Tehuelchesaurus benitezii, RICH ET AL., 1999

Carballido, José L., Rauhut, Oliver W. M., Pol, Diego & Salgado, Leonardo, 2011, Osteology and phylogenetic relationships of Tehuelchesaurus benitezii (Dinosauria, Sauropoda) from the Upper Jurassic of Patagonia, Zoological Journal of the Linnean Society 163 (2), pp. 605-662 : 608-623

TEHUELCHESAURUS BENITEZII RICH ET AL., 1999

Holotype: MPEF-PV 1125, partial postcranial skeleton ( Fig. 2B ), including 10 articulated presacral vertebrae plus an eroded element, 4 sacral vertebrae, parts of the sacricostal yoke, several ribs, right scapulacoracoid, right humerus, left radius and ulna, fragment of right ilium, right pubis and fragment of left pubis, left ischium and shaft of right ischium, both femora, and skin impressions.

Emended diagnosis: The taxon can be diagnosed by the autapomorphic presence of two accessory laminae in the infradiapophyseal fossa of the middle dorsal vertebrae. The accessory lamina 1 runs posterodorsally from the paradiapophyseal lamina and merges dorsally with the accessory lamina 2, which in turn runs posteroventrally from this point to the posterior centrodiapophyseal lamina. Further apomorphies include the presence of an accessory and laterally oriented lamina on the lateral surface of the intrapostzygapophyseal lamina and the strongly anteroposteriorly expanded and very robust distal end of the humerus. Tehuelchesaurus further differs from all other sauropods in the unique combination of the following characters: absence of the prezygoparapophyseal lamina in the middle and posterior dorsal vertebrae (reverse of the plesiomorphic sauropodomorph condition; see discussion below); presence of an accessory posterior centrodiapophyseal lamina, giving the impression of a ventrally bifurcated pcdl (convergently acquired in Titanosauria); a single intrapostzygapophyseal lamina in at least the mid-dorsals (uncertain in other dorsals; convergently present in diplodocids and some basal taxa) that supports the weakly developed hyposphene in middle and posterior dorsal vertebrae; neural spines of dorsal vertebrae longer anteroposteriorly than wide transversely (reversal to the ancestral sauropodomorph condition, also present in Jobaria (Sereno et al., 1999) and Galvesaurus (Barco, 2009) ; absence of lateral expansion in the dorsal end of the neural spine of dorsal vertebrae; absence of postspinal lamina in dorsal neural spines; greatest anteroposterior width of the acromion process of the scapula over the glenoid almost four times the minimum width of the shaft.

Comments: Rich et al. (1999) gave a differential diagnosis for Tehuelchesaurus , only distinguishing it from the allegedly closely related Omeisaurus and from the ‘cetiosaurs’ Barapasaurus and Patagosaurus . However, all the characters used in their diagnosis have a much broader distribution within sauropods, and neither any specific one of these characters nor their combination is unique for Tehuelchesaurus . Upchurch et al. (2004) noted that the characters given by Rich et al. (1999) were insufficient to diagnose the taxon and suggested the unusual, anteroposteriorly broad morphology of the distal humerus as a possible autapomorphy of Tehuelchesaurus . Although the morphology of the distal humerus might partially be due to preservational artefacts, the presence of two small posterior condyles separated by a broad groove and the very deep distal end of this bone might indeed represent further autapomorphies of Tehuelchesaurus .

Type locality and horizon: Estancia Fernandez, 43°10 ′ S, 69°15 ′ W. Lacustrine silt- and sandstones, partially tuffaceous, at the basis of the Cañadón Calcáreo Formation, most probably Oxfordian– Tithonian (see above).

DESCRIPTION

AXIAL SKELETON

The vertebral column is represented by ten dorsal and four anterior sacral vertebrae ( Figs 3–13 ), which were found in articulation. A further vertebral fragment, collected from the surface, was identified as an anterior caudal by Rich et al. (1999: 79), but rather represents a poorly preserved fragment of one of the anteriormost dorsals. Thus, the total number of dorsal vertebrae was at least 11. For reasons of convenience of description, the dorsal vertebrae preserved in articulation are numbered consecutively from 1 to 10, starting with the anteriormost vertebra, as was done also by Rich et al. (1999). The preservational quality of the bone is generally good, but most vertebrae show signs of lateromedial compression and suffered from erosion of their left sides and neural spines. Therefore, the following description is mainly based on the right side of the vertebrae. The terminology of vertebral laminae in the description follows Wilson (1999) and that of vertebral fossae follows Makovicky (1997).

General features

As noted by Rich et al. (1999), all presacral vertebrae are opisthocoelous, although the fully articulated state of the vertebral column makes a precise evaluation of the extent of their anterior convexity impossible in most vertebrae (see below). The centra are slightly higher than wide and this difference becomes less evident in posterior dorsals, although the transverse compression could have distinctly affected these proportions in different regions of the dorsal column (see below). The central length slightly increases from D1 to D3 and then gradually decreases, whereas the central height increases gradually and constantly from D1 to D10. The vertebral centra have a transversally convex ventral surface, unlike the ventrally flattened centra of some basal sauropods (e.g. Jain et al., 1979). Well-developed pleurocoels are present throughout the dorsal series and in the first sacral vertebra. These pleurocoels are subovoid in shape, having rounded anterior and posterior margins. Contrary to Rich et al. (1999), these true pleurocoels are pneumatic cavities that invade the centra and form large camerae (sensu Wedel, Cifelli & Sanders, 2000). The camerae extend anteriorly, posteriorly, ventrally, and dorsally from the pleurocoel and seem to be deeper in the anterior half (see below).

The neural arches have a relatively constant height, but vary from being subequal to lower (measured from the floor of the neural canal to the dorsal border of the proximal basis of the transverse process) than the vertebral centra toward the posterior dorsals (mainly because of the dorsoventral increase in the height of the centra). The lateral sides of the neural arches bear a large infradiapophyseal fossa. In most vertebrae this fossa bears a small, deep depression with well-defined margins that is located below the diapophysis and anterior to the posterior centrodiapophyseal lamina (PCDL), herein referred to as the posterodorsal depression. This depression has a similar topographical position as the large and deep lateral depressions present in some basal eusauropods (e.g. Barapasaurus , Cetiosaurus , Patagosaurus ; Jain et al., 1979; Bonaparte, 1986b, 1999; Upchurch, 1998; Upchurch & Martin, 2002, 2003). These depressions are developed as deep, but superficial depressions in the anterior dorsals in Patagosaurus and Barapasaurus (Jain et al., 1979; Bonaparte, 1986b, 1999), and seemingly in all dorsals in Cetiosaurus (Upchurch & Martin, 2002) , whereas they are developed as large lateral foramina that lead into large supraneural camerae in the posterior dorsals in the former two genera. Large, paired supraneural camerae are also present in Tehuelchesaurus , as in many other eusauropods, including neosauropods (e.g. Diplodocus , Camarasaurus ; Bonaparte, 1999). These cavities are positioned medial to the deep posterodorsal depressions, although the preservation of the specimen does not allow determination of whether the supraneural camerae communicate with the exterior through a supraneural foramen, as in Barapasaurus (Jain et al., 1979) and Patagosaurus (Bonaparte, 1986a) . The neural canal has been plastically deformed in some vertebrae, but some are seemingly undistorted and have a semicircular shape (being slightly wider than high), rather than being high and slit-like as in Barapasaurus (Jain et al., 1979) and Amygdalodon (Rauhut, 2003a) . A vertical ridge extends dorsally above the neural canal on the anterior side of anterior dorsal vertebrae [i.e. the ventral end of the intraprezygapophyseal lamina (TPRL); see below] and is bounded by large depressions that extend laterally onto the anterior surface of the neural arches.

The zygapophyses are relatively large, broad, and oval-shaped in the anterior dorsals, but they become gradually smaller in more posterior vertebrae. The zygapophyses project anteriorly and posteriorly, slightly exceeding the level of the anterior or posterior rim of the vertebral centra (excluding its convex anterior articular surface). The articular facets of the pre- and postzygapophyses are obliquely inclined at an angle of approximately 30° from the horizontal, as in most neosauropods. In most vertebrae, a lateral expansion of the ventral end of the prezygapophysis is present and most probably represents the lateral face of the hypantrum. Although these expansions are well developed in most vertebrae, the hyposphene– hypantrum system seems to be weakly developed, as can be observed in D9. A single lamina that joins the hyposphene with the neural canal is present and interpreted as the intrapostzygapophyseal lamina (TPOL), following Wilson (1999). The development of this lamina increases markedly in posterior vertebrae (see below).

The parapophyses are placed on the neurocentral junction in D1 and D2, whereas in more posterior elements they reach a position slightly above the half-height of the neural arch in D4–D8, but are more ventrally placed again in the last dorsals. The transverse processes and diapophyses are damaged in most vertebrae, but where they are preserved, they are short and massive and directed laterally, rather than dorsolaterally. The articulations for the capitula of the dorsal ribs on the parapophyses are concave, whereas those for the tubercula on the transverse processes are convex. The lateral surfaces of the neural arches bear well-developed lamination, which changes considerably along the preserved region of the dorsal series (see below).

The lamina that is most consistently developed in all dorsal vertebrae is the PCDL, although its presence cannot be determined in D10. This lamina is robust and slightly inclined anterodorsally in the anteriormost vertebrae, but almost vertical in more posterior elements and connects the ventral side of the diapophysis with the dorsal rim of the vertebral centrum. The Paradiapophyseal lamina (PPDL) is also well developed in all vertebrae in which the parapophysis is located above the neurocentral suture (D3–D10). This lamina is directed anterodorsally, so that the parapophysis overhangs the centrum anteriorly. A low, stout, laterally rounded centropostzygapophyseal lamina (CPOL) is present in the infrapostzygapophyseal fossa from D2 to the last presacral element, varying only in the dorsoventral placement of its anteroventral point of origin (see below). Well-developed, more or less horizontal prezygodiapophyseal and postzygodiapophyseal laminae (PRDL and PODL) are present in all dorsals, the former being much shorter and robust than the latter. Two autapomorphic laminae of Tehuelchesaurus are present between the PCDL and the paradiapophyseal lamina (PPDL). The development of these laminae and their exact positions (as well as that of other laminae) varies along the dorsal series and will be described in detail for the respective elements below. No complete neural spine is preserved, but parts of this structure can be observed in D1–D8, with D6 apparently presenting an almost complete spine, which is not bifurcated. As is usual in eusauropods, neural spine lamination is well developed, but contrary to most neosauropods (Upchurch et al., 2004) prespinal and postspinal laminae are absent. Spinoprezygapophyseal and spinopostzygapophyseal laminae (SPRL and SPOL) are well developed, but, in contrast to many sauropods (Wilson, 2002), the latter is not subdivided into a lateral and a posterior spinopostzygapophyseal lamina.

Dorsal 1 ( Fig. 3 )

The ventral margin of the centrum of D1 is approximately 1.25 times longer than the dorsal margin ( Table 1), indicating that the dorsal vertebral column flexed dorsally towards the dorsal–cervical junction. D1 is the only vertebra in which the complete anterior articular surface is exposed. This surface is strongly convex, showing the ophistocoelous condition of anterior dorsals of Tehuelchesaurus . The length of the anterior articular condyle is approximately 50% the centrum length.

Deep pleurocoels invade the centrum and expand within it, forming several large pneumatic camerae. Left and right camerae are separated by welldeveloped and thick bony septa. They are exposed on the dorsal surface of the anterior articular condyle of the centrum through a broken area, providing a longitudinal section of these camerae at different dorsoventral heights. The left camera is exposed at midheight of the centrum (at the level of the pleurocoel). The camera extends anteriorly from the pleurocoel, is lateromedially narrow and elongated, and extends throughout most of the articular condyle. The anterior region of the right camera is exposed at a more dorsal level, close to the dorsal edge of the centrum. At this level the anterior camera is anteroposteriorly shorter (extending only within the articular condyle) and wider transversely than in the more ventral level exposed on the left side. It shows fragments of a thin parasagittal septum that seems to (at least partially) divide the camera. The left camera is also exposed posterior to the pleurocoel, where it is broad and lacks a well-defined contour. The dorsal extension of this part of the camera seems to invade the neural arch (as observed on the medial surface of pedicels of the right neural arch). In contrast to the derived condition of Brachiosaurus , as well as more derived titanosauriforms (Wedel et al., 2000; Wedel, 2003a, b), this vertebra does not have a camellate structure *Length at about mid-height of centrum, excluding the anterior convexity.

†Dorsal length.

‡Ventral length.

d, damaged or deformed. Centrum width has not been evaluated as all vertebrae are more or less compressed transversely.

either in the condyles or in the zygapophyses. Therefore it resembles the condition of several nontitanosauriform neosauropods (e.g. Camarasaurus , Galvesaurus ; Wedel, 2003a; Barco, 2009). The pleurocoel is enclosed in a large depression delimited dorsally by the posterior centroparapophyseal lamina (PCPL) and the parapophysis ( Fig. 3 ). This depression has an acute posterior border produced by the almost horizontal orientation of the PCPL, although the posterior margin of the pleurocoel itself is rounded. This character also contrasts with the derived acute posterior edge of the pleurocoels present in Titanosauriformes (e.g. Brachiosaurus , Chubutisaurus , Saltasaurus ).

The parapophysis is entirely placed on the centrum, and a stout centroprezygapophyseal lamina (CPRL) originates from the dorsal rim of the centrum above it, being directed anterodorsally along the anterior margin of the neural arch ( Fig. 3 ). The four diapophyseal laminae that characterize the presacral vertebrae of most eusauropods (Wilson, 1999) are present and well developed in this vertebra. A robust PCDL is anterodorsally oriented (nearly 45°) and is almost parallel to the CPRL ( Fig. 3 ). Although the contact of the PCDL with the diapophysis is broken, it is evident that this lamina forms the vertical axis of the T-shaped diapophysis. The anterior centrodiapophyseal lamina (ACDL) runs posterodorsally from the lower third of the CPRL to approximately the half height of the PCDL ( Fig. 3 ), being oriented at an angle of 45° with respect to the longitudinal axis. The ACDL of this vertebra is relatively reduced, as the anteroposterior length of its neural arch is much smaller than in subsequent vertebrae. However, the ACDL is much more robust than its homologous structure, the PPDL, in the other dorsals. The two other laminae are the PRDL and the PODL, the former being more robust and shorter than the latter and forming the horizontal axis of the T-shaped diapophyseal process.

Two large triangular fossae are present on the lateral surface of the neural arch, located one above the other and separated by the ACDL ( Fig. 3 ). The dorsal fossa is bounded by the CPRL anteriorly, the ACDL posteroventrally, the dorsal portion of the PCDL posteriorly, and the PRDL dorsally. The apex of this triangular fossa is directed ventrally, whereas it is broad dorsally. The extension of this fossa is much greater than in the two subsequent vertebrae, expanding over approximately half of the height of the neural arch. Given the position and development of this fossa in the subsequent vertebrae (D4–D5; see below), we refer to this fossa as the infraprezygapophyseal fossa (INPF; Fig. 3 ). This fossa bears a small but distinct posterodorsal depression in its posterior half, just above the point of contact between the ACDL and the PCDL. As noted above, this depression might represent a lateral exit of an extensive supraneural camerae, but the presence of such camerae in this vertebra cannot be established with certainty (they are absent in the anterior dorsals in some advanced, non-neosauropodan eusauropods; see Upchurch & Martin, 2002). The lower triangular fossa (infradiapophyseal fossa; INDF) is bordered by the PCDL posteriorly, the ACDL anteriorly, the parapophysis anteroventrally, and the PCPL posteroventrally. This fossa is small but deep and is much smaller than those of subsequent vertebrae, given the relatively lower position of the ACDL in dorsal 1.

The anterior surface of the neural arch becomes progressively wider dorsally, leading to the prezygapophysis ( Fig. 3 ), which is completely preserved on the right side of the vertebra. The postzygapophyses are placed posteriorly, close to the level of the diapophysis in the next vertebra (D2). Due to the articulation with the subsequent vertebra, their articular surfaces cannot be observed, but the posterior and lateral surfaces of the right postzygapophysis and the posterior base of the neural spine, which was positioned posteriorly above the postzygapophysis, are well preserved and bear the basal part of the SPOL and spinodiapophyseal laminae (SPDL). Only a minor portion of the SPOL is preserved in D1 and resembles the condition in more posterior vertebrae, in which this lamina is more completely preserved (see below). The SPDL of this vertebra is oriented at an angle of approximately 25° from the longitudinal axis, thus being more horizontal than the SPDL in subsequent elements. A small triangular fossa is present on the lateral surface of the postzygapophyses, within the lateral recess formed by the dorsal junction of these two laminae. This fossa becomes progressively deeper anteromedially and is undivided, in contrast to the condition of subsequent vertebrae (see below). A similar depression is also present in other vertebrae (D2–D3) and may represent an autapomorphic character of T. benitezii ( Fig. 4 ).

Dorsal 2 ( Fig. 4 )

This vertebra and D3 differ considerably from each other and from dorsal 1, thus representing a transitional series between the morphology of the anterior dorsals and the mid to posterior dorsal vertebrae. The centrum length of D2 is similar to the ventral length of D1, but unlike this vertebra there is no difference in the centrum length at its dorsal and ventral edges. A deep pleurocoel is enclosed in a large and posteriorly acute depression that is approximately 1.3 times larger than that of D1. This depression is delimited dorsally by a stout PCPL ( Fig. 4 ). This lamina is not horizontal as in D1, but is slightly inclined anterodorsally, as the parapophysis is placed more dorsally. The parapophysis is located at the level of the neurocentral suture, having almost 80% of its dorsoventral extension placed on the neural arch. The articular surface of the parapophysis is large and comma-shaped, having an acute ventral edge and a rounded dorsal edge ( Fig. 4 ). This contrasts with the parapophysis of the preceding vertebra (which is significantly smaller) and with that of the subsequent vertebrae (which are oval in shape).

The neural arch of D2 is anteroposteriorly longer than that of D1 but shows the same basic lamination pattern, with the addition of accessory laminae and an associated fossa. The prezygoparapophyseal lamina (PRPL) of D2 is well developed and forms the anterior margin of the lateral surface of the neural arch ( Fig. 4 ), but it is less robust, shorter, and more anteriorly directed than that of D1. A similar difference is present in the PCDL of D1 and D2, in which it has a sharp ventral end and dorsally reaches the diapophysis, forming a prominent vertical base of this process. An additional lamina extends ventrally from the dorsal end of the PCDL of D2 and runs parallel to it, along its anterior margin. A small, trough-shaped concavity is present between the two laminae along their dorsal half ( Fig. 4 ). The ventral end of this lamina seems to be incomplete and may have been displaced from its natural position, given that this region of the neural arch seems to be broken and slightly deformed. The lamina that corresponds to the ACDL in the first dorsal is here better understood as PPDL, as the parapophysis has almost entirely moved onto the neural arch. It is significantly different from the corresponding lamina of the preceding vertebra. First, it is more dorsally located, running from the midpoint of the CPRL to the accessory lamina described above (close to the dorsal end of the PCDL). Second, the orientation of the PPDL is more horizontal than the ACDL in D1. Third, the anteroventral end of the PPDL is less robust than that of the homologous lamina in D1, forming a thin and sharp lamina. Fourth, the posterodorsal end of the PPDL is robust but not as developed as in D1. Rich et al. (1999: fig. 7) labelled a more ventrally placed lamina, which extends from the posterodorsal rim of the parapophysis as ‘ventral part of broken infraprezygapophyseal (ACDL) lamina’ in this vertebra. However, repreparation of the vertebra revealed that this lamina is indeed not broken, but represents a short additional lamina (see below). The PODL and the PRDL of D2 are well developed and form the horizontal laminae of the diapophyseal process. The PRDL is more developed in this vertebra than in subsequent elements.

The lateral surface of the neural arch of D2 bears a series of depressions and fossae that differ from those of other elements. The most extensive of these is a large trapezoidal fossa enclosed by the PCDL (posteriorly), PPDL (anterodorsally), PRPL (anteriorly), PCPL (posteroventrally), and the parapophysis (anteroventrally). This depression is much larger than the lower fossa of D1 and is also interpreted as the infradiapophyseal fossa ( Fig. 4 ). Within this fossa, three distinct depressions are present. The largest of them extends over the posterodorsal corner and excavates the base of the anterior surface of the PCDL (and its accessory lamina). This depression is dorsally bounded by the PPDL and is in identical topographical position as the posterodorsal depression developed in more posterior dorsals (see below). A distinct, kidney-shaped depression is present in the posteroventral end of the infradiapophyseal fossa. This depression is approximately 5 cm long and relatively shallow, but is only present in this vertebra. The third depression is developed at the anteroventral corner of the infradiapophyseal fossa. This depression excavates the posterior surface of the base of the parapophysis and is dorsally bounded by an accessory lamina. This lamina [referred to as the ACDL by Rich et al. (1999)] originates in the posterodorsal corner of the parapophysis and becomes rapidly lower posterodorsally, disappearing at about mid-length between the parapophysis and the PCDL within the infradiapophyseal fossa ( Fig. 4 ). The infraprezygapophyseal fossa of D2 is much more reduced than in D1. This fossa is deep, subcircular ( Fig. 4 ), and is enclosed between the PPDL and the PRPL along its ventral half. The dorsal border of this rounded fossa forms a sharp edge that is oriented horizontally, 5 cm below the PRPL. The lateral surface of the neural arch also bears a shallow and elongated concavity that runs enclosed between the PPDL and the PRDL.

The dorsal surface of the right prezygapophysis cannot be observed because it is in natural articulation with the postzygapophysis of D1, but the anterior surface of the prezygapophyseal process is well exposed in this vertebra. The lateral half of this process is convex (as in D1) but the medial half is markedly concave. This concavity represents the circumneural cavity (sensu Bonaparte, 1999) and is located dorsal to the neural canal. The dorsal limit of this concavity is formed by the medial region of the prezygapophysis, which seems to lack a hypantrum. The left and right circumneural cavities are separated from each other by a sharp vertical strut of the TPRL. The dorsal (paired) part of the TPRL is not preserved in this vertebra. The postzygapophyses of D2 are badly damaged, except for their medial sides and articular surfaces (still in articulation). The medial surfaces of both postzygapophyses are almost in contact with each other. The postzygapophyseal process is supported by a well-developed CPOL, not observed in the previous vertebra (D1). This lamina is thin and delicate in comparison with the CPOL of subsequent vertebrae and extends from the midheight of the PCDL posterodorsally to the anteroventral rim of the postzygapophysis ( Fig. 4 ). The postzygapophysis is connected to the dorsal end of the PCDL by an accessory lamina that runs horizontally and flares anteriorly. This accessory lamina is located dorsal to the CPOL, and a deep depression is enclosed between them.

Dorsal 3 ( Fig. 5 )

This vertebra was cut transversally at the level of the PCDL ( Fig. 5 ) at the time this specimen was collected, although this lamina is only broken at its ventral end. The pleurocoel present in this vertebra has a similar anteroposterior length as that present in D2, but they differ in their general shape. The pleurocoel of this vertebra has the major dorsoventral height at its posterior end and not at the anterior end as in D2. Although a pleurocoel fossa is present, its dorsal edge is formed by a pronounced edge, which obviously corresponds to the PCPL, but does not take the form of a true lamina; this lamina is thus absent in this and all subsequent vertebrae. The fossa is developed as a small posterior depression that is distinctly and abruptly inset from the lateral surface of the centrum. The vertebral centrum is cut transversally at the level of the anterior margin of the pleurocoel posterior fossa. Several isolated spaces filled with sediments are observed in this cross-section, which are interpreted as pneumatic cavities [camerae in the sense of Wedel (2003a, b); Wedel et al., 2000]. These small spaces are present both in the centrum and in the neural arch. The latter obviously represent small supraneural camerae that are separated by a stout midline septum. The small size of these camerae, also in comparison with their development in more posterior dorsals (see below), is consistent with the observation that such camerae are usually first developed in more posterior dorsals in some non-neosauropodan basal eusauropods, such as Patagosaurus and Barapasaurus (see Upchurch & Martin, 2002). The pneumatic cavities in the centrum of D3 are markedly smaller than those observed in D1 (and D8), but this difference is interpreted as a product of the posterior location of the breakage in the former, rather than a decrease in pneumaticity. This is in accordance with the observation that the pneumatic cavities become smaller posteriorly in D1, as described above.

The lateral surface of the neural arch has a different lamination pattern than the preceding vertebrae, and is unique in having an accessory lamina absent in all other preserved vertebrae (see below). Additionally, two other incipiently developed laminae are present in D3, which are topographically homologous with two well-developed autapomorphic laminae present in subsequent vertebrae. Therefore, these laminae are also interpreted as autapomorphic.

The parapophysis is entirely placed on the neural arch, so that the ACDL is divided in two laminae (Wilson, 1999), the ACPL that connects the parapopshysis with the anterior centrum edge, and the PPDL that connects the parapophysis to the diapophysis. The ACPL runs posteroventrally from the parapophysis to the centrum and forms an angle of approximately 35° with the longitudinal axis. The articular surface of the parapophysis is oval, with its major axis is oriented anteroventrally–posterodorsally. In this vertebra the PPDL connects the parapophysis with the ventralmost part of the diapophysis, where the PCDL forms the ventral pillar of this process ( Fig. 5 ). In contrast, the PPDL of subsequent vertebrae extends from the posterodorsal rim of the parapophysis to the anterior part of the diapophysis at the contact with the PRDL (see below). Although the right PCDL is partially broken, especially at its ventral end, it seems that this lamina was not very robust. The ventral end of the PCDL can be seen on the left side of this vertebra. This lamina has an expanded and bifurcated ventral end, separated by a triangular depression. The anterior branch is rounded whereas the posterior branch is thinner and sharper ( Fig. 5 ), as in titanosaurs (Salgado et al., 1997). These two branches fuse gradually dorsally, forming a single PCDL that connects to the ventral side of the diapophysis. According to Salgado, Apesteguía & Heredia (2005), the posterior branch corresponds to the original PCDL, whereas the anterior branch represents an accessory posterior centrodiapophyseal lamina (APCDL). As in D2, the PRDL is present in this vertebra but this lamina is less developed, being lower and broader than that of D2. Consequently, the T-shaped diapophysis is less evident in lateral view in D3. A triangular infradiapophyseal fossa is present in this vertebra ( Fig. 5 ). This fossa is broad and becomes progressively deeper toward its dorsal edge. Its ventral limit is not well defined, but the fossa becomes gradually shallower ventrally, in contrast to the condition of the preceding vertebra, in which this fossa is clearly delimited by the PCPL. The infradiapophyseal fossa is enclosed by the PPDL anterodorsally and PCDL posteriorly ( Fig. 5 ). A deeper and triangular posterodorsal depression is present within the infradiapophyseal fossa. This depression is defined by the PCDL posteriorly, the PPDL dorsally and a posteromedially oriented bony crest anteroventrally ( Fig. 5 ). Two autapomorphic laminae are present in the anterodorsal margin of the infradiapophyseal fossa and below the PPDL. These laminae are incipient ridges and their topographical position is the same as that of the well-developed accessory laminae present in the next vertebra (see below). Consequently they are interpreted as weakly developed autapomorphic laminae, homologous to those present in following vertebrae. An accessory lamina, which is only present in this vertebra, runs parallel to the PPDL along its posterior course and is located dorsal to it and ventral to the PRDL ( Fig. 5 ). This lamina originates within the posterior part of the infraprezygapophyseal fossa and meets the PCDL posteriorly, dorsal to the connection of this lamina with the PPDL. Anteriorly, the dorsal rim of the infraprezygapophyseal fossa runs in the anterior extension of this lamina. The infraprezygapophyseal fossa is much smaller than in D2 and D1 and is dorsally bordered by the accessory lamina only present in this vertebra and ventrally by the PPDL.

The neural spine is largely incomplete but seems to be similar to the neural spine of D6 (see below). Well-developed SPOL and SPDL are present on the postzygapophysis and the posterior end of the lateral side of the neural spine, respectively. These laminae enclose a depression that is larger than that of D1 and is subdivided in two distinct con- cavities separated by a bony strut. The ventralmost depression is conical and deeper than the dorsal fossa. The SPRL are also present but nothing can be said about their development in detail due to poor preservation.

Dorsal 4 ( Fig. 6 )

The pleurocoel of this vertebra does not differ in shape or length from that of D3 and also bears a posterior fossa. This fossa is very similar both in shape and in length to that of D3, but differs from the fossa of subsequent vertebrae in the presence of an abrupt, step-like posterior border. From D4 to D9 (the last vertebra in which a small pleurocoel fossa is present) the pleurocoel fossa becomes progressively smaller. The parapophysis is more dorsally located on the neural arch than in D3 and is connected to the centrum by the ACPL, which is set at an angle of approximately 35° to the vertical, resembling the condition of subsequent dorsal vertebrae. The PPDL connects the parapophysis with the anterior margin of the diapophysis ( Fig. 6 ), rather than with its ventral margin, as is the case in D3. The articular surface of the parapophysis is smaller than in D3 and is oval in outline ( Fig. 6 ).

The lamination of the neural arch of this vertebra differs from that of D3 and remains more or less constant from this vertebra to D9, with only minor changes in the shape or orientation of some laminae. This is especially true for the two autapomorphic laminae (named here as AUT 1 and AUT 2), which are better developed than in more anterior vertebrae. Two depressions are present just below these laminae. The larger depression is situated below AUT 1 and is rectangular in shape.

As in the anterior vertebra, the robust PCDL has an expanded and bifurcated ventral end, with a broad posterior branch and smaller anterior branch, enclosing an oval depression between them ( Fig. 6 ). The PRDL is a very low strut and merges dorsolaterally with the PPDL.

A stout CPOL is present in this vertebra and in most of the following vertebrae. A lateromedially flattened and well-developed TPOL is present and clearly visible in lateral view, and bears an accessory lamina on its lateral surface that projects laterally and runs steeply posterodorsally, approximately parallel to the posterior margin of the TPOL. This accessory lamina, here defined as accessory TPOL (aTPOL), can be seen in all vertebrae where the TPOL is observed, and is interpreted as an autapomorphic lamina only present in Tehuelchesaurus . This lamina converges dorsally towards the CPOL ( Fig. 6 ).

A trapezoidal infradiapophyseal fossa encloses two deeper depressions, as they are also present in D3. One of these depressions lies posteroventral to the parapophysis and is oval in outline and larger than that of the latter vertebra. This depression is here termed infraparapophyseal depression. The depression located below the AUT 1 is oval rather than rectangular, as in D3. A bony strut separates this depression from the infraparapophyseal depression, and the AUT 1 runs posteriorly from this bony strut ( Fig. 6 ). As in D3 this lamina joins with the laterally oriented AUT 2, which is smaller and less developed than the AUT 1 ( Fig. 6 ). Due to the absence of homologous anchor points either at the beginning or at the end of these accessory laminae, no formal names in accordance with the nomenclature proposed by Wilson (1999) are proposed here, and they are simply referred to as autapomorphic lamina 1 and 2 (AUT 1 and AUT 2; Fig. 6 ). The more anterior lamina (AUT 1) runs posterodorsally from a small infraparapophyseal depression. The posterodorsal end of the AUT 1 joins with the lateral end of the AUT 2.

The anterior and posterior margins of the neural spine are formed by the paired SPOL and SPRL laminae, whereas the lateral surface of the spine bears a stout SPDL. The SPDL and SPOL run dorsally and are expanded laterally, whereas the SPRL projects anteriorly and it is not laterally expanded. A triangular depression is defined by the SPDL and SPOL. The rounded fossa present in D3 is also present in D4, although it is shallower than in the preceding vertebra.

Dorsal 5 ( Fig. 7 )

This vertebra is transversely broken close to the posterior margin of the pleurocoel ( Fig. 7 ), slightly anterior to the level where D3 has been cut. The pleurocoel is slightly shorter than that of the more anterior vertebrae and the posterior fossa behind the pleurocoel is small and shallows gradually posteriorly. As in D3 the pneumatic cavities of the centrum and neural arch can be seen in the exposed cross-section. The pleurocoel opens into camerae through its ventral and dorsal margins, especially in its anterior half where these fossae seem to be larger. As in the other vertebrae, the paired pneumatic cavities present in the centrum are true camerae (sensu Wedel et al., 2000; Wedel, 2003a, b). The paired pneumatic cavities of the neural arch are supraneural camerae similar to those of some basal eusauropods and some basal neosauropods (e.g. Barapasaurus , Patagosaurus , Camarasaurus ; Bonaparte, 1986b, 1999). The camerae of the centrum and supraneural camerae extend ventrally and dorsally from the pleurocoel, respectively. These pneumatic cavities are short and at least at this more posterior level, the centrum and the neural arch are only moderately pneumatized. Although the right infradiapophyseal fossa communicates with the right supraneural camera through a small opening, it cannot be determined whether this opening is natural. The opening bears numerous small fragments of thin, laminar bone, which may represent remains of thin bony septa that enclosed the supraneural camera laterally, as in Camarasaurus or Diplodocus (Upchurch et al., 2004) . As in D3 the reduced pneumaticity (compared with D1 or D9) is interpreted as a result of the posterior section at which the vertebra was cut.

This vertebra was damaged laterally and the diapophysis and PODL are incomplete, but the general development of the laminae seems to be similar to that of D4. The expanded ventral margin of the PCDL is less developed than in D4, and especially its anterior branch is less robust. The length from the beginning of the bifurcation to the end of the anterior branch is half as that of the anterior vertebra. Although the PRDL is present as a weakly developed lamina in D4 and D6, this lamina seems to be completely absent in D5. Accessory autapomorphic laminae are better developed in this vertebra (and D6) than in any other element. In contrast to the situation observed in D3 and D4 AUT 1 and AUT 2 are equally developed and without a difference in length between them ( Fig. 7 ). AUT 2 is not as laterally oriented as in D4 but is dorsoanteriorly oriented and meets the AUT 1 in an approximately right angle. AUT 1 extends posterodorsally and follows the course of the PPDL. Two deeper depressions are placed below these accessory laminae. As in D3 there is an almost rectangular depression below the AUT 1, and a depression with a small opening that seems to communicate with the right supraneural camera is situated just below the AUT 2. As in D4, a robust CPOL is present, but with a lateral position, in contrast to the more medial position that this lamina has in D4. The TPOL and its accessory lateral lamina are also present in this vertebra.

As in D4, the neural spine is formed by the paired SPOL, SPRL and SPDL. The principal difference between the neural spines of the former vertebra and D5 is that the SPOL of the latter is less laterally expanded and the SPDL is less developed.

Dorsal 6 ( Fig. 8 )

Due to erosion of the posterior end of D5, the anterior convex articular surface of this vertebra and the strongly concave posterior articular surface of D5 can be observed. The pleurocoel is very similar in size and shape to that present in the previous vertebra, but the fossa situated posterior to this structure is somewhat smaller.

The general features of the lamination of the neural arch do not differ notably from those of the preceding vertebra, except for the orientation and development of the two autapomorphic laminae (AUT 1 and AUT 2; see below). Although the PCDL does not have a marked accessory lamina (APCDL), it notably widens ventrally ( Fig. 8 ), as also occurs in some titanosaurs. As in middle and posterior dorsal vertebrae in which this lamina is present, the CPOL of this vertebra is well developed. The TPOL is present in at least all vertebrae from D4 to D10, but in D6 this lamina is absent, and there are no signs of it, although there is a slightly convex ridge above the neural canal that does not seem to be broken. The PRDL, absent in D5, is present again in this vertebra showing a similar development and orientation to that observed and described in D3.

The infradiapophyseal fossa is slightly smaller than in D5. The strongly posteroventrally oriented ACPL forms the ventral margin of this fossa. This fossa is further delimited by the PCDL and the strongly developed AUT 2 (posteriorly), PPDL (anterodorsally), and ACPL (anteroventrally). The two autapomorphic laminae (AUT 1 and AUT 2) show a different development and orientation than in D5. In D6 both laminae project laterally from the depth of the infradiapophyseal fossa, whereas they project more ventrally in D5. In contrast to D3 and D4, AUT 1 of this vertebra is less developed than AUT 2. The former extends from the dorsal end of the infraparapophyseal fossa to the anterodorsal end of AUT 2. The fossa below AUT 1 is rounded and larger than in anterior vertebrae. AUT 2 extends anterodorsally from the dorsoventral midpoint of the PCDL to its dorsal contact with the AUT 1 ( Fig. 8 ). A small rounded depression is formed between AUT 1 and the PPDL. AUT 2 separates the infradiapophyseal fossa from a small triangular depression also delimited by the diapophysis (dorsally) and the PCDL (posteriorly).

The neural spine of this vertebra is the most complete, with its preserved height being 16 cm. The SPOL joins with the SPDL at its half-height and the resulting composite lamina extends dorsally where it meets the arched SPRL. As a result of this union there is an elongate depression on the lateral surface of the neural spine, as in many basal eusauropods, with the exception of the most basal forms. As this union is usually placed between half and two-thirds of the height of the neural spine, the mid-dorsal neural spines in Tehuelchesaurus would have been 30–35 cm high at the most, and thus approximately as high as the neural arch (discounting the spine), or lower. The spine is very slightly inclined posteriorly in D6, and seemingly also in D5 and D7. Rich et al. (1999: 76) stated that it cannot be said whether the neural spines of Tehuelchesaurus were bifurcated; however, at least this spine does not show any signs of bifurcation, and the bases of the more anterior spines are also more consistent with un-bifurcated structures. This spine is unusual for a sauropod in that it is longer anteroposteriorly than wide transversely, probably at least partially as a result of the only weakly developed SPDL.

Dorsal 7 ( Fig. 9 )

The pleurocoel of this vertebra is slightly shorter than that present in D8 and is not oval as in the anterior vertebra, but more rounded. The posterior pleurocoel fossa is very weakly developed and this vertebra is the last element in which this fossa is clearly present ( Fig. 9 ).

Unlike the situation observed in D6 and similar to that observed from D3 to D5 this vertebra has a ventrally bifurcated PCDL. The anterior branch of the ventrally bifurcated PCDL is continuous in direction with the CPOL ( Fig. 9 ). As in the posterior dorsals of this taxon, the anterior branch of the PCDL might be an anterior continuation of the CPOL (and therefore not homologous to the APCDL of titanosaurs). However, the different orientation of the anterior branch of the PCDL and the CPOL in more anterior vertebrae (see Figs 6 , 7 ) argues against this interpretation. Thus, for the moment, we use the term APCDL for this structure.

A deep and triangular infradiapophyseal fossa is posteroventrally defined by the APCDL. Within this fossa, a smaller, deeper, triangular depression is present. A broad, rounded ridge that runs anterodorsally from its contact with the anterior projection of the PCDL delimits this deeper zone anteroventrally whereas the latter forms its posteroventral margin ( Fig. 9 ). Two other anterior and posterior indentations are located within this triangular depression and might represent foramina that communicate with the supraneural camerae, but this area cannot be prepared and therefore the presence of such a communication cannot be confirmed. The anterior of these possible foraminae is located just above the anterodorsally oriented ridge whereas the posterior is at the level of the PCDL (above the anterior projection of the APCDL). Accessory autapomorphic laminae are also present but they are less developed than in D6 or D5. The AUT 1 runs posterodorsally from the midpoint of the PPDL and, together with the latter lamina, encloses a triangular fossa ( Fig. 9 ). The AUT 2 is weakly developed and is partially broken. This lamina runs anterodorsally and meets the AUT 1 at the posterodorsal end of this lamina. A well-developed TPOL is present and has an accessory laterally expanded lamina (aTPOL), as in D4 and D5 ( Fig. 9 ). This accessory lamina is well developed and runs dorsally, disappearing at the level of the CPOL.

This vertebra presents a well-developed SPDL, which is vertical and does not contact the SPOL, as is the case in the preceding vertebra. The SPOL is well developed, but is not laterally expanded and is posteriorly oriented. A shallow ventral depression is present between this lamina and the anteroventrally expanded SPDL, and another rounded depression is present between the SPRL and the SPDL.

Dorsal 8 ( Figs 10 , 11 )

As mentioned above, there is no fossa posterior to the pleurocoel in this vertebra, although a shallow concavity is present in this position. The length and shape of the pleurocoel are similar to that described for D7. The ACPL is a very robust lamina and its orientation does not differ from that in mid-dorsals. The PCDL and the CPOL merge at the base of the neural arch, forming a robust and dorsally oriented lamina. A small anterior branch runs from this lamina to the ventral end of the ACPL, and is interpreted as the APCDL. The prezygapophysis of the next vertebra (D9) was disarticulated and thus the contact between CPOL and the right postzygapophysis as well as the hyposphene ridge can be seen in D8 ( Fig. 10 ). The TPOL is present but its accessory lamina seems to be absent in D8, although it is present in all other vertebrae in which the TPOL is present. In most sauropodomorphs the TPOL is absent in vertebrae that have a hyposphene, with the No significant difference is observed between the neural spine of this vertebra and that of D6. The paired SPOL are posterolaterally projected, forming an angle of nearly 45° with respect to the sagittal plane, and the paired SPRL are craniolaterally oriented at a similar angle. The paired SPDL are laterally oriented, and probably suffered plastic deformation, given the asymmetric position and orientation of these laminae ( Fig. 11 ).

exception of diplodocids, in which this lamina supports the hyposphene from below (Wilson, 1999; Upchurch et al., 2004; Apesteguía, 200s5). Tehuelchesaurus has a well-developed TPOL in almost all its mid to posterior dorsals and at least this vertebra shows a weakly developed hyposphene.

The infradiapophyseal fossa is ventrally defined by a low ridge, similar to that present in D7, in which this ridge forms the anteroventral margin of a deeper area. A rounded depression is located within the infradiapophyseal fossa, close to its posterior edge. This depression is more extensive than that in D7 and might be associated with a supraneural foramen, as in previous vertebrae, but preparation of this fossa is incomplete. The depression located at the anterior region of the infradiapophyseal fossa of D7 is absent in D8. This vertebra is the last element in which the AUT 1 and AUT 2 are present. The AUT 1 is very weakly developed as a low projection extending from the dorsal end of the PPDL ( Fig. 10 ). The AUT 2 runs anterodorsally and meets the AUT 1 just below the diapophysis ( Fig. 10 ).

The neural spine is eroded close to its base, exposing the dorsal view of a longitudinal section ( Fig. 11 ). Dorsal 9 ( Figs 12 , 13 )

This vertebra is poorly preserved and the identification of its laminae is difficult, but no noticeable difference seems to be present in its general morphology compared with D8. Thus, the description is focused on its pneumatic features and not in the lamination. As is the case with D3 and D5, this vertebra was cut during the extraction of the specimen. The vertebra is cut along a transverse plane and the cross-section is relatively more anteriorly placed than those of D3 and D5, being just at the anterior margin of the pleurocoel. The exposed cross-section allows observation of the pneumatic features of this vertebra, both in the centrum and in the neural arch. The pleurocoel is round and slightly smaller than that of D10. The right pleurocoel was prepared and its ventral and dorsal communication with the right camera of the centrum can be observed ( Fig. 12 ). The cut exposes this large communication ( Fig. 12 ), which occupies the entire pleurocoel, except for its posterior border. A similar bony strut is located in the posterior region of the pleurocoels of other vertebrae. Paired camerae extend from the pleurocoel into the centrum, expanding both ventrally and dorsally ( Fig. 12 ). The pneumatic spaces present in the centrum are true camerae (sensu Wedel et al., 2000) and the neural arch bears large pneumatic spaces interpreted as supraneural camerae. The exact percentage of pneumatic space cannot be measured with confidence, given the strong lateromedial compression and deformation of the specimen, but, as preserved, at least half of the crosssection is occupied by the pneumatic spaces. The camerae of the centrum communicate with the supraneural camerae (exposed on the right side), but unfortunately the neural canal cannot be confidently identified. The left and right sides differ in the degree of pneumaticity ( Fig. 12 ), but this difference seems to be product of the angle of the cross-section that is slightly oblique to the transverse plane of the vertebra (exposing a more posterior section on the left side).

Dorsal 10 ( Fig. 13 )

This vertebra resembles the first sacral more than any of the presacral vertebra. The centrum is stout and has a large, dorsally placed pleurocoel, without any posterior depression. In contrast to more anterior dorsals, the pleurocoel is more extended dorsally and almost as high as long.

The ACPL is not anterodorsally oriented, as in other presacral vertebrae, but the ventral half of this lamina runs dorsolaterally and the dorsal half is posterodorsally inclined (as a product of the deformation of the specimen; Fig. 13 ). The parapophysis is not preserved in this vertebra, and its former position cannot be determined with any certainty, given the deformation of the ACPL. The diapophysis is placed anterior to the end of the ACPL but this may also be a product of anteroposterior crushing of the specimen. In contrast to the condition of other presacral vertebrae, a well-developed prezygoparapophyseal lamina is present in this vertebra and in the first sacral (S1). The PCDL is absent and the paradiapophyseal lamina seems to be well developed, as in S1. The neural spine of this vertebra is mostly eroded and no laminae can be observed.

Sacrum ( Fig. 13 )

The pleurocoel in the first sacral vertebra is markedly smaller than those of the preceding dorsals. Of the sacrum, three complete and a fragment of a fourth centrum are preserved. All preserved sacrals are fused, but the first centrum is separated from the second by a strong posterior expansion, similar to the expanded articular ends in the dorsal vertebrae, whereas the second and third centra are fused without any visible suture and are separated only by a minor ventral swelling. The expansion between the third and the fourth sacral vertebra is again more pronounced than that between the second and the third. The ventral sides of the first three sacrals seem to have been broadly rounded ventrally, but are somewhat deformed. All the neural arches are damaged, more so in the posterior than in the anterior elements, and the neural spines are missing.

The first sacral has a dorsoventrally expanded parapophysis that spans the neurocentral suture ( Fig. 13 ). Behind the parapophysis, a small pleurocoel is present in the dorsal part of the centrum. As in the last dorsal, a stout centropostzygapophyseal lamina extends anteriorly from the postzygapophysis up to the posterior edge of the parapophysis. A small, short lamina extends from the parapophysis posterodorsally. As in D10, no posterior centrodiapophyseal lamina is present. A stout sacral rib extends ventrally from the anteriorly directed transverse process, and spans from the neural arch to the mid-height of the centrum.

In the second sacral vertebra, the attachment for the sacral rib extends from the mid-height of the centrum to the top of the neural arch ( Fig. 13 ). It expands further laterally and, apparently, formed part of the sacricoastal yoke. The lateral part of the sacricoastal yoke is probably represented by a long, irregular bony plate that was found alongside the sacral vertebrae, but is too poorly preserved to correlate with the broken attachments of the sacral ribs. The attachment of the sacral rib consists of a thin vertical lamina and a stout, ventral antero-posterior expansion. At approximately the level of the neurocentral suture, a stout ridge expands from the posterodorsal part of the vertebral centrum and meets the sacral rib anterodorsally and laterally. A thin, strongly damaged lamina runs from the anterior part of this ridge dorsally, but its dorsal extent is uncertain, owing to the damage to the neural arch. No pleurocoel is present in the centrum, but a shallow depression is found on the posterior part of its lateral surface, between the ventral expansion of the attachment for the sacral rib and the posterodorsal ridge.

Only the centrum is preserved of the third sacral vertebra, including the ventral part of the attachment of the sacral rib, which is identical to that of the preceding vertebra.

The morphology of the vertebrae and their attachments to each other thus suggest that the sacrum of Tehuelchesaurus consisted of at least four vertebrae. Based on the presence of a pleurocoels and the neural arch lamination, the first sacral is interpreted as a dorsosacral, whereas the abrupt change in morphology in the next vertebra indicates that this and the subsequent element are the true sacrals. Thus, the fourth sacral would be a caudosacral. If one assumes that Tehuelchesaurus had five sacrals, as is the case in other neosauropods (Upchurch, 1998; Wilson & Sereno, 1998; Wilson, 2002), the last sacral should therefore be a second caudosacral. This is in contrast to the model of formation of the sacrum in neosauropods proposed by Wilson & Sereno (1998), who reconstructed the incorporation of two dorsosacrals, and in accordance with the interpretation of McIntosh et al. (1996) for Camarasaurus lewisi .

Ribs

All of the ribs preserved with the type of Tehuelchesaurus are dorsal ribs and are fragmentary. They seem to have the typical morphology of the dorsal ribs of sauropods, with a double-headed proximal end, an L-shaped cross-section in the proximal shaft, and a rather massive, plate-like distal shaft. None of the fragments has any sign of pneumatization, and no pneumatic foramina or depressions are present in the few preserved parts of the proximal ends of the ribs. Although a few fragments of more slender ribs are present, the dissociation of ribs and vertebrae and the fragmentary nature of the rib material make it impossible to say whether there was a distinction of ‘thoracic’ (anterior) and ‘lumbar’ (posterior) ribs in this animal, as is present in other sauropods (Rauhut et al., 2005).

Figures

Figure 2. A, schematic drawing of the type specimen of Tehuelchesaurus as found in the quarry, showing the semiarticulated position of the skeleton. B, outline reconstruction of Tehuelchesaurus, indicating recovered elements (rib fragments not shown). Scale bars = 100 cm. For abbreviations see text.

Figure 3. Tehuelchesaurus benitezii, ventrolateral view of the first preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 4. Tehuelchesaurus benitezii, ventrolateral view of the second preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 5. Tehuelchesaurus benitezii, ventrolateral view of the third preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 6. Tehuelchesaurus benitezii, ventrolateral view of the fourth preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 7. Tehuelchesaurus benitezii, lateral view of the fifth preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 8. Tehuelchesaurus benitezii, lateral view of the sixth preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 9. Tehuelchesaurus benitezii, lateral view of the seventh preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 10. Tehuelchesaurus benitezii, lateral view of the eighth preserved vertebra. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.

Figure 11. Tehuelchesaurus benitezii, horizontal crosssection through the base of the neural spine of the eighth dorsal vertebra. Scale bar = 10 cm. For abbreviations see text.

Figure 12. Tehuelchesaurus benitezii, transversal cut of the ninth preserved dorsal vertebra, showing the complex internal camerate system in grey. For position of crosssection see Figure 13. Scale bar = 10 cm. For abbreviations see text.

Figure 13. Tehuelchesaurus benitezii, lateral view of the posterior half of the ninth and tenth preserved dorsal vertebra and the four anterior sacral vertebrae. The cross-section in Figure 12 corresponds to the anterior end of the articulated series. Stereophotographs (A) and line drawing (B). Scale bar = 10 cm. For abbreviations see text.