fused complex is sometimes called otoccipital; Knoll et al., 2012; Knoll et al., 2015; Royo-Torres & Upchurch, 2012). They bear two elongate paroccipital processes that extend lateroventrally to articulate with the squamosal and the posterior end of the quadrate (Figs. 5 and 6). Medially, the exoccipital-opisthotic borders almost the entire foramen magnum except for a small dorsal contribution of the supraoccipital. The exoccipital forms the dorsolateral corners of the occipital condyle. As in Suuwassea and Diplodocus CM 11161, the exoccipital almost excludes the basioccipital from the participation in the dorsal surface of the occipital condyle (Harris, 2006a). The lateral surface of the condylar neck is pierced by two foramina, which are the exits for cranial nerve XII (Fig. 11; Knoll et al., 2012), unlike the condition in Amargasaurus, which only has a single exit (Paulina Carabajal, Carballido & Currie, 2014). The two exits for cranial nerve XII are anteriorly bordered by the crista tuberalis, which separates them from the opening for cranial nerves IX–XI (the metotic foramen; Knoll et al., 2012; Royo-Torres & Upchurch, 2012), and extends from the ventral edge of the paroccipital process onto the basioccipital, of which it forms the posterolateral edge until it reaches the basal tubera (Fig. 11). The metotic foramen is anteriorly bordered by the crista interfenestralis, which separates two well developed fossae between the crista tuberalis and the otosphenoidal crest (Makovicky et al., 2003).

The paroccipital processes of Galeamopus pabsti SMA 0011 have slightly convex external surfaces, but do not bear a ridge as in Kaatedocus (Tschopp & Mateus, 2013b). The ventral edge of the paroccipital process is straight, only the dorsal corner of the distal end is expanded dorsally, resulting in a distinctly concave dorsal edge. The lateral margin of the paroccipital process is subtriangular, with a longer, vertically oriented dorsal portion, and a shorter, laterally inclined ventral part. In lateral view, it is straight, unlike the curved ends of the element in Suuwassea and Galeamopus hayi (Harris, 2006a; Tschopp, Mateus & Benson, 2015).

Basioccipital and basisphenoid. The basioccipital forms the main portion of the occipital condyle. It is relatively short and connects the articular surface of the occipital condyle with the basal tubera (Fig. 11), which are of about the same width (Table 1). The articular surface of the occipital condyle is offset from the condylar neck. Narrow ridges connect the midline of the ventral aspect of the condylar neck with the posteromedial corners of the basal tubera and the lateral face of the neck with the crista tuberalis. The ridges result in concave lateral surfaces of the basioccipital and concave posterior surfaces of the basal tubera. The concavity on the posterior surface of the tubera is anteroventrally confined by a distinct, transversely convex ridge, which separates the posterior and ventral surfaces of the tubera (Fig. 11). The basal tubera are box-like, and medially separated by a distinct, but relatively narrow notch. The ventral edges of the tubera form a nearly straight line in posterior view, whereas the anterior edges are angled in a wide V-shaped manner in ventral view. Anteriorly, the basipterygoid processes attach to the tubera. In the reconstructed skull, the processes are mounted slightly dorsal to their actual location, above the anteroventral end of the otosphenoidal crest (Fig. 12). When articulated properly, they would be elongate (5.3 times longer than wide; Table 1), straight, and would form a narrower angle than as mounted. This is important because shorter and more widely diverging basipterygoid processes are typical for Apatosaurus, whereas narrower angles are typical in Diplodocus (Berman & McIntosh, 1978). The processes are not as well connected at their base as is the case in Kaatedocus (Tschopp & Mateus, 2013b). The distal ends of the basipterygoid processes are expanded.

Orbitosphenoid. The orbitosphenoids delimit the endocranial cavity anteriorly and attach to the frontals dorsally, each other medially, and the laterosphenoids posterolaterally. Each orbitosphenoid is relatively wide dorsally and has an anteroventral process, which is expanded at its end and separates the two openings for cranial nerves II medially (the optic foramen) and III laterally (the oculomotor foramen; Fig. 12; Janensch, 1935; Harris, 2006a; Balanoff, Bever & Ikejiri, 2010; Knoll et al., 2015). Unlike the condition in Suuwassea or Europasaurus (Harris, 2006a; Sander, Mateus & Laven, 2006), the optic foramen of Galeamopus is bridged over by bone medially. Anterodorsally, the two orbitosphenoids form the olfactory fenestra together with the frontals (the exit for cranial nerve I; Fig. 12; Janensch, 1935; Balanoff, Bever & Ikejiri, 2010), and posterolaterally, at the junction with the laterosphenoid, the foramen for cranial nerve IV (the trochlear foramen; Balanoff, Bever & Ikejiri, 2010) defines the outline of the orbitosphenoid.

Laterosphenoid. The laterosphenoid mainly consists of a crest that bears the antotic (or capitate; Knoll et al., 2012; Knoll et al., 2015) process posterodorsally and extends anteroventrally to join the otosphenoidal crest. It connects to the frontal and parietal posterodorsally, the orbitosphenoid anterodorsally, and the prootic posteroventrally. As for the orbitosphenoid, the laterosphenoid outline is defined by various openings: cranial nerves III and IV anteriorly at the junction with the orbitosphenoid, the trigeminal foramen posterodorsally (cranial nerve V; Balanoff, Bever & Ikejiri, 2010), as well as the oculomotor foramen and the abducens foramen anteroventrally (Fig. 12; Balanoff, Bever & Ikejiri, 2010). Dorsal to the opening for cranial nerve IV, there is a separate, small opening for the orbitocerebral vein, similar to the condition in Diplodocus and other sauropods, but different from Amargasaurus (Paulina Carabajal, 2012; Paulina Carabajal, Carballido & Currie, 2014). The antotic crest separates the trigeminal foramen from the other openings. The antotic process is dorsoventrally higher than anteroposteriorly long, and tapers laterally to a rounded tip, which contacts the postorbital.

Prootic. The prootic lies between the laterosphenoid anterodorsally, the parietal and paroccipital processes posterodorsally, and the basisphenoid anteroventrally. The prootic bears the well-developed otosphenoidal crest, which extends relatively far laterally, but is very thin dorsoventrally. It does not end in an additional transverse expansion anteriorly, as would be typical for dicraeosaurids (Janensch, 1935; Paulina Carabajal, Carballido & Currie, 2014). The otosphenoidal crest extends between the foramen for cranial nerve V more anteriorly and the ones for cranial nerves IX–XII more posteriorly (Paulina Carabajal, Carballido & Currie, 2014). Posterodorsally, the otosphenoidal crest extends to the base of the paroccipital processes, and bifurcates to enclose the foramen for cranial nerve VII. The two branches reunite before reaching the paroccipital process, similar to the condition in Amargasaurus (Paulina Carabajal, Carballido & Currie, 2014).

Pterygoid. The left pterygoid is only partly prepared (Fig. 13). The pterygoid connects the quadrate posterolaterally with the basipterygoid processes posteromedially, the ectopterygoid and palatine anterolaterally, and the vomer anteromedially. The two elements would join along the midline of the skull. The pterygoid of SMA 0011 resembles the same bone in the indeterminate diplodocine CM 3452 in its dorsoventrally deeper shape compared to Camarasaurus and Giraffatitan (McIntosh & Berman, 1975). A shallow articulation facet for the basipterygoid processes lacks the hook-like process present in dicraeosaurids and Camarasaurus (Wilson, 2002; Whitlock, 2011a).

Ceratobranchial. Only the right ceratobranchial is preserved, but appears to be almost complete (Fig. 14). The ceratobranchial is a bone of the hyoid apparatus, with no bony connections to the rest of the skull (Wilson et al., 2016). It is a narrow bone, with a distinct upward curve at midlength. The anterior ramus becomes transversely flattened towards its anterior end, which bears a shallow longitudinal groove on the medial side. The ceratobranchial slightly widens dorsoventrally where it curves upwards and towards the squamosal, as was shown in Tapuiasaurus (Zaher et al., 2011; Wilson et al., 2016). The posterodorsal end is rounded and offset from the shaft by a distinct rim.

Mandible

The mandibles preserve both dentaries and surangulars, and the left angular. Two additional, thin bones might represent the prearticulars. No articular, splenial, and coronoid is preserved.

Dentary. Both dentaries are preserved. The dentary is the anterior-most bone of the lower jaw and the only one bearing teeth. Posteriorly, it is followed by the surangular dorsally and the angular ventrally (Figs. 4–7). Internally, it would be overlain by the splenial ventrally, but this is not visible due to the mount. The dentary is a thin bone, with a dorsoventrally high dentigerous portion (Table 1), having the typical, ventrally projecting ‘chin’ of flagellicaudatans (Fig. 6; Upchurch, 1998; Whitlock, 2011a). The anteromedial portion is marked by several small, irregularly placed pits. A relatively larger, distinct foramen pierces the lateral surface at midheight below the posterior-most tooth. The medial wall of the dentigerous portion of the dentary projects further dorsally than the medial wall. Posterior to the tooth bearing portion, the dentary tapers in dorsoventral height, the right one much more so than the left. The symphysis is oblong and strongly anteriorly inclined. There are at least eleven, possibly twelve, dentary teeth.

Surangular. Both surangulars are present. This bone is very flat transversely, curves ventrally at its posterior end and bears a foramen at its highest point, which is also the highest point of the entire lower jaw (Figs. 4–7). The jaw does not bear a coronoid eminence.

Angular. Both angulars are incomplete anteriorly. They are concave externally, due to the laterally curving ventral edge. They taper relatively continuously anteriorly, but abruptly at their posterior ends (Figs. 4–7), where they expand transversely in order to accommodate the articular, which is not preserved.

?Prearticular. Both prearticulars appear to be present, but are partly hidden in the mount or only partially prepared, and separately stored in the SMA collections (Fig. 15). They are thin, elongate bones that taper posteriorly. A very shallow groove marks the probable lingual surface, extending anteroposteriorly, following the somewhat sinuous curve of the dorsal edge of the bone. In its anterior half, the bone becomes slightly thicker mediolaterally and curves outwards.

Teeth

The teeth have the typical diplodocoid, peg-like shape, and have a Slenderness Index (SI) of approximately 4 (Fig. 16; Tschopp, Mateus & Benson, 2015: tab. S16). They are slightly wrinkled but do not have denticles. Worn teeth usually have a single wear facet at a low angle to the long axis of the tooth, but some teeth also show two facets that are conjoined apically. In these teeth, the lingual facet is more steeply inclined than the labial one. The crown tips are slightly wider than deep, which is especially visible in replacement and/or unworn teeth, which have a very weakly spatulate upper-most crown. The enamel is distributed evenly on all sides, and no grooves mark the lingual face. In the jaws, the teeth are inclined anteriorly relative to the long axis of the jaw, and set side-by-side without overlapping each other.

Cervical vertebrae (Figs. 17–32; Table 2)

Preservation. Thirteen cervical vertebrae are present, as is the right proatlas. The cervical vertebrae were found partly articulated. The proatlas and atlas were recovered among the disarticulated skull elements. Axis to CV 5 were lying semi-articulated in close association, followed by the slightly disarticulated CV 6–8. After a short gap of 0.3 m, CV 9 and 10 were found articulated, and finally a block of five articulated elements including the cervico-dorsal transition was recovered at a distance of about 1 m from the next nearest vertebrae (Fig. 3). The gap between CV 8 and 9 is interpreted to be too short to accommodate yet another element, which in this area of the neck already reach lengths of at least 150% the distance of the gap. Also, measurements of the posterior cotyle of CV 8 and the anterior condyle of CV 9 more or less fit to each other, taking the deformation of CV 8 into account (Table 2). Thus, the only reasonable position, where cervical vertebrae could be missing, is between CV 10 and the block including the cervico-dorsal transition. None of the cervical ribs were fused to their centra, and certain anterior to middle ribs were found at some distance from the vertebrae. However, combining the positional information from the quarry maps and the size and side of the ribs, an attribution of most of them to their respective centra was possible. Five ribs belonging to the articulated cervico-dorsal transition were found in place, yielding crucial information about the changes in morphology from the neck to the back. Two pairs of them are transitional in shape, but can still be interpreted as cervical ribs due to the presence of an anterior process and their short posterior shaft (see below). They belong to the second and third articulated vertebra of the transitional block. One pair and a single rib are definitive dorsal ribs, and were found semi-articulated with the last two vertebrae in the block.

Proatlas. The right proatlas is preserved and complete (Fig. 17). It connects the braincase with the atlantal neurapophyses. The proatlas of SMA 0011 is strongly curved and tapers distally. The proximal articular surface is ovoid, with the largest width located in the dorsal half. The medial surface is concave, the lateral one convex. The proatlas of SMA 0011 is different from the element in Kaatedocus (see Tschopp & Mateus, 2013b: figs.  3–6) due to its much narrower distal tip.

Atlas. The atlantal centrum is not fused to the neurapophyses (Fig. 18). It has a well-developed anteroventral lip as is typical for diplodocids, and convergently present, although less evident in several other sauropods (Mannion, 2011; Whitlock, 2011a). A large foramen lies between the posterolateral projections at the posteroventral edge of the centrum. The lateral surface of the centrum is concave and bears a foramen as well, resembling an incipient pleurocoel. The neurapophyses have a relatively wide base, and turn upwards and backwards to articulate with the prezygapophyses of the axis. A wide medial process occurs anteriorly, as in the specimen AMNH 969 (Holland, 1906). This process articulates with the proatlas, and is much better developed than in Diplodocus USNM 2672 or Kaatedocus (Marsh, 1896; Hatcher, 1901; Tschopp & Mateus, 2013b). A small but distinct subtriangular process occurs on the opposite side of the medial process of the atlantal neurapophyses of SMA 0011, projecting laterally. The posterior wing of the neurapophysis does not taper as in Kaatedocus siberi (Tschopp & Mateus, 2013b), but remains subrectangular with a widely rounded distal end. This morphology was proposed as an unambiguous autapomorphy for the genus Galeamopus by Tschopp, Mateus & Benson (2015), but is also present in the dicraeosaurid Amargasaurus cazaui MACN-N 15 (Paulina Carabajal, Carballido & Currie, 2014). However, the wide distal ends of the neurapophyses remain autapomorphic for Galeamopus within Diplodocidae.

Axis. The axis of SMA 0011 (Fig. 19) has a closed but still slightly visible neurocentral synostosis, and unfused cervical ribs. The centrum is opisthocoelous. The pleurocoel extends over almost the entire centrum, and contains horizontal ridges at its anterior and posterior end. No vertical subdivision of the pleurocoel occurs. Anteriorly, the pleurocoel extends onto the dorsal surface of the parapophysis. The ventral surface of the centrum bears a distinct longitudinal keel medially, which widens anteriorly and posteriorly, where it also becomes rugose. The centrum is diagenetically transversely compressed ventrally, but it is clear that the ventral surface was constricted at midlength, and it appears that the wider posterior part of the ventral keel was laterally accompanied by shallow depressions. The parapophysis is rounded and faces anterolaterally and slightly ventrally. The diapophysis projects somewhat posteriorly, but does not bear a distinct posterior process. The neural arch is high and weakly posteriorly inclined. The prezygapophyses are not preserved. The only well-defined laminae are the PODL and the PRSL. The PRSL is slightly expanded transversely at its anteroventral end, similar to, but not as distinct as in AMNH 969 (Tschopp, Mateus & Benson, 2015). In lateral view, the PRSL is slightly concave ventrally, and straight in the upper part. The spine top is rugose, weakly expanded transversely, and situated entirely anterior to the postzygapophyseal facets. This anterior placement of the summit is unusual for sauropods, but present in Diplodocus carnegii CM 84 (Hatcher, 1901). Unlike CM 84, however, the neural spine summit of SMA 0011 has a posterior projection, similar to the condition in Giraffatitan (Janensch, 1950). The margin of the SPOL is strongly concave in lateral view, becoming vertical in the upper part. Small epipophyses are present laterally above the postzygapophyses. They do not project posteriorly. A large rugose area is present on the lateral side of the spine, slightly above mid-height (Fig. 19). It is subtriangular, broader towards the SPOL, with a pointed, elongate tip towards the center of the SDF. This rugosity could be homologous to the distal lateral expansion in the axis of Camarasaurus and Suuwassea (Madsen, McIntosh & Berman, 1995; Harris, 2006b), but the neural spine top is much more elevated in SMA 0011. Such a rugosity appears to be absent in the axis of Diplodocus carnegii CM 84 (Hatcher, 1901). The postzygapophyses of the axis of SMA 0011 slightly overhang the centrum posteriorly, and bear subtriangular facets with a straight anterior border.

Postaxial cervical vertebrae (Figs. 20–32). The cervical centra are all opisthocoelous and relatively elongate. As is typical for nearly all sauropods, the most elongate elements are the mid-cervical vertebrae (Table 2). All cervical centra have well-developed pleurocoels extending over almost the entire length of the centrum, also invading the dorsal surfaces of the parapophyses. The internal structure of the pleurocoel varies along the column: the anterior and posterior horizontal ridges described in the axis disappear by CV 4 and are present in only the right pleurocoel in CV 3 and 4 (Figs. 20 and 21). A vertical subdivision into anterior and posterior pneumatic fossae becomes visible in CV 3, and is pronounced from CV 5 backwards (Fig. 22). The subdividing ridge is oriented anterodorsally-posteroventrally, as in most sauropods. The posterior pneumatic fossae of CV 5–7 bear a large, slightly ellipsoid foramen at their anterior end, which pierces the median wall (Figs. 22–24). Whereas the median wall is thin posterior to this hole, it is transversely expanded anterior to the hole. The wider anterior margin of the hole bears a vertical groove that leads into a pneumatic foramen on the posterior face of its expanded portion (Fig. 23). Such a hole in the median wall is extremely rare in sauropods. Diplodocus carnegii CM 84 was reported to have confluent pleurocoels in posterior cervical vertebrae (Hatcher, 1901), “Morosaurus” agilis USNM 5384 shows this peculiarity in CV 3 (Gilmore, 1907), and a Camarasaurus axis has the same feature (AMNH 5761/X1, Osborn & Mook, 1921: pl. LXVII). Deep pneumatic openings are also present in mid-cervical centra of Galeamopus hayi HMNS 175, but these were left filled with sediment, and it remains unclear if these pierce the median wall or not (E Tschopp, pers. obs., 2010). The posterior pneumatic fossae of CV 5 and 6 of SMA 0011 become pointed posteriorly, due to the development of a shallow posteroventral fossa, which diagnoses most diplodocines (except Kaatedocus; Tschopp & Mateus, 2013b). From CV 6 backwards, the anterior pneumatic fossa becomes subdivided by a horizontal ridge at about mid-height. The ventral portion of the anterior fossa becomes vertically divided in CV 9 (Fig. 26). The latter is also the first element in the series to show a separation of the posterior-most portion of the posterior pneumatic fossa. Additionally, CV 10 has a horizontally subdivided posteroventral fossa (Fig. 27). In the first element of the articulated transitional series, the pleurocoel becomes less complex again (Fig. 28).

In the first preserved posterior cervical vertebra, the anterior condyle is damaged, so that it reveals the internal structure. The condyle is composed of large internal cavities, surrounded by 2–4 mm thick, relatively dense bony struts. The arrangement appears symmetric, with a subtriangular cavity dorsomedially, and two subcircular cavities following on both sides (Fig. 28).

The parapophyses become slightly anteroposteriorly elongate in CV 3 and 4. These structures project ventrolaterally in all elements, but not to the degree present in Apatosaurinae (Gilmore, 1936; Upchurch, Tomida & Barrett, 2004; Tschopp, Mateus & Benson, 2015). The anterior surface of the parapophyseal ramus, which connects the facets and the anterior condyle, is very distinct, dorsoventrally expanded, and rugose. The fossa on the dorsal surface of the parapophysis is subdivided by a short, oblique ridge in CV 6 and more posterior elements. In CV 9 and 10, the parapophyseal facet is subtriangular, anteroposteriorly elongated, and wider posteriorly than anteriorly.

The ventral surface is hourglass-shaped and narrow in anterior and mid-cervical vertebrae, but becomes relatively wide in more posterior elements. The ventral surfaces of CV 3 and 4 bear a distinct longitudinal keel on their anterior halves, with prominent pneumatic foramina lateral to the keel in CV 3, and less prominent ones in CV 4. In CV 3, a shallow ventral ridge also occupies the posterior end. The ventral surfaces of CV 5 and more posterior vertebrae are concave without any traces of ridges or pneumatic foramina. Posteriorly, the ventral surfaces are bordered by distinct posteroventral flanges. These flanges become rugose ventrally in the posterior cervical vertebrae.

None of the centra are fused with the corresponding cervical ribs. The neurocentral synostosis is closed but visible in the anterior and posterior cervical vertebrae, whereas in posterior mid-cervical vertebrae it is completely open. Where it is closed, the zigzagging neurocentral synostosis is more visible anteriorly than posteriorly (Fig. 29). In the most anterior and posterior elements, the synostosis becomes extremely faint to completely obliterated posteriorly. It lies on top of the centrum, such that the entire pedicels of the neural arches are detached in the unfused elements. The synostosis line is highest in the anterior half and descends anteriorly and posteriorly.

The neural arch is high in anterior cervical vertebrae, but becomes lower posteriorly. In all elements, it appears very fragile and slender, with very thin but distinct lamination. In posterior cervical vertebrae, the neural arch is somewhat displaced anteriorly, reaching close to the anterior condyle, but being well distant from the posterior edge of the centrum. The displacement reaches its maximum in the posterior-most cervical vertebrae.

The prezygapophyses project anteriorly and slightly dorsally in most elements. Close to the cervico-dorsal transition, they become more elevated. They bear suboval facets in CV 3, with the long axis extending anteroposteriorly. From CV 4 onwards, the facets become subtriangular, with the tip located medially. The facets are transversely convex as in all diplodocines (McIntosh, 1990b; Wilson, 2002; Whitlock, 2011a). Only in CV 5 are they concave, but this appears to be due to taphonomic distortion. In CV 7 and 8, the articular facets are elevated on pedestals, but no transverse sulcus is present posteriorly, unlike in Kaatedocus (Tschopp & Mateus, 2013b). The prezygapophyses cap the PRCDF dorsally, which in CV 5 and 6 is subdivided by a vertical accessory lamina connecting ACDL and PRDL right at the diapophysis. Anteriorly, the prezygapophyses are ventrally supported by the CPRL, which is single in anterior cervical vertebrae. From CV 7 backwards, the CPRL is divided, with one distinct and few short, weak accessory lamina in the PRCDF. The accessory laminae subdividing the PRCDF become stronger in more posterior elements. The anterior-most portion of the lateral surface is marked by distinct pre-epipophyses in CV 4 and more posterior elements, however, they only extend considerably anterior to the prezygapophyseal facet in CV 9 and 10. Posteriorly on the prezygapophyseal process, the anterior portion of the SDF develops a deep, but not well defined fossa in CV 3.

In anterior cervical vertebrae, the SPRL is distinct on the prezygapophyseal process, disappears around midlength of the dorsal portion, and becomes visible again on the spine top. In mid-cervical vertebrae, the SPRL is weak to almost absent on the prezygapophyseal process, as is typical for Diplodocinae (Tschopp & Mateus, 2013b). Thus, the SPRF and SDF of anterior and mid-cervical vertebrae are not distinctly separated from each other at the base of the neural spine. In posterior cervical vertebrae, the SPRL is distinct. Due to a backwards curve of the spine top in anterior cervical vertebrae, the SPRL has a somewhat sinuous appearance in lateral view in these elements. Below the backwards curve, the SPRL extends almost vertically in CV 3 to 6, becomes slightly posteriorly inclined in CV 7 and 8, and anteriorly inclined in CV 9 and more posterior vertebrae. A PRSL is present at the base of the neural arch in unbifurcated spines, which reach back to CV 7.

The diapophysis is entirely located in the anterior half of the vertebra. The transverse processes of SMA 0011 do not form such distinct posterior processes as those present in Kaatedocus (Tschopp & Mateus, 2013b). The diapophysis is supported by distinct ACDL, PRDL, PODL, and PCDL. The ACDL and PRDL are separated along their entire length, a feature typical for apatosaurines, and usually absent in diplodocines (Tschopp, Mateus & Benson, 2015). The PCDL is almost horizontal, and the PODL steeply inclined in CV 3, but in CV 4 and more posterior elements, they approach each other, forming a more acute angle anteriorly. In anterior elements, the PODL and PCDL unite before curving laterally, but more posteriorly they remain separate as the ACDL and PRDL, and the POCDF is therefore extended onto the posterior surface of the diapophysis. In CV 5, two nearly parallel PCDL occur: the ventral one connects to the ventral edge of the diapophysis, whereas the dorsal one joins the PODL anteriorly (Fig. 22). In more posterior elements, the PCDL bifurcates anteriorly (Figs. 23 and 24). In the posterior CV, this bifurcation is very strong, with the more dorsal branch being nearly vertical (Figs. 25–27). The CDF lies directly ventral to the diapophyseal process. In the posterior cervical vertebrae of SMA 0011, a short but stout accessory lamina occupies the posterior portion of the fossa. In CV 10, there is a vertical accessory lamina posterior to the dorsal branch of the PCDL, subdividing the POCDF (Fig. 27). Dorsomedial to the accessory lamina, the POCDF is pierced by a large foramen, such that the POCDF is interconnected with the SPOF (Figs. 27 and 30). A similar state appears to be present in the anterior cervical vertebrae of Dicraeosaurus hansemanni MB.R.4886 (E Tschopp, pers. obs., 2011), a partial mid-cervical vertebra of Suuwassea emilieae ANS 21122 (Harris, 2006b: fig. 8B), and Brontosaurus yahnahpin Tate-001, but in these taxa, the borders of the opening seem to be broken. Fossae at the same location occur in many taxa, including Diplodocus or Supersaurus (Hatcher, 1901; E Tschopp, pers. obs., 2013), but none of them opens up into a large foramen as in SMA 0011 (Fig. 30).

The SDF is of generally simple morphology. In CV 5 and 6, a shallow but dorsally well delimited fossa is located close to the spine summit. In CV 6 and 7, the SDF bears a distinct, dorsoventrally elongate fossa posterolateral to the SPRL, at about mid-height of the metapophysis. From CV 7 backwards, a vertical accessory lamina follows the SPRL posteriorly, as in Diplodocus carnegii CM 84 (Hatcher, 1901). No subfossae are present in the SDF of posterior cervical vertebrae, but in mid- and posterior cervical vertebrae, the SDF becomes clearly delimited dorsally, just below the anteroposterior narrowing of the spine top. CV 10 furthermore bears a stout, slightly anteriorly inclined lamina where the SDF is deepest, but the lamina does not connect to any surrounding lamina.

The neural spine undergoes distinct changes in development and orientation from anterior to posterior. In anterior cervical vertebrae, it is vertical, and dorsoventrally tall, reaching well above the postzygapophyses. The axis, as well as CV 3 and 4 have a distinctly posteriorly curving spine summit, as can also be seen in the corresponding elements of Brontosaurus yahnahpin. There is an abrupt change in height from CV 5 to 6, resulting in a smaller total height of CV 6 compared to CV 5. Such a development has only been described in Dicraeosaurus (Janensch, 1929), but neural spines are often incomplete, where anterior cervical vertebrae have been found (e.g., Diplodocus carnegii CM 84, Apatosaurus louisae CM 3018; Hatcher, 1901; Gilmore, 1936), which makes a thorough assessment of this character difficult. However, SMA 0011 is clearly different from the state in Kaatedocus siberi AMNH 7530 and SMA 0004, in Barosaurus sp. AMNH 7535, and in the indeterminate diplodocine CM 3452, where the anterior cervical neural spines are low, and total vertebral height continuously increases throughout the vertebral column (Tschopp & Mateus, 2013b; Tschopp, Mateus & Benson, 2015). From CV 6 backwards, the cervical neural spines of SMA 0011 decrease in relative height, compared to pedicel height (Table 2), and become anteriorly inclined. Towards the cervico-dorsal transition, neural spine height increases again, such that the posterior cervical vertebrae have highly elevated spine summits. In the first two vertebrae of the transitional block, the spine summits are most strongly anteriorly inclined, and the dorsal-most parts of the neural spines are anteroposteriorly short but elongated dorsoventrally. Bifurcation of the spine is present only from CV 8 backwards, which is more posterior compared to Diplodocus or Apatosaurus (Wedel & Taylor, 2013), but not as posterior as in Barosaurus (McIntosh, 2005). Unbifurcated neural spines slightly expand transversely towards their distal end, similar to the state in Suuwassea emilieae (Harris, 2006b). Posteriorly, the SPOLs are thin but project far posterodorsally, and connect to each other across the spine summit. Therefore, they enclose a distinct, wide and deep SPOF. Elements with bifid neural spines have a median tubercle. The lateral surface of the neural spine summits becomes rugose in posterior vertebrae. CV 9 has a distinct dorsoventral ridge on the medial side of the metapophysis, which connects the summit with the median tubercle, as in Kaatedocus siberi SMA 0004 (Tschopp & Mateus, 2013b).

Following the changing orientation and elevation of the spine, the SPOL also has a quite variable morphology from anterior to posterior cervical vertebrae: the lamina is strongly concave in CV 3, and less so in CV 4, due to the more expressed backwards leaning of the spine top in CV 3. The SPOL is gently curved in CV 5, but strongly concave in CV 6, where it forms a 90° angle. Due to the increasing anterior inclination of the spine, the SPOL becomes more gently concave in CV 7 and more posterior elements. Its posterior portion, where it unites with the epipophysis, is almost horizontal. The epipophysis is well developed in all cervical vertebrae, often overhanging the postzygapophyses. It constitutes the posterior end of the SPOL, and is often pointed. The postzygapophyseal facets are suboval to subcircular in the anterior cervical vertebrae, but become subtriangular more posteriorly, with the tip pointing medially. They are concave and thus face both downwards and outwards. They are ventrally supported by a vertical, single CPOL.

Penultimate and posterior-most cervical vertebra

Preservation. The two posterior-most vertebrae are the second and third elements in the block preserving the cervico-dorsal transition. They are still embedded in matrix, and only the right sides are prepared (Figs. 31 and 32). The diapophysis is not preserved in either vertebra, and the posterior-most element also lacks the right metapophysis and postzygapophysis. The anterodorsal part of the right lateral surface of the centrum of the posterior-most vertebra is reconstructed, including the neurocentral synostosis.

Compared to more anterior cervical vertebra, the two posterior-most vertebrae have a considerably taller diapophysis, and less distinct epipophyses. Their centra are opisthocoelous and have an intermediate elongation compared to more anterior cervical vertebrae and the first dorsal vertebra (Table 2). The lateral surface is marked by elongate pleurocoels that occupy the central and anterior portion of the centrum. In the posterior-most element, the pleurocoel is more restricted towards the anterior than in the penultimate one, being almost entirely situated above the parapophysis. The parapophysis lies ventrolateral to the pleurocoels, which extend onto its dorsal face. Posteroventral flanges are present, but become less distinct in the posterior-most centrum. The ventral surface is transversely concave and broad, with a shallow longitudinal ridge located anteriorly.

The neural arch height above the synostoses is more or less equal to centrum length, not counting the condyle (Table 2). As in anterior and posterior cervical vertebrae, the neurocentral synostosis is closed, but still visible in its anterior half. The neural spine is divided. The prezygapophyseal facet is broad, and projects slightly anterior to the condyle in both vertebrae, although the ramus is more vertically oriented in the posterior-most cervical vertebra than in the penultimate one. A weak pre-epipophysis is present, but does not extend beyond the prezygapophyseal facet. The SPRL is strongly concave, due to the strong anterior inclination of the spine top. The PRDL does not contact the ACDL directly, but they are interconnected by a vertical lamina below the diapophysis. The latter is thus slightly elevated above the centrum, and dorsoventrally high. The broken diapophysis of the posterior-most element reveals large open spaces internally that are surrounded by narrow laminae of relatively dense bone tissue. Both the ACDL and the PCDL are only slightly inclined, and connect to the ventral most part of what would have been a dorsoventrally tall diapophyseal ramus before it was broken off. The POCDF is subdivided by a strong, laterally facing, almost vertical accessory lamina, forming a posteroventral branch of the anterior end of the PODL. This differs from more anterior elements, where the accessory lamina in the POCDF faces posteriorly. Unlike the mid-cervical vertebrae, the posterior elements do not have any fenestra connecting the POCDF with the SPOF. The spine summits are anteroposteriorly narrow, and inclined anteriorly, but the inclination decreases in more posterior elements. The lateral surface of the spine is marked by the SDF, which is well delimited dorsally, similar to the state in the first posterior cervical vertebra. From the top of the SDF, the spine of the posterior-most elements forms a narrow anterodorsal projection. The medial surface of the spine is slightly anteroposteriorly convex and smooth.

Dorsal vertebrae (Figs. 33–37; Table 3)

Preservation. The dorsal series of SMA 0011 was found in two parts, with one and a half dorsal vertebrae preserved with the neck vertebrae, and the posterior-most six elements preserved with the appendicular material. A third block including three anterior to mid-dorsal vertebrae with associated dorsal ribs was collected from a position between the two main parts as described above (Fig. 3), and was initially included as part of the specimen. However, these most probably do not belong to the holotype specimen due to different size, preservation, and an apparently older ontogenetic stage (based on neurocentral closure patterns).

Dorsal vertebrae 1 and 2. Both elements are broken and deformed such that it is difficult to understand their morphology in detail (Figs. 33 and 34). The first dorsal vertebra lacks the right diapophysis and neural spine, such that the medial surface of the left metapophysis is visible in the mount (Fig. 33). The dorsal portion of the centrum and ventral half of the neural arch are crushed, and various pieces of each became intermingled. The second dorsal element preserves a very deformed, anterior half of the centrum, which is not fused with the neural arch (Fig. 34). A part of the neural arch is severely crushed and intermingled with the fractured pieces of the first dorsal vertebra (Fig. 33).

The dorsal vertebrae are considerably shorter than the posterior-most cervical elements, but remain of about the same length along the dorsal column (not considering the condyle; Table 3). The first dorsal vertebra has a strongly opisthocoelous centrum, whereas DV 2 is only slightly opisthocoelous. Both vertebrae bear distinct pleurocoels on the anterodorsal corner of their lateral sides. These pleurocoels are shorter than the ones of the posterior-most cervical elements, and excavate the neural arch pedicels internally. The position of the parapophysis is difficult to see in both elements, but appears to be still on the centrum in DV 1 (anterodorsal to the pleurocoel), whereas the centrum of DV 2 does not show any traces of a parapophysis. The ventral side of DV 1 is well delimited by posterior ridges between the lateral and ventral surfaces. A broad, but distinct midline ridge marks the anterior half of the ventral side of the first dorsal centrum. The articulation surface of the second centrum for the neurocentral synchondrosis is broad and curved. The neural canal is narrowest at midlength of the centrum. The internal structure of the centrum consists of large chambers, separated from each other by thin, well-defined laminae, which are not symmetrical (Fig. 34D).

The neural arches of the dorsal vertebrae are higher, but more anteroposteriorly compressed, than in the posterior-most cervical elements. The prezygapophysis is relatively short. The SPRL is oriented almost vertically, and no strong anterior inclination of the neural spine is present anymore. The medial side of the first dorsal neural spine is gently convex, and slightly longer anteroposteriorly than in the posterior-most cervical vertebrae. Postzygapophyses are not preserved.

Mid- to posterior dorsal vertebrae (probably DV 5–10). Dorsal vertebra 5 lacks its right neural arch, diapophysis, and spine (Fig. 35). Dorsal vertebra 6 lacks the anterior part of the centrum, the right diapophysis, parapophysis, and prezygapophysis, and the spine top. In dorsal vertebra 7, the right diapophysis, parapophysis, and the spine top are missing. Dorsal vertebrae 8 and 9 lack the right diapophysis and parapophysis. The last dorsal vertebra lacks the neural spine process, whereas the arch below the postzygapophysis, the diapophysis, and the prezygapophyses are preserved (Fig. 36).

The mid- and posterior dorsal centra are short, and generally amphiplatyan to amphicoelous. Only DV 5 shows a weak anterior condyle. The pleurocoel is largest in DV 6–8, occupies the dorsal half of the centrum and extends slightly onto the pedicels, below the neurocentral synchondrosis. The ventral surface is convex, and not well separated from the lateral side. The centrum is slightly shorter ventrally than at mid-height. In DV 6 and 7, a zigzagged line marks the neurocentral synostosis at the dorsal edge of the centrum. Dorsal vertebrae 8–10 have the centra and neural arches detached, but no obvious articulation surface is visible on either element, indicating that that closure has initiated but not entirely completed, such that centra and neural arches could be detached easily. The neural arch is high, with highly elevated postzygapophyses, resulting in longer pedicels than neural spines in at least DV 5–8. Pre- and postzygapophyses are on more or less a horizontal line. The pedicels below do not show a strong lamination, but the ACPL, PCDL, and CPOL can be well distinguished. Dorsal vertebrae 6–9 furthermore show a weakly developed PCPL. An accessory lamina can be found in DV 7, connecting the PCDL with the PODL, and in DV 8 between the PRPL and the PRDL. Only a single hyposphene is visible (in DV 5), relatively long dorsoventrally, and transversely expanded ventrally, resulting in a high and narrow trapezoid. The width of the ventral end (39 mm) is slightly more than twice the minimum width of the hyposphene (16 mm). The posterior surface of the hyposphene is transversely concave. It is ventrally supported by a single, vertical lamina. The parapophysis lies at mid-height on the pedicels in DV 6, at two thirds in DV 7 and at three fourths in DV 8. More posteriorly, the parapophysis seems to have been attached to the prezygapophysis.

A single transverse process is preserved completely (the left of DV 5; Fig. 37). It projects more or less straight laterally (although it was reconstructed as being strongly inclined, see Fig. 35), curving very gently ventrally towards its distal tip. The process is widest dorsally, and dorsoventrally concave both on its anterior and posterior sides. The diapophyseal facet points ventrolaterally and is strongly expanded posteriorly.

The spine is relatively low in DV 5–8, and only in DV 9 and probably 10 does it exceed the pedicel height. The spines are situated above the posterior-most portion of the centrum, and are vertically oriented. This differs from the strongly anteriorly inclined posterior dorsal neural spines of Diplodocus (Hatcher, 1901; Gilmore, 1932). The SPRL is vertical in DV 6, strongly dorsoventrally convex in DV 7 and 8, and slightly convex in DV 9. The SPDL is short and only expressed at its ventral end. Dorsally it merges with the SPOL, which extends onto the lateral surface of the spine. The POSL, or possibly medial SPOL, is straight and vertical. Due to the preservation and mounting (partly embedded in matrix), it cannot be distinguished at this point how far back the bifurcation proceeds. The last definitively bifid neural spines are present in DV 5.

Ribs

Cervical ribs (Figs. 38–49; Table 4). The cervical ribs are thin, fragile elements. None of them are fused with their respective centra. They are composed of a rib shaft, an anterior process, and the capitulum and tuberculum. The ribs are concave internally, with a lamina connecting the tuberculum with the capitulum internally, producing two separate fossae anteriorly and posteriorly.

The axial cervical rib has almost no tuberculum and is thus a straight, elongate, and dorsoventrally compressed sheet of bone, which becomes slightly higher around midshaft but tapers again posteriorly (Fig. 38). The capitulum is not offset from the posterior shaft, and faces anteromedially. The capitular facet is much longer than wide, such that it articulates with both the axial parapophysis, and to a small extent also with the posteroventral projections of the atlas.

Anterior to mid-cervical ribs are longer than their corresponding centra, unlike the situation in Apatosaurus louisae CM 3018 (Gilmore, 1936), but they overlap only a small portion of the following vertebra. The anterior process is distinct but very short in CR 3, and pointed in CR 3–5 (Figs. 39–41). This process becomes very broad and rounded anteriorly in mid- and posterior cervical ribs (Figs. 42–49). At the base of the anterior process, mid- and posterior cervical ribs bear a dorsal lamina, which connects the capitulum with the tubercular edge of the anterior process (Fig. 45). Thereby, it forms the anteromedial rim of a deep triangular fossa, which is otherwise bordered by a transverse lamina between capitulum and tuberculum and the lateral margin of the anterior process. This fossa is further subdivided by a second oblique ridge, parallel to the first, in posterior cervical ribs (Fig. 47B). The tuberculum is posteriorly inclined in anterior cervical ribs, and triradiate in cross-section at midlength. The three axes are oriented anteriorly, posteriorly, and medially. The tubercular facet is generally wider than long. The capitulum bears a pneumatic foramen dorsally, posterior to the origin of the lamina connecting the capitulum with the tuberculum (Fig. 42). The capitular facet is ovoid in CR 3, with the wider end anteriorly. It becomes subrectangular to reniform in more posterior ribs, with the longer axis being oriented anteroposteriorly, and the sometimes concave margin being the dorsolateral one. The ventral surface of the cervical rib is marked by striations (Fig. 46), probably for muscle or tendon insertions.

The two posterior-most cervical ribs (Figs. 48 and 49) bear progressively shorter anterior processes, compared to more anterior cervical ribs. The dorsal oblique lamina disappears, and also the transverse lamina connecting capitulum and tuberculum becomes less pronounced. The angle between capitulum and tuberculum widens considerably, approaching 90° in the posterior pair. The posterior process shortens and tapers strongly. A distinct longitudinal ridge marks the ventral surface, as in anterior dorsal ribs. One right posterior cervical rib (field number M 6/16-3) has a pronounced, anteriorly projecting spur close to the origin of the transverse lamina on the capitulum, which might be an ossified tendon insertion, and is absent on the left element of the pair (Fig. 48). The pneumatic fossa on the capitulum is reduced in the first pair of posterior-most cervical ribs, and totally absent in the second pair. The capitular facet becomes ovoid again, resembling the shape of the facet in CR 3. In the posterior-most pair of cervical ribs, the capitular facet is nearly circular, and supported by a strong, subtriangular capitular neck. The tubercular facet is longer than wide, and thus resembles rather dorsal ribs than cervical elements. In the posterior-most cervical ribs, the posterior process does curve slightly downwards, and not strictly posteriorly as in more anterior elements.

Dorsal ribs (Figs. 50–56; Table 5). Several ribs have been recovered associated with the dorsal series, but whereas the sequence from anterior to posterior appears relatively clear, based on the quarry position, the exact position of the single elements can only be confidently determined for some elements.

DR 1 has a capitulum and a tuberculum which stand in a right angle to each other (Figs. 50 and 55). The anterior surface of the rib head bears a distinct, narrow, proximodistal ridge, which originates from the tubercular facet and extends in a nearly straight line distally onto the rib shaft (Fig. 50), where it fades out. At the base of the capitulum, a broader, slightly less distinct ridge separates from the narrow one and curves for a short distance onto the anterior surface of the capitulum, joining its proximal edge at about midlength (Fig. 50). Both the tubercular and capitular facets are anteroposteriorly compressed, rugose articular surfaces. The posterior surface of the capitulum is flat. The posterior surface of the tuberculum is marked by two longitudinal ridges: a longer, narrower medial one, and a shorter and broader lateral one (Fig. 50). Together, they form a distinct proximal fossa just below the tubercular facet, which fades out more distally towards the rib shaft. The rib shaft has a V-shaped cross-section at its base and flattens distally. The anterior side changes from being distinctly convex (due to the presence of the proximal longitudinal ridge) to even slightly concave once the ridge disappears. The distal end of the shaft tapers nearly to a point, and is marked by a sharp longitudinal ridge on the posterior surface, which extends from below midshaft to the tip and thus creates a distinctly triangular cross-section, with an inverted orientation compared to the cross-section at the base of the shaft.

Dorsal rib 2 has a much shorter tuberculum, which is mainly due to the fact that the bony shelf connecting capitulum and tuberculum is more extensive in this element compared to the first dorsal rib (Fig. 51). The longitudinal ridge on the anterior surface of DR 2 is less pronounced and wider than in DR 1, and no perpendicular ridge occurs at the base of the capitulum. Also the short, longitudinal ridges on the posterior surface of the tuberculum of DR 1 do not occur on DR 2, so that the rib head is uniformly concave posteriorly. When articulated with the dorsal vertebra, the shaft of DR 2 curves backwards and tapers until about midlength. From here, the anterior and posterior edges remain subparallel, just to minimally expand distally towards the distal-most tip.

A probable DR 3 preserves only the shaft, which is wider and more triangular than circular in cross-section. The distal end is expanded (Fig. 52).

More posterior ribs continue the trends observed from DR 1 to DR 3. The shape of the rib head changes such that the capitulum projects obliquely dorsomedially instead of perpendicular to the long axis of the shaft. The capitular facet becomes gradually stronger throughout the series, whereas the tuberculum becomes shortened. The rib head thus has a subtriangular shape in axial view in more posterior elements. In at least the last three dorsal ribs (but maybe additional posterior dorsal elements are lacking), the capitulum curves dorsally at its end, such that the capitular facet comes to face dorsomedially instead of more strictly medially as in more anterior ribs. The relatively thin sheet of bone between capitulum and tuberculum remains flat internally throughout the entire series (contrary to the state in most other diplodocines, in which this area is marked by an oblique accessory ridge; Tschopp, Mateus & Benson, 2015). None of the ribs bear pneumatic foramina. The shafts are marked by a longitudinal groove on the lateral edge in mid- to posterior dorsal ribs, and have an ovoid to slightly subtriangular cross-section, with a transversely oriented long axis, and a slightly more angular posterior surface. The last three or more dorsal ribs decrease significantly in shaft width, compared to more anterior elements, and obtain a subcircular cross-section similar to DR 1.

The left dorsal rib 2 bears bite marks on its distal end (Fig. 56). The bite marks are eleven parallel, slightly curved grooves on the external side of the rib, which extend from the posterior edge anteroventrally. The distance between the marks on the posterior edge varies from 16 to 26 mm, with a mean distance of 20.75 mm.

Sternal ribs (Figs. 57 and 58). Several morphotype C elements (sensu Tschopp & Mateus, 2013a) were recovered associated with SMA 0011. They are rod-like, narrow bones (Fig. 57). Some have a rather circular, and others a laminar cross-section, and all have smooth margins. A single, flattened morphotype E element (field number M5/4-2) is expanded on one side, where it has rugose margins (Fig. 58). No additional information can be gleaned to date that would help to confirm or discard the interpretation of Claessens (2004) and Tschopp & Mateus (2013a) that these elements are sternal ribs.

Forelimb (Figs. 59–67; Table 6)

Scapulae. Both scapulae lack the dorsal part of the acromion and of the distal end of the blade (Fig. 59). The acromion and the blade form an acute angle, but the acromial ridge is only very slightly developed. The lateral surface anterior to the acromial ridge is concave. Medially, the acromion is concave. The glenoid surface is transversely concave and faces slightly more medially than laterally. The surface is widest anterodorsally, close to the articulation with the portion of the glenoid on the coracoid, and tapers posteroventrally. The ventral edge is mostly straight, and does not bear a triangular process as present in some Camarasaurus specimens, or Dystrophaeus (Osborn & Mook, 1921; McIntosh, 1997). The distal end of the blade is slightly expanded ventrally as in Brontosaurus excelsus YPM 1980 (Upchurch, Tomida & Barrett, 2004). The dorsal, or acromial edge of the scapula is much more concave than the ventral one, due to the stronger extensions of the dorsal portion of the acromion and the indicated, wider distal expansion of the shaft, which starts more anteriorly on this edge than on the ventral one. No oval rugose tubercle is present on the base of the shaft, unlike in Brontosaurus excelsus YPM 1980 (Upchurch, Tomida & Barrett, 2004), although a slightly elevated structure occurs in the left scapula.

The left scapula bears distinct bite marks medially along the broken posterodorsal edge of the acromion. There are at least ten subparallel grooves oriented perpendicular to the broken edge, and varying in length from 19 to 73 mm. Also on the lateral side, the left scapula bears short, subparallel grooves, which mark the slightly elevated structure at the base of the shaft. Seven grooves are present. Given that this structure was probably the attachment site for soft tissue (the M. scapulohumeralis cranialis, according to Remes, 2008), the theropod might have bitten only there in order to detach the muscle from the bone.

Coracoid. The right coracoid is preserved, which is only observable in lateral view due to the way it is mounted. The coracoid is somewhat tear-drop shaped (Fig. 59), with a concave anterodorsal edge, and a strongly, continuously convex, narrow dorsal margin, unlike the squared coracoids of apatosaurs (Riggs, 1903; Bakker, 1998). The coracoid foramen is completely enclosed, but the coracoid is not fused with the scapula. The bone is gently convex dorsoventrally. It curves slightly medially at its anterior margin. No distinct notch is present anterior to the glenoid surface. The glenoid is strongly transversely expanded at its center, and tapers posterodorsally and anteroventrally. The articular surface is barely visible in lateral view. The glenoid surface and the articulation surface with the scapula enclose an angle of about 155°.

Humeri. The humeri are both complete but slightly compressed anteroposteriorly, the right humerus more so than the left (Figs. 60 and 61). The humeri are widely transversely expanded at their proximal ends, both laterally and medially. The distal ends are expanded as well, but less so. The proximal portion of the anterior side is concave transversely. A small, rugose tubercle marks this concavity, as in most diplodocids (Tschopp, Mateus & Benson, 2015), but it is more laterally positioned compared to the apatosaur AMNH 6114 or Galeamopus hayi HMNS 175 (Fig. 60). The deltopectoral crest of G. pabsti SMA 0011 does not extend to midshaft (Table 6). Its distal end is distinct and follows the lateral margin. It is not transversely expanded as would be typical for titanosaurians (Wilson, 2002; Curry Rogers, 2005). The lateral surface of the crest is concave anteroposteriorly, but this depression is probably exaggerated taphonomically. The humeral head is well offset from the shaft and centrally located. The posterior surface is transversely convex in its proximal half, but becomes concave distally, where it develops a shallow intercondylar groove. Two ridges mark the distal end anteriorly, indicating the extensions of the medial and lateral condyles. The ridges are relatively well visible and extend proximally up the shaft (Fig. 61B). The medial condyle is much wider than the lateral one.

Ulna. The ulna lacks the proximal-most portion of the anterior arm of the condylar processes. The bone is strongly anteroposteriorly compressed in its proximal half (Fig. 62). It is generally slender, with a triradiate proximal end. The anterior arm is considerably longer than the lateral one, even though this is enhanced due to compression. The ulna has concave posterolateral and posteromedial surfaces. The lateral arm is somewhat wider than the anterior one. The distal part of the anterior surface bears two strong and elevated, longitudinal ridges. They both taper distally and proximally, and have a smooth surface. Proximally, the more lateral of the two ridges extends above midlength (Fig. 62D). Distally, the more medial ridge is more pronounced, reaching the distal articular surface. The distal end is expanded medially and somewhat anteroposteriorly. The articular surface is subrectangular in outline.

Radius. The radius is complete, but its proximal end is compressed, and the distal end taphonomically sheared such that the entire bone appears sigmoid (Fig. 63). The proximal articular surface has thus a narrow, ellipsoid outline, but would probably be slightly more subcircular if undeformed. The shaft is subrectangular in cross-section. As in the ulna, also the distal end of the radius is slightly expanded transversely. The posterior surface bears two longitudinal ridges on its distal portion for the articulation with the ulna. The lateral ridge is stronger and marks the posterolateral edge of the radius. It extends from the distal articular surface about one third up the shaft. The more medial ridge is weakly developed and shorter. It does not reach the distal articular surface. The distal surface is subrectangular, with convex medial and lateral margins and weakly concave anterior and posterior borders. The lateral half of the distal articular surface is beveled.

Carpal. The carpal is an irregular, relatively thick element (Fig. 64). It does not bear distinct articular surfaces, and was found slightly disarticulated, such that an orientation of the carpal within the manus was not possible to definitely confirm. Only one element was found. The entire bone is relatively rugose and was found between the radius and mtc I–III. This is the same arrangement as found in the articulated manus of the indeterminate diplodocine WDC-FS001A (Bedell & Trexler, 2005; Tschopp, Mateus & Benson, 2015), but different from apatosaurines, where the carpal overlies mtc II–IV (CM 3018 and UW 15556; Hatcher, 1902; Gilmore, 1936). If the orientation of the carpal did not change during diagenesis, the surface articulating with the radius is strongly convex transversely, but some abrasion has occurred, and the internal bone structure is visible both medially and laterally. It is therefore possible that the complete element would be more block-like in shape, as known from other diplodocine specimens (WDC-FS001A, Bedell & Trexler, 2005). It is relatively narrow anteroposteriorly at its medial end. The lateral side is about double the anteroposterior length, thanks to a distolateral, posteriorly projecting process. Anterior and posterior surfaces are fairly smooth. Distally, there are no distinct articulation surfaces for the metacarpals, unlike the state in Camarasaurus (Tschopp et al., 2015). The carpal of SMA 0011 is taller proximodistally than the elements known from the apatosaurines CM 3018 and UW 15556 (Hatcher, 1902; Gilmore, 1936).

Metacarpals. All metacarpals are complete and were found articulated. Metacarpal I was recovered flipped 180° such that the distal articular surface was at the level of the proximal articular surface of the remaining metacarpals. This displacement indicates that mtc I was not rigidly included in the columnar metacarpal structure adapted for weight-bearing. Given that digit I bears a large ungual, it did not have a primarily graviportal role, and was probably therefore not so strongly bound to the other metacarpals. The metacarpals of SMA 0011 are relatively elongate bones (Fig. 65), but less so than in Camarasaurus (Tschopp et al., 2015). Metacarpal III is the longest, followed by mtc II, IV, I, and V (Table 6). Metacarpal I and II have subrectangular to trapezoidal proximal articulation surfaces, contrasting with triangular ones in mtc III and IV.

Metacarpal I is relatively stout. The proximal surface is concave anteroposteriorly and flat transversely. It is slightly deeper laterally than medially. The lateral edge of the articular surface is strongly concave, whereas the medial one is somewhat convex in proximal view. The posterior surface bears two small but distinct nutritional foramina on the distal half. The distolateral portion of the shaft is crushed, resulting in a triangular lateral surface. The distal condyles are well separated from each other and anteroposteriorly convex. The lateral condyle is much longer proximodistally than the medial one. This results in a strongly inclined distal surface, such that the proximal phalanx projects posteromedially (and distally) in the articulated manus.

Metacarpal II has very distinct, straight anteromedial and anterolateral edges. The proximal and distal ends are slightly expanded in all directions. The proximal articular surface is wider anteriorly than palmarly and slightly convex. The shaft is thicker medially than laterally. The proximal portions of both the medial and lateral surfaces are concave, laterally more than medially. A slightly rugose, longitudinal ridge separates the medial from the palmar surface, and extends distally from the proximal end for about two thirds the length of mtc II. The distal surface slightly curves into the anterior surface. Its lateral and medial condyles are only visible in distal and posterior view. The medial condyle is larger than the lateral one.

Metacarpal III is the most elongate element of the manus. The proximal articular surface is subtriangular to ovoid. No distinct transition from the anterior onto the medial surface occurs on mtc III. The anterior and palmar faces unite laterally at a distinct ridge. The medial surface is concave proximally. The concavity is bordered by two distinct longitudinal, somewhat rugose ridges extending distally half way down the shaft. In the articulated manus, these ridges would face internally. The proximal and distal articular surfaces are slightly twisted. The distal surface is ovoid to subrectangular and does not extend considerably onto the anterior face. The articular facet is flat transversely and convex anteroposteriorly.

Metacarpal IV has a P-shaped proximal articulation surface, with a concave medial edge. As in mtc III, the shaft of mtc IV is twisted, and a distinction of the anterior face is not possible. A distinct ridge connects the posterior apex of the proximal articular surface with the posteromedial corner of the distal articular surface. The distal articular surface is subtriangular as well, with the apex anteriorly, and inclined medial and lateral edges. Two condyles are visible posteriorly. The apex of the distal articular surface curves onto the anterior face.

Metacarpal V is short and widely expanded both transversely and anteroposteriorly at its proximal end. It is somewhat drop-shaped in proximal view, with the tip facing palmarly. The shaft is twisted anti-clockwise, in proximal view. The medial surface is slightly concave for the reception of mtc IV. The free lateral face is gently convex. The medial and lateral surfaces meet at a ridge in their proximal halves. The distal end is partially reconstructed, but the preserved parts indicate that it is transversely expanded. A distinct concavity marking the posterolateral corner of the distal articular surface is of taphonomic origin, having collapsed while being closely attached to the distal articular surface of mtc IV during diagenesis.

Manual non-ungual phalanges. The manual non-ungual phalanges are relatively short and robust (Fig. 66). They are wider than long, as is typical for the eusauropod manus (Bonnan, 2003). The phalanges were found disarticulated, but closely associated with the metacarpals. A definitive assignation to distinct digits can be inferred for phm I–1 and II–1, but the identification of the other three non-ungual phalanges remains uncertain. Based on comparisons with the articulated manus of the Camarasaurus SMA 0002, we identified the elements as phm IV–1, V–1, and II–2. However, they could also be phm III–1, IV–1, and V–1, respectively. The latter arrangement would imply a clearly advanced stage in phalangeal reduction compared to Camarasaurus, but would be supported to some degree by the closer association of the nubbin-like phalanx with mtc IV and II than with mtc II or phm II–1. Nonetheless, given that the other phalanges were dislocated and scattered around the entire metacarpus, the burial location of the vestigial phalanx should not be taken as strong evidence for its articulated position.

The proximal surface of manual phalanx I–1 is concave anteroposteriorly. The phalanx I–1 has a concave posterior surface, with a proximally projecting palmar lip. Its medial surface is shorter than the lateral one, enhancing the angulation of the ungual phalanx even more. The lateral surface is concave proximodistally. The lateral extension of the posterolateral edge forms a thin, short crest (Fig. 66). Nothing similar is present in the manus of Camarasaurus (Osborn, 1904; Tschopp et al., 2015), but too few articulated proximal manual phalanges are known in diplodocids in order to decide if this might be autapomorphic in SMA 0011 or is instead more widespread within the clade. A phalanx figured by Jensen (1985: fig. 1E) appears to show a similar development of the posterolateral edge, but has not been identified below Sauropoda indet (Jensen, 1985). The phm I–1 of SMA 0011 has well-developed medial and lateral distal condyles with a distinct intercondylar groove occurring palmarly. The entire distal surface is subtrapezoidal, being longest palmarly, than medially, laterally and finally dorsally.

Manual phalanx II–1 has a concave proximal surface, which is oval in outline. It is only minimally wider than the shaft. The medial surface is broader, but shorter than the lateral one. The anterior surface is convex transversely. The posterior surface is marked by a bulge at the center of its proximal portion, and a pit distal to it. The distal articular surface is expanded transversely, and the condyles extend onto the medial and lateral surfaces. In anterior view, the distal surface is nearly flat, whereas in distal view, the palmar margin is concave.

Manual phalanx II–2 (as identified herein) is a vestigial, suboval bony nubbin. A distinct ridge separates the proximal and distal surfaces, which are convex and rough.

The probable manual phalanges IV–1 and V–1 are very similar, with IV–1 being slightly larger. They have concave proximal articular surfaces, transversely more so than anteroposteriorly. The surfaces are suboval in outline, and their anterior margins are pronounced laterally. The anterior surfaces are concave proximodistally, but slightly convex transversely. Medial and lateral surfaces are very narrow. The distal surfaces are without condyles. They have a continuous, rounded surface in dorsal view, which curves proximally at its medial and lateral end, almost reaching the proximal articular surface. The medial and lateral surfaces are thus practically nonexistent. The lack of medial and lateral condyles implies that these elements were the terminal phalanges of these digits.

Manual ungual. One ungual is present, situated on the first digit (Fig. 66). It is a long, high, and transversely compressed element. The proximal surface is ovoid, with a narrow anterior tip, and a widened palmar portion, where the articular surface lies. Above the articular surface, the proximal surface projects somewhat proximally, and is rugose. This rugosity extends as a short ridge posteriorly, onto the articular surface. The articular surface is inclined such that when articulated, the ungual would be slightly laterally deflected, compared to the long axis of the preceding phalanx. The medial surface is convex anteroposteriorly. A short groove marks the distal-most portion, which is slightly elevated (about 1 mm) above the more proximal portion of the claw, and shows a different surface texture (Fig. 67). The latter might represent fossilized remnants of the keratinous sheet covering the claw. The lateral surface is almost flat, with a long, proximodistally extending, straight groove covering the distal half of the surface. The palmar surface is strongly convex proximally and flat distally.

Hindlimb (Figs. 68–77; Table 7)

Terminology. The pubis and ischium are described as if they were oriented vertically and horizontally, respectively. The distal shaft of the ischium thus has long dorsal and ventral edges in lateral view.

Ilium. The right ilium is preserved, but was found in such a bad state that the medial side had to be covered immediately with plaster (B Pabst, pers. comm., 2014). Therefore, no morphological information can be gleaned from that side. The ilium lacks a large part of the posterodorsal portion of the iliac blade, and the distal-most end of the pubic peduncle (Fig. 68). The preacetabular process has a very pointed apex, which is directed anterolaterally, and is relatively broad transversely. The anterior portion is strongly concave, with the ventral margin facing ventrolaterally. The ventral preacetabular border and the pubic process form an angle of 90°. A triangular depression is located laterally at the base of the pubic process, with a horizontal and medio- and lateroventrally inclined sides. This is similar to the putative diplodocid ilium from Spain (CPT-1074; Royo-Torres & Cobos, 2004; E Tschopp, pers. obs., 2012), and has also been reported in other sauropod taxa (e.g., Cetiosaurus oxoniensis, Lirainosaurus astibiae, and Jobaria tiguidensis; Upchurch & Martin, 2003; Díez Díaz, Pereda Suberbiola & Sanz, 2013; Tschopp, Mateus & Benson, 2015). The pubic peduncle is distinctly concave transversely on its posterior face, but fractures indicate that the concavity is exaggerated and that the transverse width of the pubic peduncle would be slightly larger otherwise. The ischial tubercle faces ventrolaterally. The acetabular margin is thinnest just posterior to the pubic peduncle, and extends transversely towards the articulation surfaces of the ischium and pubis.

Pubes. Both pubes are almost complete, but lack a portion of the ischial articulation. The pubes are relatively slender (Fig. 69). The obturator foramen is completely enclosed and located in the proximal third of the ischial articulation. It is subtriangular in outline and oriented dorsomedially-ventrolaterally. Even though eroded, the anterodorsal corner of the pubis does not seem to bear a very pronounced, hook-like ambiens process, unlike the condition seen in Diplodocus or Supersaurus (Hatcher, 1901; Lovelace et al., 2007). This corner is laterally expanded, and from here, the pubis slightly tapers along the acetabular surface. The medial surface of the proximal half of the bone is proximodistally concave and transversely slightly convex. The latter convexity becomes more pronounced towards midlength, where the ventral margin curves back from the expanded ischial articulation to the narrow midshaft. The dorsal edge of the pubis is gently concave. Its anterior end is expanded both transversely and anteroposteriorly. The narrowest portion of the shaft lies below midshaft, at about two thirds of the entire length of the pubis. The ischiadic articulation is not preserved in its entire length, but broken surfaces indicate that a distinct ridge extended from the ischiadic facet along the ventromedial margin of the shaft to the distal articular surface. The reconstructed length of the ischiadic articulation is about 38% the total length of the pubis (Table 7). The distal end is convex, expanded anteroposteriorly, but not transversely. It is heavily rugose, concave laterally in distal view, and convex medially.

Ischium. The ischium lacks the distal half of the shaft (Fig. 70). It is mounted on plaster, such that only the medial view is accessible. Its proximal portion is wide and concave. The acetabular surface is inclined, such that the medial border forms a thin crest. This crest is relatively straight in medial view, but concave and curved in proximal view. Unlike the state in rebbachisaurids, the acetabular surface does not expand towards the articulation surfaces for the ilium and the pubis (Mannion et al., 2012). The iliac process has no distinct neck and is relatively narrow. The pubic articulation is much longer, and slightly convex in medial view. It curves slightly medially towards its ventral end. The shaft is weakly convex at its base, separating the concave acetabular portion from the again shallowly concave posterior shaft. The dorsal and ventral margins are parallel, only the posterior-most preserved portion of the dorsal edge indicates a slight dorsal expansion towards the end, as is typical for diplodocids (McIntosh, 1990a; McIntosh, 1990b; Upchurch, 1998; Wilson, 2002). No distinct ridges or scars can be seen on the internal surface.

Femur. The greater trochanter and the distal end are not preserved in the femur of SMA 0011 (Fig. 71). The medial edge is gently curved below the femoral head, not as distinct as in Dyslocosaurus (McIntosh et al., 1992). The head is separated from the shaft ventrally, but does not project far medially. It is slightly wider transversely than anteroposteriorly, and has a strongly rugose surface. The lateral margin of the shaft is slightly convex proximally, forming a weak lateral bulge, but no medial deflection of the proximal end occurs. The shaft is crushed at its center, but it is obvious that the medial side was anteroposteriorly wider than the lateral one. There is no indication of a large foramen opening at the center of the anterior surface, as occurs in some specimens of Diplodocus and Dicraeosaurus (Wilson, 2002; Whitlock, 2011a; Tschopp, Mateus & Benson, 2015), but some parts in that area are reconstructed. The fourth trochanter is entirely located on the posterior surface of the shaft, but close to the medial border proximally. The distal end of the fourth trochanter curves distinctly laterally towards the midline of the shaft. The fourth trochanter is medially accompanied by a shallow depression proximally and two rugose tubercles centrally and distally (Fig. 71D). The shaft is 1.5 times as wide as it is anteroposteriorly thick (Table 7). The more distally located tubercle of the two is the more developed. The preserved, distal-most part of the shaft slightly expands transversely.

Tibia. The tibia is complete, but compressed anteroposteriorly (Fig. 72). It is slightly expanded at both ends. The proximal end is longer transversely than anteroposteriorly, but this is partly due to taphonomic compression. The outline of the proximal articular surface is subrectangular as in apatosaurines, and unlike the subtriangular state as in diplodocines (Lovelace et al., 2007). However, it is unclear how much this shape is influenced by the compression. The cnemial crest is somewhat displaced distally, and is thicker distally than proximally. It projects laterally. Posterior to the crest, a fossa occurs for the reception of the fibula, which is posteriorly bound by a wide longitudinal ridge of about the same length as the cnemial crest. The lateral side of the shaft is much narrower than the medial one. A small convexity marks the distal end of the lateral edge. The distal articular surface has the typical step-like arrangement as in all sauropods, for the articulation with the ascending process of the astragalus.

Fibula. The fibula is a slender bone, with a strongly anteroposteriorly expanded proximal end, and a less expanded distal end (Fig. 73). The proximal end is transversely compressed. It has a pointed anterior end, which projects somewhat medially, similar to Diplodocus carnegii CM 94 (Hatcher, 1901). A distinct, but proximodistally short ridge extends from the posterior end down the shaft (Fig. 73B), for about 9 cm. The medial surface is marked by a subtriangular area with a striated rugosity, which covers about the proximal-most 20–25% of the shaft. The attachment site for the iliofibularis muscle, which is often called fibular trochanter, is situated slightly above midheight, as in Diplodocus (Whitlock, 2011a), and has an oval outline. Towards its distal end, the shaft expands more strongly transversely than anteroposteriorly, but the distal articular surface still remains anteroposteriorly longer than transversely wide. In particular the distomedial edge expands to articulate with the fibular facet on the astragalus. The distal articular surface has an oval outline.

Astragalus. The astragalus is wedge-shaped in both anterior and proximal views (Fig. 74). The anteromedial corner is reduced. Posteriorly, the astragalus is marked by a high ridge connecting to the ascending process. The latter extends backwards to the posterior end. The high, 42 mm wide ridge separates the two fossae for the articulation with the tibia medially and the fibula laterally. The ridge itself is slightly concave transversely, and bound by two distinct, dorsoventrally extending margins. The two margins end in two pronounced, bulge-like posteroventral expansions. The two expansions are separated by a strongly concave posteroventral margin in ventral view, similar to the condition considered autapomorphic in Janenschia robusta (Bonaparte, Heinrich & Wild, 2000). The tibial fossa is larger than the fibular fossa and subdivided by a shallow, oblique, anteroposteriorly oriented ridge into a medial and a lateral portion. The medial portion is pierced by three large foramina. The fibular fossa is relatively uniform, with the anterior edge forming a distinct lip-like lateral extension. The fibular fossa is thus visible in posterior view, a diplodocoid synapomorphy convergently acquired by Jobaria (Whitlock, 2011a). The distal roller is flattened due to compression, and appears to be subdivided horizontally into three distinct parts: an anteriorly facing portion, an anteroventral face, and a ventral part.

Pes. The pes was found associated with the astragalus, tibia and fibula, but slightly out of articulation. The absence of a calcaneum might therefore be due to taphonomy. Metatarsals I and II were found somewhat separated from mts III–V, with the phalanges php I–1 and III–1 in between. The first ungual was lying above the astragalus, whereas digit II was found in articulation. No other phalanges were found associated, but a small left pedal ungual was recovered mingled with the skull elements, and was therefore used in the mount. It is here described, but attribution to SMA 0011 must be considered preliminary.

Metatarsals. All left metatarsals were recovered complete (Fig. 75). Metatarsals III and IV are the longest elements, and mts I and II the stoutest (Table 7).

Metatarsal I is very robust, and the bone surface collapsed diagenetically in two areas on the dorsal and the lateral surface. The first metatarsal has a D- to drop-shaped proximal surface, which is wider anteriorly than plantarly and has a concave lateral margin. The anterior surface is considerably shorter medially than laterally, resulting in angled proximal and distal surfaces, compared to the long axis of the shaft. The anterior surface bears few nutrient foramina, as is the case in Cetiosauriscus and Suuwassea, but not in camarasaurids (Harris, 2007; Tschopp, Mateus & Benson, 2015; Tschopp et al., 2015). The posterior surface is convex proximally and bears a small foramen centrally on its distal half. The medial surface is slightly convex anteroposteriorly, the lateral one concave for the reception of mts II. Distally, the lateral condyle projects much further than the medial, and develops a distinct posterolateral process, as is typical for diplodocids (McIntosh, 1990a; McIntosh, 1990b). The distal part of the anterolateral edge is marked by a rugose tubercle accompanied by a particular bone surface structure resembling a net of veins (Fig. 76). The distal articular surface bears a distinct intercondylar groove visible in anterior and plantar view.

Metatarsal II has a more squared proximal surface compared to mts I, but with concave medial and lateral margins. The anterior surface is less trapezoidal than in mts I. However, the proximal and distal articular surfaces of mts II are still angled to the long-axis of the shaft. As in mts I, mts II has a strong posterolateral process. The distal portion of the anterolateral edge bears a distinct rugosity, which does not extend onto the anterior surface, unlike in Dyslocosaurus AC 663 or Cetiosauriscus NHMUK R3078 (McIntosh et al., 1992; Tschopp, Mateus & Benson, 2015). Metatarsal II of SMA 0011 has a very distinct anteromedial edge, but a less developed anterolateral one. No intercondylar groove can be seen between the distal condyles in anterior view, but a shallow groove occurs posteriorly.

Metatarsal III is elongate, with a narrow shaft and greatly expanded proximal and distal ends. The proximal and distal articular surfaces stand perpendicular to the shaft axis. The proximal articular surface is subtriangular, with anterior, lateral, and medioplantar margins. It is relatively flat, and does not show distally curving edges as in mts I and II. A strong, narrow projection occurs on the posteromedial corner. A weak, narrow rugosity marks the distal end of the anterolateral edge of the shaft. The proximal portions of the medial and lateral faces are anteroposteriorly concave. The distal articular surface is subtriangular, with the lateral side being much shorter than the medial. It is anteroposteriorly convex and transversely nearly flat.

Metatarsal IV is elongate like mts III, but the proximal expansion reaches further down the shaft. The proximal end is slightly twisted in respect to the long axis. It is subtriangular in outline, with a rather straight lateroplantar margin, unlike the shape of mts IV of the camarasaur SMA 0002 (Tschopp et al., 2015). The surface is flat, as in mts III. The shaft is smooth, and maintains the subtriangular shape of the proximal articular surface. It is concave transversely on its lateroplantar surface, and does not bear any distinct rugosities. The distal end has only incipient condyles, which are hardly recognizable in either anterior or distal view. In distal view, the articular surface is trapezoidal, with a shorter anterior than plantar margin.

Metatarsal V has the typical paddle-shaped outline known from almost all sauropods (Bonnan, 2005). The proximal articulation surface is subtriangular, with the apex pointing anteromedially. From there, a ridge extends distally, separating the proximal portion of the anterior surface from the medial one. The ridge disappears in the distal half. The shaft is smooth, unlike in mts V of the camarasaurid SMA 0002 (Tschopp et al., 2015). The posterior surface is flat transversely, but a lip-like posterior extension of the proximal surface overhangs the face (Fig. 75). The distal surface is a single, convex facet.

Pedal non-ungual phalanges. The left pes of SMA 0011 preserves three proximal non-ungual phalanges and the second non-ungual phalanx of the second digit (Fig. 77). They are relatively short bones with subsequently less well-developed distal condyles, from php I-1 to php III-1.

Pedal phalanx I–1 is slightly wedge-shaped, with a considerably shorter lateral than medial surface. Therefore, the distal condyles face laterodistally, resulting in the typical lateral deflection of the pedal unguals of eusauropods (Bonnan, 2005). The proximal articular surface is subtrapezoid, with two distinct, concave facets for the two distal condyles of mts I. In the medial facet, a deep pit is located close to the midline, and somewhat more anteriorly than plantarly. A similar pit was interpreted as the result of osteochondrosis in the camarasaurid SMA 0002 (Tschopp et al., 2016). The anterior surface is transversely narrower than the posterior surface. It is clearly separated from the medial surface, but grades continuously into the lateral one. The posterior surface is transversely concave, with a smooth transition into the distal articular surface. Laterally, the proximal and distal articular surfaces nearly meet in the plantar half. The distal condyles are in an angle to each other, with the medial one being oriented nearly vertically, whereas the lateral one is oblique, resulting in an articular facet that is anteriorly narrower than plantarly.

Pedal phalanges II-1 and III-1 are similar to each other in general shape. The former is slightly broader than php III-1, which has subequal widths and lengths (Table 7). The medial condyle of both phalanges is transversely compressed, but projects considerably further distally than the lateral one. The proximal articular surface of php II-1 bears a deep pit as in php I-1. Laterally, the proximal facets of both php II-1 and III-1 taper, such that the outline becomes subtriangular in proximal view.

The pedal phalanx II-2 is a proximodistally shortened element, which basically only consists of proximal and distal articular surfaces and a short medial face. The proximal articular surface has two facets for the condyles of php II-1. It is at an angle to the long-axis as indicated by the orientation of the short medial surface. The distal articular surface has a relatively wide medial condyle, and a thin and narrow lateral one. The orientation of the two condyles is subparallel.

Pedal unguals. Three left unguals are preserved and mounted in the left pes of SMA 0011 (Fig. 77). The third ungual was found at some distance to the associated pes, together with skull material, but would fit in size for digit III. As mounted, this amounts to a pedal phalangeal formula of 2-3-2-0-0. This, however, is most probably underestimated, as comparisons with other diplodocid feet indicate (Hatcher, 1901; Gilmore, 1936; Janensch, 1961; Bonnan, 2005). The pedal unguals are sickle-shaped and decrease in length from the first to the third. Ungual III is the most stout element, because the proximal width remains more or less the same from ungual I to III, whereas the length decreases. The pedal unguals I and II are strongly transversely compressed, but this is exaggerated due to taphonomy. The anterior edge is strongly curved and narrow. It is S-shaped in ungual I, because of deformation in the proximal-most part. The medial surfaces are convex, and the lateral sides are anteroposteriorly concave on their anterior portion and convex plantarly. The pedal unguals are wider transversely in their plantar half, especially at the proximal end, where the wider area bears the proximal articular surface. A groove marks the lateral surface, and follows more or less the curvature of the claw. The plantar surface of pedal ungual I is marked by a deep oblique groove, extending from the proximomedial corner to about midlength of the lateroplantar edge. Such a groove has not been described previously, and does not occur in the other two unguals of the same pes. The groove might be caused by taphonomy, because according to the quarry map, a sternal rib was found above it. During diagenesis, this rib could have been pressed onto the claw resulting in such a relatively wide, but elongate groove. The plantar surfaces of pedal unguals I and II bear a weak tubercle, resembling that of Tastavinsaurus sanzi Ars1-3 (Canudo, Royo-Torres & Cuenca-Bescós, 2008: figs. 10A–19B).