The Appendicular Skeleton of Neuquensaurus, a Late Cretaceous Saltasaurine Sauropod from Patagonia, Argentina

Neuquensaurus, from the Late Cretaceous of Argentina and one of the first dinosaurs described from Patagonia, is one of the most derived sauropod dinosaurs, and its proportions and size place it among the smallest sauropods ever known. In this context, Neuquensaurus is central to understanding late stages of sauropod evolution. This contribution offers a full description of the appendicular skeleton of Neuquensaurus. The anatomical analysis reveals that the appendicular skeleton of Neuquensaurus exhibits unique characteristics only shared with closely related saltasaurine titanosaurs; for example, the laterally directed preacetabular lobe of the ilium, the prominent fibular lateral tuberosity, and the presence of an intermuscular line on the femoral shaft, which is proposed here as a synapomorphy of Saltasaurinae. Neuquensaurus also displays many reversals to primitive character states, such as the presence of a prominent olecranon process of the ulna, a trochanteric shelf, a lesser trochanter and an ischial tuberosity. Additional characters that allow its evaluation in a phylogenetic context are here provided. Among them are the extremely deflected femoral shaft, the elliptical femoral cross-section, and the anterolaterally oriented cnemial crest.

"Titanosaurus" australis was erected and first described by Lydekker (1893) based on a series of associated caudal vertebrae and some elements of the limbs recovered from Neuquén Province, Patagonia, mostly belonging to the same individual (Lydekker 1893: 4). As noted by Wilson and Up− church (2003: 139), Lydekker does not specify how many in− dividuals those elements belongs to, and the fragments of the girdles and limbs were not associated with the type caudal vertebrae (Wilson and Upchurch 2003: 139). Huene (1929) later referred to "Laplatasaurus" araukanicus Huene, 1929 some elements previously assigned to "T". australis by Ly− dekker and made an extensive description of that material, with the inclusion of numerous elements (mostly belonging to several adult and sub−adult individuals) collected in the early 20 th century in the course of fieldwork carried out by the Museo de La Plata, Argentina. The collected bones were discovered intermixed; hence Huene couldn't deter− mine single individuals: "The separation (of the bones) piti− fully had to be made by examination; therefore, errors are not excluded" (Huene 1929: 23, translated from the Spanish). Huene made a classification of the limb bones housed at the Museo de La Plata and assigned to the genus Titanosaurus, according to their peculiar shape and relative proportions, recognizing two Patagonian taxa: "Titanosaurus" australis and "Titanosaurus" robustus Huene, 1929. Huene (1929 classified the long bones of "Titanosaurus" australis and "T". robustus "…without determining or differentiating the vertebral material of each species … Huene (1929) used the name of "Titanosaurus" australis in an arbitrary way to iden− tify the form possessing slender limb bones and creating for the remainder the species "T". robustus, without taking into account the fact that the type material of the species "T". aus− tralis … consists of a series of caudal vertebrae" (Powell 2003: 43). Though Huene's descriptions were detailed and helpful, they were not extensively comparative with other sauropods yet known. Those taxa received scant attention for some 50 years, until Bonaparte and Gasparini (1978) re−stud− ied limb bones referred by Huene (1929) to "T". robustus (i.e., left femur, left ulna, right ulna, and left radius). They specified lectotype for those materials, indicating that they may correspond to the same individual (Bonaparte and Gas− parini 1978: 397). Powell (2003, adapted from his disserta− tion written in 1986) also revised the specimens of Titano− saurus and reconsidered the anatomy and validity of both species of the genus. He observed that the Indian type species of the genus Titanosaurus (Titanosaurus indicus Lydekker, 1877) more closely resembles "Laplatasaurus" araukanicus than "T". australis. Accordingly, the latter was included in a new genus, thus erecting Neuquensaurus australis as a new taxonomic entity with a modified diagnosis (Powell 2003), while N. robustus was regarded as a nomen dubium (Powell 2003;Wilson and Upchurch 2003). Subsequently, McIntosh (1990) tentatively referred "T". australis and "T". robustus to the genus Saltasaurus, arguing that the differences be− tween those taxa noted by Bonaparte and Powell (1980) are not of taxonomic importance (McIntosh 1990: 395). Powell (1992) and later Wilson and Upchurch (2003) did not recog− nize genus−level differences between those species. Salgado et al. (2005) recently described a new specimen of N. austra− lis, adding to information on axial and appendicular elements known for the species, and provided a revised diagnosis. Ad− ditionally, Salgado et al. (2005) include in their description of the new specimen other elements that were found associ− ated with the latter and they "...provisionally interpreted [them] as belonging to the same genus" (Salgado et al. 2005: 625). Moreover, other newly discovered material potentially belonging to Neuquensaurus remain undescribed and are in− cluded in the present analysis.
The present study is focused on the appendicular anat− omy of Neuquensaurus. Bearing in mind its disarticulated condition (which has made a detailed study of its osteology difficult), the new discoveries of the last years, and unpub− 1 m  Wilson and Sereno (1998).
Remarks.-All currently materials referred to the appendi− cular skeleton of N. australis mentioned by Lydekker (1893), Huene (1929), Powell (2003), and Salgado et al. (2005) are listed in Appendix 1, as well as those specimens referred by Huene (1929) and Powell (2003) to N. robustus. The new specimens not yet published are also included in Appendix 1. In some cases, the original remains are presumed to be miss− ing, so interpretations were based on illustrations in Huene (1929). In the worst cases, neither material nor drawings were available at all (see Appendix 1 for details).

Description
It is remarkable that a complete analysis including the axial skeleton of both Neuquensaurus australis and Neuquen− saurus robustus is needed to asses the definitive assignment of all the elements to the former valid taxon and to elucidate the taxonomic status of the latter. Such study is, of course, out of the scope of this contribution. I will focus the descrip− tions primarily on the multiple associated hind limb elements that Huene (1929) referred to "Titanosaurus" australis. Ele− ments previously referred to "Titanosaurus" robustus will be described in each section devoted to the respective element only if there is a reason to believe that they probably repre− sent N. australis. Where they differ in morphology from N. australis, each will appear at the end of the corresponding section, with some comments. Any elements for which refer− ral to N. australis is dubious will be treated in a separate de− scriptive section as cf. Neuquensaurus.
The phylogenetic relationships of Titanosauria remains obscure, in part, because the fragmentary nature of most gen− era. To avoid nomenclatural ambiguities, I will follow the phylogenetic definitions for Titanosauria as follows: Titanosauria Bonaparte and Coria, 1993: Andesaurus delgadoi Clavo and Bonaparte, 1991, Saltasaurus loricatus Bonaparte and Powell, 1980, their most recent common an− cestor, and all descendants.
Pectoral girdle Huene (1929) mentioned the existence of thirteen bones of the pectoral girdle of Neuquensaurus. However, of those, eleven are present today in the collections of the Museo de La Plata (Appendix 1). There is also a left coracoid fused to a doi:10.4202/app.2009.0099 fragment of scapula (MLP−CS 1298) that was previously de− scribed as a fragment of ilium (Huene 1929;Powell 2003) and is re−described herein. Scapula ( Fig. 2A, B).-The description of the scapula is mainly based on MLP−CS 1096 ( Fig. 2A) because is the better−preserved specimen. The scapula and coracoid are co−ossified, as in Opisthocoelicaudia (Borsuk−Białynicka 1977). The scapula consists of two well−defined portions, a wide proximal part and a narrow distal, elongated scapular blade. The whole structure has a sigmoid shape in dorsal view and is medially curved. The ventral margin of the scap− ular blade is straight while the dorsal edge is sigmoid, with its proximal portion narrower than the distal one, which is ex− panded as in Saltasaurus. The dorsal margin of the proximal portion of the scapula is rugose, at the site where the anterior portion of the M. levator scapulae originated. The acromion is medially curved and laterally concave, as in Saltasaurus. The contact between the latter and the scapular blade is U−shaped. The scapular blade has a longitudinal ridge on its lateral surface (Huene 1929;Salgado et al. 2005). The proxi− mal portion of the scapula is in contact with the coracoid, forming a nearly 90°angle glenoid fossa, resulting in a sub−triangular shape of the fossa. The glenoid is thick and faces ventrolaterally. The glenoid lip of the scapula presum− ably faced anteroventrally. There is a depression on the proximolateral surface of the scapula, also seen in Salta− saurus, which is interpreted as the supracoracoideal fossa, origin site of the M. deltoides scapularis (M. scapulohume− ralis anterior sensu Borsuk−Białynicka 1977). There are also a fragment of right scapula (MLP−CS 1129) and a fragment of left scapula (MLP−CS 1301) that may belong to N. austra− lis. The former was referred by Huene (1929: 36) to probably pertain to the same individual as MLP−CS 1096, which is likely. Both specimens have the same proportions and gen− eral outline, and also present a flat, rugose muscular scar pos− terior to the acromion. On the other hand, the also fragmen− tary specimen MLP−CS 1301 has the general outline and the medially−curved scapular blade of MLP−CS 1096. As pointed out by Huene (1929), it is highly probable that MLP−CS 1301 belongs to a juvenile specimen of N. australis.
The general aspect of the scapula of N. australis resem− bles that of other Titanosauria, such as Saltasaurus, Opistho− coelicaudia (Borsuk−Białynicka 1977), Lirainosaurus (Sanz et al. 1999), and Alamosaurus (Gilmore 1922). However, the scapula of the Patagonian specimen differs from these titano− saurs in having a glenoid lip that ends in a right angle.
Coracoid ( Fig. 2A-C).-There are three preserved cora− coids (MLP−CS 1096, MLP−CS 1298, and MLP−Ly 14). Two of them (MLP−CS 1096 and MLP−CS 1298) are co−ossified with the respective scapula. The coracoid corresponds in size to the proximal portion of the scapula. It is stout and firmly fused to the scapula along its posterior surface although the suture is not evident. The ventral margin of the element con− tacts the anterior margin in a nearly 90°angle, giving the coracoid a quadrangular outline, resembling that of Salta− saurus (Wilson 2002;Powell 2003). The glenoid (ventral) portion is notably rugose and thick, particularly the infra− glenoid lip. The medial surface is flat, while the lateral sur− face presents a concavity, origin site of M. supracoracoideus. There is a slight ridge anterior to the concavity, perpendicu− lar to the dorsal margin, which Huene (1929: 37) referred to as the attachment site of pectoral musculature. It may actu− ally correspond to the M. coracobrachialis. In MLP−CS 1096 and MLP−CS 1298 the coracoid foramen is not evident, al− though there is a slender non−perforated hole close to the margin of the scapular contact.
Sternal plates (Fig. 2D, E).-There are two sternal plates mentioned by Huene (1929) as belonging to Neuquensaurus australis (MLP−CS 1104 and MLP−CS 1260). The general outline of the sternal plate is crescentic as in other Titanosauria (Salgado et al. 1997;Wilson 2002;Curry Rogers 2005), with lateral margins strongly concave. The anterior portion is ro− bust and becomes thinner towards its lateral and posterior bor− ders. The anteroventral region has a stout crest which runs anteroposteriorly, and was the origin site of M. pectoralis (Huene 1929;Borsuk−Białynicka 1977). The crest is ventro− laterally oriented. The dorsal surface is almost flat. The right sternal MLP−CS 1104 and the left sternal MLP−CS 1260 are very similar in size and general proportions, so that they are symmetrically equal. As pointed out by Huene (1929: 36), it is very probable that pertain to a single individual.
Crescentic sternal plates are also present in Rapetosaurus (Curry Rogers 2009); Alamosaurus (Lucas and Hunt 1989), Opisthocoelicaudia (Borsuk−Białynicka 1977) and Saltasau− rus (Powell 2003). Nonetheless, the most interesting features of the sternal plates present in N. australis are their large size and the presence of the large anteroventral ridge. A similar ridge is present in Saltasaurus (Powell 2003: pl. 39b), al− though it is much less developed than in the Patagonian spec− imens.
There is also a right sternal plate (MLP−CS 1295) re− ferred by Huene (1929) and Powell (2003) as a left sternal of N. robustus. I consider these as belonging to N. australis due to their close resemblance, general outline, and the presence of the well developed anteroventral crest (contra Huene 1929: 36).

Forelimb
Several elements of the forelimb are represented, including well preserved right and left humeri, ulnae and radii; how− ever many others elements described by Huene (1929) are missing.
Humerus (Fig. 3).-Nine humeri are preserved in total. The humerus is a robust bone, as in other Titanosauria (robust− ness index, RI = 0.305-0.339, Appendix 2A), but more slen− der than that of Opisthocoelicaudia (RI = 0.37, Wilson and Upchurch 2003). The proximal and distal portions are ex− panded, particularly the former, reaching in some cases (e.g., MLP−CS 1050) 50% of the total length of the bone. The proximal portion is slightly medially oriented with respect to the distal end, as in Saltasaurus (Powell 2003). It is medio− laterally expanded and anteriorly concave. The humeral head is rounded and well developed. The lateral margin of the diaphysis is also concave. The proximal surface has its greater robustness in the central part, corresponding to the humeral head, being more slender on its lateral and medial margins. The dorsal edge of the proximal end is straight and forms a 90°angle with the lateral margin, as in Saltasaurus and Opisthocoelicaudia (Borsuk−Białynicka 1977: fig. 7B). The most notable features of the proximal portion are the above−mentioned mediolateral expansion and the robust deltopectoral crest, which runs down the lateral edge of the anterior face of the proximal half of the bone: this is longitu− dinally oriented, and slightly medially twisted. This structure has a rugose surface, which was the site for the attachment of the abductor musculature (i.e., M. pectoralis, M. dorsalis scapulae, and M. deltoides scapularis). There is a deep sur− face on the anteroproximal portion of the humerus, medial to the deltopectoral crest, which is interpreted as the site of in− sertion of the M. coracobrachialis ("coracobraquial breve" sensu Huene 1929; see also Powell 2003). The posterior sur− face has a tuberosity placed posteroventrally to the delto− pectoral crest. This structure is also seen in Opisthocoeli− caudia (Borsuk−Białynicka 1977: fig. 7D) and Saltasaurus, although it is less developed in these taxa than in Neuquen− saurus. This tuberosity was the site of insertion of M. latissi− mus dorsi, not the "braquial inferior" (contra Huene 1929).
The humeral shaft is mediolaterally expanded and its cross section is approximately elliptical, with its antero− posterior length 70% of the mediolateral breadth (eccentric− ity index, ECC index = 1. longitudinally oriented convex area flanked by two depres− sions, which correspond to the origin site of the humeral heads of M. anconeus. The distal end of the humerus is less expanded than the proximal one. The condyles are asym− metrical, being the lateral condyle the more robust. They are separated by the cuboid fossa, which is well developed in saltasaurines.
A proximal portion of a right humerus (MLP−CS 1019) was referred by Huene (1929) to "Titanosaurus" robustus. As he pointed out, this bone lacks the acute angle between    (Huene, 1929) nomen dubium (MLP−CS definitively taxonomic identity to that bone, I find no reason to consider MLP−CS 1019 as different from N. australis. The most notable features of Neuquensaurus humeri are their robustness and the mediolateral development of the proxi− mal portion, as well as the almost right angle between the dor− sal and lateral margins of the proximal portion, also seen in other Titanosauria, such as Saltasaurus (PVL 4017−92, Powell 1992), Opisthocoelicaudia (Borsuk−Białynicka 1977), Alamo− saurus (Lucas and Hunt 1989), and Magyarosaurus (McIntosh 1990: fig. 16.10).
Ulna (Fig. 4).-Ten ulnae were mentioned by Huene (1929), although only eight of those can be located, one of those of dubious affinities. The proximal portion of the ulna is wide and is formed by three structures. Two conspicuous ridges anteromedially and anterolaterally directed, respectively, frame the olecranon on both sides: the anterolateral (AL), and the anteromedial (AM) processes (Bonnan 2003: 607). The third structure is the olecranon process, which is placed posterolaterally and was the insertion site of the tendons of M. anconeus. It is a well defined structure, although it does not protrude above the articular surface. Those three ele− ments (the olecranon plus the two processes) define a tri− radiate proximal cross−section. The radial (anterior) and me− dial surfaces are concave. There is a longitudinal ridge on the radial surface that corresponded to the origin site of M. pro− nator quadratus (Huene 1929;Meers 2003). There are also two left ulnae (MLP−CS 1053 and MLP−CS 2004) which Huene (1929: 39) and Powell (2003: 39) both referred to N. australis. However, those bones does not resemble the slender appearance of the ulna of N. australis, but have the stout look and extremely developed ulnae of N. robustus (MLP−CS 1094 and MLP−CS 1095), which constitute part of the lectotype designed by Bonaparte and Gasparini (1978). The olecranon process of MLP−CS 1053 (Huene 1929: pl. 11: There are two additional radii with dubious affinities. The ra− dius is a rather bent bone. Its proximal end is more expanded than the distal one; it is roughly oval in proximal view and its dorsal margin is straight, with rugosities on the proximal and distal ends. The proximal portion is medially expanded. The anti−ulnar (anterior) face is straight, while the ulnar (posterior) face is convex. On the latter there is a furrow flanked by two ridges oriented obliquely from the anteromedial to the posterolateral side of the diaphysis ("interosseous ridge", Curry Rogers 2009). The medial ridge could correspond to the insertion site of the M. pronator teres (see also Huene 1929;Borsuk−Białynicka 1977). The distal surface of the bone is el− liptical and its distal margin is oriented obliquely with respect to the long axis of the diaphysis, from the ventromedial to the dorsolateral side. robustus and has less expanded proximal and distal ends: hence, its assignation to N. australis is probably correct. (Fig. 6).-The only carpal (MLP−CS 1234) tentatively assigned to "T". australis by Huene (1929: pl. 12: 1) is missing. The overall shape is rounded although its proximal surface is almost pyramidal. No other anatomi− cal details can be gleaned from Huene's drawing.  There are three metacarpals (MLP−CS 1197, MLP−CS 1189, and MLP−CS 1238, Fig. 6) that were erroneously as− signed by Huene (1929) as metatarsals of "Titanosaurus" robustus (Powell 2003). Metacarpal II (MLP−CS 1197, Fig.  2A) is columnar, with expanded ends. The proximal portion is rugose and has a triangular outline, with the apex on the palmar side. The medial side of the triangle is convex and ar− ticulated with the concave surface of metacarpal I (Ape− steguía 2005). On the proximomedial side there is a short, longitudinal ridge facing downward, which is the articulation area for metacarpal I. The lateral and anteroproximal sides of the bone are flat. However, there is a longitudinal ridge that extends from the middle of the shaft to the distal portion, close to the distal end. The distal part of metacarpal II is qua− drangular in outline and bears rugosities.

Carpus and manus
Metacarpal III (MLP−CS 1189, Fig. 2B) is similar to metacarpal II in general outline. Its proximal portion has a triangular shape, with slight convex sides. The diaphysis is columnar with a triangular cross−section, while the distal end is quadrangular in cross section. Rugosities are present on the proximal and distal portions and the anterior side of the shaft is flat. The lateral side has a ridge flanked by rugosities.
Metacarpal IV (MLP−CS 1238, Fig. 2C) has a character− istic subrectangular cross−section in proximal view (Ape− steguía 2005). The lateral and medial sides are concave for articulation with metacarpals V and III, respectively. The an− terior surface is almost flat and the palmar side is broader proximally. On the lateral and medial sides of the proximal end there are two ridges flanking both sides that probably correspond to attachment sites for tendons (Huene 1929).
As in other neosauropods (e.g., Diplodocus, Camara− saurus, Brachiosaurus, Rapetosaurus, Opishocoelicaudia), the proximal end of metacarpals II and III form triangular wedges in proximal views. Metacarpal IV of Neuquensaurus shares with other titanosaurs the presence of a subrectangular proximal end with concave sides for articulation with meta− carpals III and V (Apesteguía 2005).

Pelvic girdle
Several ilia, ischia and pubes were previously described (Lydekker 1893;Huene 1929;Powell 2003;Salgado et al. 2005). Eleven incomplete ilia of Neuquensaurus australis and materials referred to Neuquensaurus robustus were de− scribed by Lydekker (1893) Salgado et al. (2005) assigned to N. australis an almost complete pair of ilia (MCS−5/16) fused to the sacrum, and a fragment of ischium (MCS−5/24). There is also a fragment of ischium that proba− bly pertains to the genus that has not previously been de− scribed (MPCA−CS 001) and is described here for the first time. 408 ACTA PALAEONTOLOGICA POLONICA 55 (3)  Ilium (Fig. 7).-The description of the ilium is based on the original material described by Lydekker (1893) (MLP−Ly 17), a fragment of left ilium described by Huene (1929) (MLP−Av 2069) as belonging to "Titanosaurus" robustus, which I consider more probably that of N. australis, and those elements described by Salgado et al. (2005) (MCS− 5/16). The ilium has both expanded preacetabular and post− acetabular portions. The preacetabular lobe of MCS−5/16 (Fig. 7A) is subhorizontally oriented and projects laterally, as in others titanosaurs (Borsuk−Białynicka 1977;Salgado et al. 1997Salgado et al. , 2005Jain and Bandyopadhyay 1997;Upchurch 1998). The whole iliac blade has a "twisted" configuration, so that the outside surface of the preacetabular lobe faces up− ward, whereas the outside surface of the postacetabular lobe faces downward (Salgado et al. 2005). The pubic peduncle is transversely expanded and anteroventrally directed, and its ventral (distal) surface has a triangular shape, with one of the vertices pointing inwards. The ischiadic peduncle is poorly developed, as in other sauropods (Wilson 2002). The shape of the preacetabular lobe is semicircular and it faces anterodorsally when the ilium is oriented with the ischial and pubic peduncles in the same plane (Salgado et al. 1997). The fragment of left ilium described by Huene (1929) (MLP−Av 2069) as belonging to "T". robustus, I consider more closely similar to that of N. australis because of its general proportions, the mediolaterlal development of the pubic peduncle and the same angle between the preace− tabular lobe and the pubic peduncle.
An anteroventrally directed pubic peduncle is also re− ported in Opisthocoelicaudia (Borsuk−Białynicka 1977: fig.  12). The most noteworthy feature of the ilium of Neuquen− saurus is the lateral projection of the preacetabular lobe with respect to the long axis of the ilium (Salgado et al. 2005). This condition is related to the great development of the hind limb extensor musculature (Otero and Vizcaíno 2008). This condition is also present in other saltasaurines, such as Salta− saurus  and Rocasaurus (MPCA−Pv 46), and non−titanosaur sauropods, such as Camarasaurus (Osborn and Mook 1921: figs. 49, 87). Ischium (Fig. 8).-The description of the ischium is based on MCS−5/24 and a hitherto undescribed, well preserved but in− complete right ischium (MPCA−CS 001). Additionally, Huene (1929: 40, pl. 14: 3) mentioned the existence of a fragment of a left ischium (MLP−CS 1261) that resembles MCS−5/24. The ischium is, as in other titanosaurs (see Salgado et al. 1997: fig. 5), a short bone with a relatively broad blade. This could be re− lated to the development of the site of origin of the adductor musculature (Otero and Vizcaíno 2008). MCS−5/24 is slender, more so than MPCA−CS 001. The latter is a robust bone, showing thickened articular surfaces. The iliac peduncle is well developed and stout, and is separated from the main body of the ischium (Curry Rogers 2005: character 332), as in Saltasaurus and Rocasaurus. The pubic peduncle, only pre− served in MPCA−CS 001, is extensive, as in other titanosaurs. In MPCA−CS 001 there is a protuberance on the lateral surface of the posterior margin, over the line of the pubic contact, also reported in MLP−CS 1261 (Huene 1929: 41). This is the ischial tuberosity, an elongated process with rugosities over the surface, which was the site of origin of the M. flexor tibialis internus 3 (Borsuk−Białynicka 1977; Hutchinson 2001aHutchinson , 2002. The posterior margin is similar to that of Saltasaurus, and differs from Rocasaurus in being less concave. The assignment of MPCA−CS 001 to Neuquensaurus australis is mainly based on the presence of the ischial tuberosity, which is mentioned by Huene (1929: 41). This tuberosity can not be seen in MCS−5/24 because the ischial blade is damaged. The ischial tuberosity is also reported in other Titanosauria, such as Opisthocoelicaudia The ischium of Neuquensaurus has a similar morphology to that of other Titanosauria (e.g., Saltasaurus, Rocasaurus, Aeolosaurus, Isisaurus, Alamosaurus) in which the whole structure has a semilunate shape with a distally expanded blade.
Pubis (Fig. 9).-Five incomplete pubes of Neuquensaurus are preserved. Only MLP−CS 1102 has a relatively well−pre− served shaft. The pubis is an expanded bone with thick proxi− mal and distal margins. The proximal end is wider than the entire shaft, while the distal end is as wide as the shaft. There is a longitudinal crest on the ventral surface of the bone, close to the lateral margin ("ventral crest", Powell 2003: fig. 43:  1b). The presence of the longitudinal crest determinates two parallel areas, which were the sites of origin of the M. Hindlimb Femur (Fig. 10).-Eleven femora of Neuquensaurus are pre− served. The femur is a large, mediolaterally expanded bone, as in other titanosaurs (mediolateral/anteroposterior breath more than 1.35, Appendix 2D). The femoral head is prominent, ro− bust and positioned dorsomedial to the greater trochanter, as in Rocasaurus and Saltasaurus. In MCS−5/27 and MCS−5/28, which was referred by Salgado et al. (2005) to N. australis, and MLP−CS 1480 (lectotype of N. robustus), the femoral shaft is straighter at its medial than at its lateral margin, be− coming more sudden at the beginning of the femoral head. The greater trochanter is present and placed laterally to the femoral head, although positioned at a lower level. In better−preserved materials (e.g., MCS−9), superficial rugosities can be observed over the surface of the greater trochanter. Distal to the greater trochanter, on the posterolateral surface of the proximal por− tion of the femora, there is a sigmoidal ridge that is here inter− preted as the remainder trochanteric shelf (Fig. 10A 3 , A 4 ), site of insertion the extensor M. ischiotrochantericus (Otero and Vizcaíno 2008). This structure was maintained throughout archosaurian evolution, being present both in ornithischians and saurischians (Novas 1996;Hutchinson 2001b, see Discus− sion). A lateral bulge, very well developed in MCS−9, is pres− ent on the lateral surface of the shaft, distal to the greater trochanter, as in other Titanosauriformes (Salgado et al. 1997;Wilson and Sereno 1998). The curved outline of the lateral bulge determines the medial deflection of the proximal third of the shaft (Salgado et al. 1997;Wilson and Sereno 1998;Wil− son 2002;Curry Rogers 2005; but see Discussion). The fourth trochanter, restricted to the posterlateral surface of the shaft, is low and rugose and is the site of insertion of the caudofemoral musculature.
There is a linea intermuscularis cranialis on the anterior surface of the shaft (Fig. 10A 1 , A 2 ) as in other Archosauri− formes (Hutchinson 2001b). This is a structure related to the distribution of the femoral extensor muscles of the leg (e.g., M. femorotibialis, Otero and Vizcaíno 2008). The linea inter− muscularis cranialis ("arista longitudinal" sensu Huene 1929: 42) is a dorsoventrally elongated crest along the midline of the anterior surface of the shaft. It begins at the level of the lateral bulge, extending to the intercondylar zone. A similar structure is also seen in Saltasaurus (PVL 4017−83, "long rugosity" sensu Powell 2003), Rocasaurus (MPCS−Pv 46), and Bona− titan (MACN−RN 821). The linea intermuscularis cranialis has its origin at the base of the greater trochanter, extending distally on to the posterior surface of the femur. That structure is particularly well developed in MCS−9. The distal condyles are prominent, with the fibular condyle more developed than the tibial one. As mentioned by Wilson and Carrano (1999), the distal portion of the femur has the condylar surface form− ing a dorsomedial angle with respect to the major axis of the femur. It determinates a dorsomedial inclination or "beveled condition" of the femoral shaft when in articulation with the tibia, producing a "wide−gauge" gait, typical of titanosaurs.
The femur of Neuquensaurus resembles that of Salta− saurus and Rocasaurus in its general robustness, the promi− nent femoral head, and the extreme mediolateral development of the diaphysis. It differs from the much more slender femur of Bonatitan (Appendix 2D). The femora of Neuquensaurus, Saltasaurus, Rocasaurus, and Bonatitan share the presence of the intermuscular line and prominent trochanteric shelf (Otero 2009). The femur of Neuquensaurus differs from that of other titanosaurs (e.g., Lirainosaurus, Sanz et al. 1999;Rapeto− saurus, Curry Rogers 2005) in having the femoral head oriented medially, rather than dorsomedially, when the distal condyles are aligned to the horizontal plane (Wilson and Carrano 1999).
The element MLP−CS 1480, which is part of the lectotype of N. robustus specified by Bonaparte and Gasparini (1978), has no substantial differences respect the femora referred by Huene (1929) and Salgado et al. (2005) to N. australis and are not here deemed of taxonomic differences. I consider those el− ements as belonging to the same taxon (see Discussion).
Tibia (Figs. 11, 12).-There are four complete tibiae, two right and three left. There are also two incomplete tibiae (MLP CS 1303, MLP CS 1093) and one complete tibia (MLP CS 1123) mentioned by Huene (1929) that is now missing. One of the complete tibiae assigned to N. australis (MCS− 5/25, Salgado et al. 2005) is articulated with the fibula (Fig.  12). The tibia is a relatively short and robust bone (Appendix 2E) as in Saltasaurus (PVL 4017−84), with well developed proximal and distal ends. The tibial diaphysis is elongated anteroposteriorly and flattened mediolaterally. The proximal articular face is oval−shaped and the cnemial crest is well−de− veloped, robust and anterolaterally oriented. The distal artic− obturator foramen 100 mm ular surface is heart−shaped. There is a concavity on the inner face of the cnemial crest in which the fibula articulates. The dorsal edge of the crest has a rugose area, which was the site of insertion for the common tendon of the extensor muscula− ture (e.g., Mm. iliotibiales, Mm. femorotibiales, Otero and Vizcaíno 2008). The area behind the crest has a proximo− distally elongate concavity that follows the length of the crest (see also Salgado et al. 2005). There is no significant differ− ence between MLP−CS 1264 (referred to N. robustus,  Bonaparte and Gasparini (1978); left femur in anterior (C 1 ) and posterior (C 2 ) views.
tibia (see above description). It is a slender, mediolaterally− compressed bone (Appendix 2F). The fibulae have a slightly sigmoid shaft. The proximal portion is anteroposteriorly elongated and the distal articular face is oval−shaped. The most notable feature of the bone is the lateral tuberosity, on the proximolateral surface of the bone. This structure, a characteristic feature of Eusauropoda (Wilson and Sereno 1998), is extremely well developed in Neuquensaurus (Powell 2003). With an oval−shaped geometry, it originates on the anteroproximal portion of the shaft and runs posterodistally toward the distal portion. This tuberosity was the site of insertion of the M. iliofibularis (Otero and Vizcaíno 2008), not the M. flexor digitorum longus (contra Borsuk−Białynicka 1977). (Figs. 12, 14).-There is a right astragalus fused to the tibia (MCS−5/29, Fig. 12G, N), hence the proxi− mal surface is not available for description. The astragalus is transversely narrower than the distal tibial surface (Salgado et al. 2005 Two first metatarsals, assigned to N. robustus (MLP−CS 1179 and MLP−CS 1185), are known. Metatarsal I is short and stout, with expanded proximal and distal ends, with the former more so than the latter. Its general outline is sub− rectangular, and its medial and lateral margins are concave. The distal articular surface is oriented obliquely with respect to the long axis. The element MLP−CS 1179 is considerably bigger than MLP−CS 1185; hence the latter is presumed to belong to a juvenile individual. Both materials were referred by Huene (1929) as "left" metatarsals; however they corre− spond to the right side. The first metatarsal of N. australis drawn by Huene (1929: pl. 17: 3), now missing and pre− sumed lost, is similar to those assigned to N. robustus, al− though the former is more slender.

Tarsus and pes
Metatarsal II (MLP−CS 1183) was originally described by Huene (1929) as a first metatarsal. It is re−described herein as a left metatarsal II. It is quite longer and gracile than meta− tarsal I. The diaphysis is transversely compressed, as seen in other titanosaurs (González Riga et al. 2008; Curry Rogers 2009). As in metatarsal I, the distal articular end has two de− fined condyles. The lateral one faces quite downwards.
There are two metatarsal III (MLP−CS 1191 and MCS−10). Its proximal end has an asymmetrical drop−shape, while the distal end has a symmetrical sub−rectangular shape. Huene (1929: 44) mentioned the existence of other two metatarsal IIIs; however, no materials or drawings could be found.
Metatarsal IV (MLP−CS 1193) has been described by Huene (1929) as the second metatarsal. It is longer than metatarsals I and II. Its proximal end is flat, has a sub− rectangular cross−section, and bears two notches at both me− dial and lateral sides for articulation with metatarsals II and IV, respectively. The proximal end is expanded and oriented perpendicularly with respect to the distal end. Metatarsal IV is the longest, with a sub−oval cross−section. The flat proxi−  the proximal end characteristically expanded dorsoventrally and a narrower distal end (McIntosh 1990). The proximal portion is rugose, the lateral surface is convex, and the me− dial surface also bears rugosities and has a longitudinal ridge. There is a left metatarsal V (MLP−CS 1182), referred by Huene (1929) to "Titanosaurus" robustus, which is stouter than MLP−CS 1180 (the proximal major breadth/shaft length ratio of MLP−CS 1182 is 2.8, and the same ratio in MLP−CS 1180 is 5.3). There are two preserved pedal phalanges. One of them (MLP−CS 1206) was previously incorrectly assigned to the manus (Huene 1929;see above). MLP−CS 1206 is very short, longitudinally compressed and mediolaterally expanded. The other phalanx (MLP−CS 1184) corresponds to the first digit. It is a short and robust bone with a slight depression on its proximal end, which articulates with the metatarsal I. The distal end is triangular.
Metatarsals I, II and V of Neuquensaurus are the smallest and robust. Metatarsal I is the stoutest of the pes and resem− bles that of other neosauropods, such as Apatosaurus (Bonnan 2005), Rapetosaurus (Curry Rogers 2009), and Opisthocoelicaudia (Borsuk−Białynicka 1979). Metatarsals III and IV are the largest and are most gracile, as in other sauropods (McIntosh 1990), and are very similar to those of Rapetosaurus (Curry Rogers 2009). The "paddle−like" con− dition of metatarsal V of Neuquensaurus is shared with other sauropods (Bonnan 2005).

cf. Neuquensaurus
Two additional fragments of scapulae (MLP−CS 1296 and MLP−CS 1292) were mentioned by Huene (1929) as belong− ing to "Titanosaurus" australis. However, those elements do not resemble MLP−CS 1096 and MLP−CS 1129 at all. One of those (MLP−CS 1292) is a small scapular blade, but much more gracile that MLP−CS 1301. The other element (MLP− CS 1296) is a scapular blade that lacks the distal portion. Due to its bigger size and fragmentary condition it is difficult to refer MLP−CS 1296 to Neuquensaurus australis. The speci− men MCS−7 was found associated with other remains as− signed to N. australis (Salgado et al. 2005). As Salgado et al. (2005: 630) pointed out, "it is smaller than the associated bones and lack the characteristic species features" (e.g., strongly sigmoid dorsal ridge). Also, MCS−7 has a straight scapular blade while MLP−CS 1096 has a medially curved blade. As Salgado et al. (2005) commented, there is no evi− dence besides the topographical association supporting re− ferral of MCS−7 to Neuquensaurus.
The fragments of left (MLP−CS 1167) and right (MLP−CS 1174) radii assigned by Huene (1929) and Powell (2003) to N. australis do not resemble those referred to N. australis nor to the lectotype of N. robustus. Actually, the formers are less ro− bust, do not have a well developed interosseous ridge, and its distal ends are less expanded. It is probably that MLP−CS 1167 and MLP−CS 1174 were wrong assigned by those authors and actually pertain to a different genus from Neuquensaurus (e.g., Titanosauria indet.) Huene (1929) assigned with doubt distal portions of metacarpal II (MLP−CS 1186), metacarpal III (MLP−CS 2003), and metacarpal IV (MLP−CS 1187) to "T." australis; however, they are much larger than the remaining metacarp− als assigned to the genus and thus their assignment is tenta− tive (see also Powell 2003).
There are three fragmentary ilia (MLP−CS 1056, MLP−CS 1057, and MLP−CS 1258) that Huene (1929) referred to "T". australis. Unfortunately those elements only preserve the pu− bic peduncle and part of the preacetabular lobe; hence their as− signment to a single species is vague. They do not resemble the original material described by Lydekker (1893) or that de− scribed by Salgado et al. (2005). The main differences lie on the mediolateral development of the pubic peduncle, which is more mediolaterally expanded in MLP−Ly 17 and MCS−5/16. Besides, the angle between the pubic peduncle and the preace− tabular lobe is approximately 80°in MLP−Ly 17 and MCS− 5/16 and nearly 60°in MLP−CS 1056, MLP−CS 1057, and MLP−CS 1258. Therefore, I consider the latter elements as cf. Neuquensaurus. Huene (1929) and Powell (2003) mention the existence of a fragment of left pubis (MLP−CS 1263) as belonging to N. australis. This bone is highly damaged and has nothing in common with the elements referred to N. australis.

Discussion and comparisons Phylogenetically relevant characters
The most significant morphological features in the anatomy of Neuquensaurus are present in the appendicular skeleton (Huene 1929;Wilson and Carrano 1999;Wilson 2002;Powell 2003;Otero and Vizcaíno 2008). As for other mem− bers of Saltasaurinae (e.g., Rocasaurus and Saltasaurus), the morphology of the appendicular skeleton of Neuquen− saurus differs from the typical sauropod limb pattern. The major anatomical changes of the appendicular skeleton of Neuquensaurus, as well as its most phylogenetically infor− mative characters, are discussed below (Fig. 15).
Olecranon.-The ulnae of sauropod outgroups possess a primitively well−developed olecranon process (Young 1941;Cooper 1981). This structure is reduced in Vulcanodon and basal sauropods (Cooper 1984;Wilson and Sereno 1998;Wil− son 2002). The reduction of that process within Sauropoda en− abled the alignment of the elbow joint, resulting in a typical columnar limb, which accommodated the extreme loadings achieved in large−bodied sauropod (Wilson and Sereno 1998;Wilson 2005). There are intermediate states of development of the olecranon process in some sauropods such as Apatosaurus ("posterior process" sensu Wilhite 2003: fig. 2.16B), with the process becoming particularly prominent within Titanosauria (Wilson and Sereno 1998;Wilson 2002;Powell 2003). The presence of a well−developed olecranon process is a character− istic reversal feature of Neuquensaurus and other members of Titanosauria (i.e., Saltasaurus, Wilson 2002: character 167). Janenschia, a camarasaurid sauropod (sensu Bonaparte et al. 2000) also displays a well defined olecranon process that pro− doi:10.4202/app.2009.0099 jects above the articular surface of the ulna (Upchurch 1995: fig. 14B), indicating that this is not an exclusive feature of titanosaurs, but that it is more broadly distributed among macronarians.
As Apesteguía (2004) pointed out, the olecranon process, coupled with others features, has been hypothesized as an im− portant functional element in the nesting behavior of Salta− saurinae.
Iliac blade.-The "twisted" configuration in the ilium of Neuquensaurus is the most significant pelvic innovation present in Saltasaurinae (Salgado et al. 2005). As a result, the pre− and postacetabular lobes are oriented outward. As in sauropod outgroups (i.e., prosauropods, theropods) and also ornithischians, Neuquensaurus has a preacetabular lobe of the ilium that is equal to or less than iliac length (Wilson and Sereno 1998: character 82). However, in Neuquensaurus as well as in other saltasaurine titanosaurs (i.e., Saltasaurus, Rocasaurus), this configuration is achieved not by a separa− tion of entire iliac blades, but by a lateral divergence of the preacetabular lobes. This is a derived feature (Wilson 2002: character 187) that was proposed to be related to the attach− ment for the iliopsoas complex (as in the extinct therizino− saurid theropods and the extant xenarthrans; Apesteguía 2004). The lateral divergence of the preacetabular lobes, thus produce an alignment between the femoral protractor lines of action and the direction of travel (Wilson 2005), increasing the anteroposterior component of the line of action of leg extensor muscles (Otero and Vizcaíno 2008).
Ischial tuberosity.-The ischial tuberosity is an ancestral feature present in basal Reptilia (Hutchinson 2001a). In sauropodomorphs it is a rounded scar on the proximolateral surface of the ischium and, in crocodilians, it is the site of in− sertion of M. flexor tibialis internus 3 (Hutchinson 2001a(Hutchinson , 2002Otero et al. in press). In Neornithes, the ischial tuberosity shifts toward the proximodorsal process of the ischium, representing another character state for the same feature, and maintaining the muscular correlate (Hutchinson 2001a(Hutchinson , 2002. The presence of a rounded ischial tuberosity in basal sauropods such as Patagosaurus (Hutchinson 2001a), as well as in terminal forms like Neuquensaurus and other titanosaurs (e.g., Opisthocoelicaudia, Borsuk−Białynicka 1977: fig. 12; Rapetosaurus, Curry Rogers 2009: fig. 41A), represents a primitive archosaurian character state that has been maintained throughout archosaurian evolution with oc− casional re−appearance within Sauropoda.
Femur deflected medially and the identity of the "lateral bulge".-The medially deflected condition of titanosaurian femora was first noted by Huene (1929: 42) in his description of "T". australis. The deflected femur, with its associated lat− eral bulge, is a synapomorphy of Titanosauriformes (Salgado et al. 1997;Upchurch 1998;Wilson and Sereno 1998;Wilson 2002). Saltasaurines exhibits an extreme condition of femoral deflection because of the asymmetry of femoral distal con− dyles, which yield a mediolateral inclination (Wilson and Carrano 1999). The great development of the lateral bulge in− creases the deflection of the proximolateral margin of the fe− mur. However, the homology of the "bulge" remains obscure. The lateral bulge was first observed by Huene (1929: 41) and named by McIntosh ("sharp deflection", 1990: 370), who in− terpreted it as a vestige or homolog of the lesser trochanter (see also , Bonnan 2004;Carrano 2005). The lesser trochanter of sauropodomorphs has been characterized as a small spine (Hutchinson 2001b: character 7). On the contrary, according to Wilson (2002: character 197), the lesser trochanter is a primitive character within Sauropoda, only present in Vul− canodon, whereas Titanosauriformes display a well−devel− oped bump or bulge that reaches the greater trochanter. Fusion of the lesser and greater trochanter is present in juvenile neornithines (Hutchinson 2001b); hence, the condition present in Titanosauriformes, in which both trochanters reach one to another, likely represents a convergence with the condition seen in modern birds and their close relatives.
Linea intermuscularis cranialis.-This structure represents an ancestral condition for archosaurs (Hutchinson 2001b). It is absent in non−saltasaurine sauropods, although it is well devel− oped in Neuquensaurus, as well as Saltasaurus (PVL 4017−83, "long rugosity", Powell 2003), Rocasaurus (MPCA−PV 46) and Bonatitan (MACN−RN 821). Although the linea inter− muscularis cranialis present in saltasaurine sauropods is not ac− tually a "line", but a crest, the topological placement is identi− cal to the linea intermuscularis cranialis present in other archo− saurs (Hutchinson 2001b), and also seems to be the structure that forms the boundary between both heads of Mm. femoro− tibiales (Otero and Vizcaíno 2008). I propose to term this struc− ture femorotibialis crest, and regard it as a different character state of the linea intermuscularis cranialis. This structure there− fore represents a derived character within Archosauria and most likely represent a synapomorphy of Saltasaurinae.
Femoral distal condyles.-Primitively, basal sauropods have straight−shafted femora. The "beveled" condition of the titano− saurian femora was noted by Wilson and Carrano (1999). Wil− son (2002: character 201) proposed it as a synapomorphy of Saltasauridae, in which the distal condyles are not perpendicu− lar to the long axis of the femur, but beveled dorsomedially relative to the femoral shaft. Neuquensaurus also displays this asymmetry in the distal femoral condyles, showing an in− crease in the size of the fibular condyle: this result in a particu− larly marked "beveled" condition (Salgado et al. 2005;Otero and Vizcaíno 2008: fig. 5.6). This means that the condyles were held level in life and that the long axis of the femur was therefore inclined dorsomedially rather than being vertical.  , 1903 andGiraffatitan brancai (Janensch, 1914) reported high values too (ECC almost 2.0, Taylor 2009). Although Titano− sauria is characterized by an extremely femoral eccentricity (Wilson and Carrano 1999), several non−titanosaur sauropod also exhibit an even more marked disparity between these two femoral diameters (e.g., Amargasaurus, Appendix 3). Hence, the change from the round femoral section of prosauropods to the elliptical cross−section of most sauropods seems to be gradual along a continuum.
The anteroposteriorly compressed femoral shaft in sauro− pods is a character usually associated with large body size (Wilson and Sereno 1998;Wilson and Carrano 1999;Otero and Vizcaíno 2008). Limb bones of large animals like sauro− pods have to support a mediolateral couple force generated doi:10.4202/app.2009.0099 by their own weight (Wilson and Carrano 1999;Carrano 2001). In the particular case of Neuquensaurus and other de− rived titanosaurs, there is an extra factor of bone bending stress, related to the standing pose which is typically wider during locomotion (Wilson and Carrano 1999;Otero and Vizcaíno 2008). This helps to explain the relationship be− tween the geometry of the titanosaur femoral morphology and the support of mediolateral bending by achieving an ex− treme eccentric cross−section.
Fibular lateral tuberosity.-The fibular lateral tuberosity has been a matter of discussion because its homology is not resolved. This structure is present in Eusauropoda (Wilson and Sereno 1998;Wilson 2002) as an elliptical bump on the lateral side of the proximal fibula, being extremely well de− veloped in Neuquensaurus. Huene (1929) interpreted it as the insertion site of M. peroneus, while Borsuk−Białynicka (1977) and Wilson and Sereno (1998) suggested it to be the origin site for M. flexor digitorum longus. Nevertheless, it seems that the lateral trochanter of the fibula was the inser− tion site for the M. iliofibularis, a flexor muscle that is pres− ent in the same topographical position on the fibular shaft of both living crocodiles (Otero et al. in press) and Neornithes and was recently inferred to be present in Neuquensaurus australis (Otero and Vizcaíno 2008).
Cnemial crest.-The cnemial crest is present and is quite distinctive among sauropods. In Eusauropoda it projects lat− erally as a thin plate (Wilson and Sereno 1998;Wilson 2002). In Neuquensaurus and other saltasaurine sauropods (e.g., Saltasaurus, PVL 4017−84) it is well developed, robust and anterolaterally oriented. An anteriorly projecting cne− mial crest is present in Vulcanodon (Cooper 1984, Wilson 2002 and sauropod outgroups (Cooper 1981;Galton 1990;Wilson 2002). Thus, the state in Neuquensaurus and its clos− est relatives is an intermediate character state for the cnemial crest. Although the functional significance of that structure remains obscure (Wilson and Sereno 1998), the anterolateral projection and the robustness of the saltasaurine cnemial crest could be related to the shift of the insertion site of the leg extensor Mm. femorotibiales, M. ambiens and Mm. ilio− tibiales (Otero and Vizcaíno 2008).

N. australis vs. N. robustus: comparisons
Many of the multiple specimens described by Lydekker (1893) and Huene (1929) seem to pertain to a single individ− ual; others, to a single genus and even species. However, sev− eral elements described by those authors as "Titanosaurus" robustus probably belong to the type species, "Titanosaurus" australis, and vice versa. On the other hand, the status of oth− ers cannot really be determined: they may belong to the type species, or they may represent another species or even genus.
There are several elements originally described and re− ferred to "T". robustus which actually I consider here as be− longing to Neuquensaurus australis. The right sternal plate MLP−CS 1295 was erroneously described by Huene (1929) as "left" and belonging to "T". robustus because the antero− ventral crest is a little bit shorter than in N. australis and the lateral border is more concave (Huene 1929: 36). However, that element has damaged its periosteum and the blade is not complete. On the other hand, the general outline and relative proportions are similar to the sternals referred to N. australis (MLP−CS 1260 and MLP−CS 1104). Hence, the differences between MLP−CS 1295, MLP−CS 1260, and MLP−CS 1104 are not here deemed of taxonomic importance. A similar situ− ation occurs with the right humerus MLP−CS 1019, in which "the proximal end … has an oblique orientation respect the proximal and laterals axis" (Huene 1929: 49, translated from the Spanish). Nonetheless, the aspect of MLP−CS 1019 is al− tered by its broken proximolateral end. Out of this, the pro− portions and the deltopectoral crest, which is well preserved, are identical to those of N. australis. The fragment of left ilium (MLP−Av 2069) close resembles the original fragment of right ilium (MLP−Ly 17) of "T". australis described by Lydekker (1893), despite that the former was referred by Huene (1929) as belonging to "T". robustus.
Likewise, many elements originally referred to "T". aus− tralis by Huene (1929) closely resembles the lectotype of "T". robustus specified by Bonaparte and Gasparini (1978). The fragment of left ulna MLP−CS 2004 was originally referred by Huene (1929) to "T". australis, but differs considerably from the other ulnae assigned to the species. Actually, MLP−CS 2004 has the olecranon process projecting above the proximal end articulation and is as robust as the ulnae included in the lectotype of N. robustus. A similar situation occurs with the radius. The elements MLP−CS 1169, MLP−CS 1172, and MLP−CS 1175, referred by Huene (1929) to "T". australis, show no taxomomic differences with the lectotype of "T". robustus MLP−CS 1171 and have the same proportions.
The condition of the femur, tibia and fibula deserves a spe− cial consideration because there are two real morphotypes of those bones and no type material to compare them. Among the specimen described by Salgado et al. (2005) and its associated elements there are two clearly morphs: one gracile and one ro− bust. The gracile elements were found associated in such a way that they may be assumed with confidence to pertain to a single individual (Salgado et al. 2005). Such elements include the left and right femur (MCS−5/27 and MCS−5/28, respec− tively), and the right tibia (MCS−5/25) and fibula (MCS−5/26). Those elements differ from MCS−9 (right femur) and MCS−6 (right tibia) despite the fact that all of them were found in the same area. The latter bones are more robust than the former al− though they are similar in length. Actually, MCS−9 more re− sembles the lectotype of N. robustus (MLP−CS 1480) and both have similar robustness indices (see Appendix 2D). Likewise, MCS−6 is similar to MLP−CS 1264, which is referred to the latter species. Therefore, I propose here to consider MCS−9 and MCS−6 as belonging to N. robustus. The right fibula de− scribed by Salgado et al. 2005 (MCS−5/26) is slender than MLP−CS 1264, referred by Huene (1929) to "T". robustus and have the lateral trochanter less developed. Its assignation to a different species seems to be correct.
Finally, several elements referred by Huene (1929) to either "T". australis or "T". robustus seems to pertain to a different species or even genus due to their differences and/or their frag− mentary condition. That is the case of the scapulae Saltasaurinae, a well defined South American clade of dwarf (?) sauropods Despite the taxonomic goings and comings of the systematic names, it is clear that there was a well defined group of sauro− pods of small−to−medium size, which inhabited southern South America, and can be differentiated from titanosaurs from the rest of the world by particular features. Such a clade, the Saltasaurinae, is clearly endemic to South America. The four species of saltasaurines: Neuquensaurus australis (Lydekker, 1983), Rocasaurus muniozi Salgado and Azpili− cueta, 2000, Saltasaurus loricatus Powell, 1980, andBonatitan reigi Martinelly andForasiepi, 2004 have been reported in Neuquén, Río Negro, and Salta provinces. It is noteworthy that saltasaurines are absent farther south than 42°S, although this may be due to the presence of the North Patagonian massif as a geographic barrier (Salgado 2000). The saltasaurines are exclusively forms of the Uppermost Cre− taceous (Campanian-Maastrichtian), hence radiation of the group seems to have occurred during a short period of time.
One of the most interesting features of saltasaurines is body size. Saltasaurines have been reported as the smallest sauropods known, only comparable to dicraeosaurine diplo− docoids (Salgado 1999(Salgado , 2000 and the titanosaurid Magyaro− saurus (Jianu and Weishampel 1999). In the case of salta− saurines and Magyarosaurus the mechanism hypothesized as responsible for their small size is heterochrony (Jianu and Weishampel 1999;Salgado 2000), which is defined as the change in the timing of ontogenetic events (McKinney 1986). As McNamara (1982: 130) pointed out, "changes in ontogenetic sequences though time occur by contraction, ex− tension, or a shift in timing of rates of morphological devel− opment". In the specific case of size reduction through suc− cessive ontogenies, the heterochronic process involved is paedomorphosis, which include neoteny (reduction in rate of development), progenesis (precocious sexual maturation re− duces the period of juvenile allometric growth), and post−dis− placement (retardation in onset of growth of particular organs) (McNamara 1982;McKinney 1986). Following McKinney and McNamara (1991) and Salgado (2000) the advantage of the progenesis is that it advances reproductive capability, assuming that this characteristic is positively in− fluenced by natural selection. On the contrary, if the selected character was the small size, both mechanisms neoteny and progenesis, could be equally plausible. In the latter case, saltasaurine sauropods could evolve as a result of predation pressure (or lack thereof). That is, if juveniles and adults of saltasaurines ancestors lived in different environments and if the predation pressures were high in the adults environments, then the selection would favored those forms that tended to delay the time of maturation or stopping the development. This is explained in that, usually, large predators avoid small prey because the energy expenditure does not compensate the earnings (McKinney and McNamara 1991).

Conclusions
The appendicular skeleton of Neuquensaurus displays many derived character states within Sauropoda, such as the out− wardly oriented preacetabular lobe of the ilium, a medially deflected femur, beveled femoral distal condyles, an eccen− tric femoral midshaft, and a well developed fibular lateral tuberosity. Most of these features characterize Saltasauridae (Wilson 2002;Curry Rogers 2005). The presence of a promi− nent olecranon, a trochanteric shelf, a lesser trochanter, and an ischial tuberosity represent reversals to primitive charac− ter states; while the linea intermuscularis cranialis present on the femur of Neuquensaurus (and also in Saltasaurus, Roca− saurus, and Bonatitan) represents novel character within Sauropoda that support, together with axial information, their inclusion into the monophyletic Saltasaurinae.

Appendix 2
Measurements of Neuquensaurus in cm. RI, robustness index, was calculated as follows: RI = average of the greatest widths of the proximal end, mid−shaft and distal end of the element/length of the element (taken from Wilson and Upchurch 2003). ECC, eccentricity index, was calculated as follows: ECC = (femoral mid−shaft width/femoral antero−posterior width) (taken from Carrano 2001