The Skeleton of the Manus of Scelidotherium (Xenarthra, Mylodontidae) Specimens from the Pleistocene of the Province of Córdoba, Argentina, and its Systematic Implications

Abstract

Dental and craniomandibular data have been predominantly used to infer relationships among mylodontid ground sloths. Recent studies indicate the osteology of the manus also provides useful data to test phylogenetic relationships in mylodontine mylodontids. Here we provide new comparative data from the study of the manus of a member of the Scelidotheriinae, Scelidotherium Owen, 1839, based on specimens from the Province of Córdoba that provide information on the variation and systematic relationships of the genus. The Scelidotherium material from the Province of Córdoba in central Argentina presents several morphological differences with material of the genus from northwestern Argentina. The manus of Scelidotherium shares several traits with Catonyx Ameghino, 1891, but not with Valgipes Gervais, 1874. These include: trapezoid with a sharp ridge on the articular surface for metacarpal II; magnum with at least half of articular surface for metacarpal II located on the dorsal half of the medial and distal aspects of the bone; metacarpal II with a notch on the axial border of its distal articular surface; and ungual process of the distal phalanx of the second digit markedly dorsopalmarly flattened. A phylogenetic analysis that includes osteological characters of the manus indicates that within the family Scelidotheriinae, Scelidotherium is more closely related to Catonyx than to Valgipes or Proscelidodon gracillimus Rovereto, 1914.

Introduction

The osteology of the South American sloth genus Scelidotherium Owen, 1839, is well known (e.g., Owen 1839, 1857; Burmeister 1881; Cuenca Anaya 1995), largely on the basis of very complete individuals belonging to the type species Scelidotherium leptocephalum Owen, 1839 (Burmeister 1881; Lydekker 1894; Cuenca Anaya 1995). The only other species undoubtedly assigned to this genus, Scelidotherium parodii Kraglievich, 1923, is known by less material, mostly restricted to the skull and humerus (Kraglievich 1923; McDonald 1987; Miño-Boilini 2012). These specimens have facilitated assessing the phylogenetic interrelationships of the species included in the genus with other scelidotheriines in a handful of published works (McDonald and Perea 2002; Cartelle et al. 2009; Corona 2012; Miño-Boilini 2012; Miño-Boilini et al. 2014). Among these different studies, the results of McDonald and Perea (2002) and Cartelle et al. (2009) are similar, but those of Miño-Boilini (2012) and Miño-Boilini et al. (2014) differ in their conclusions. A difference between these studies is the discrepancy concerning the closest relatives of Scelidotherium (including Scelidotherium parodii). In all of these studies the matrices are primarily based on cranial, mandibular, and dental characters. The data matrix in McDonald and Perea (2002) only presented two parsimony-informative postcranial characters out of a total of 22, and Miño-Boilini (2012) presented four parsimony-informative postcranial characters out of 21. Therefore, one way to resolve the incongruence between these studies may be the inclusion of more postcranial characters. A recent study (Haro et al. 2016) indicated that the skeleton of the manus provides useful characters for phylogenetic studies in mylodontine ground sloths. Due to the sister-group relationship of mylodontines to scelidotheriines (McDonald 1987; Gaudin 2004; Haro et al. 2016), the structure of the manus may also potentially provide parsimony-informative data to unravel the relationship of Scelidotherium to the other scelidotheriines.

The taxonomic usefulness of characters in the bones of the manus in this group is clear: previous systematic decisions, such as referring S. leptocephalum and Catonyx cuvieri (Lund, 1839a), but not Valgipes bucklandi (Lund, 1839b), to the genus Scelidotherium, is based largely on the structure of the manus (Winge 1915). Furthermore, previous studies have proposed characters from the manus as diagnostic features of Scelidotherium or the species S. leptocephalum, including the absence of an ungual in the pollex and the reduction of the contact between the unciform and metacarpal V (Schulthess 1920; McDonald 1987). Despite the existence of several descriptions of the bones of the hand of Scelidotherium (e.g., Burmeister 1881; Schulthess 1920; McDonald 1987; Esteban et al. 1992; Cuenca Anaya 1995), and comparative information of hand structure that may inform phylogenetic analysis (e.g., Winge 1915; Ortega 1967; McDonald 1987), this skeletal region has not yet been fully examined to detect characters useful to determine the relationships of Scelidotherium with other scelidotheriines.

Recognition of variation is also important in dealing with the phylogenetic relationships of a taxon (Boscaini et al. 2019). Comparative studies of morphology of specimens from different regions or ages that belong to the same species are more likely to increase our knowledge of intraspecific variation than comparisons of specimens from the same region or age, because regional and temporal variation is added to random or sexual variation. For example, McDonald (1987) indicated there is size variation between S. leptocephalum specimens of different age. The latter author also considered there is some evidence suggesting geographical variation in body size for Catonyx chiliense (Lydekker, 1886). Another example is a study of C. cuvieri from Uruguay, which increased our knowledge of its variation, which overlaps that of Scelidotherium in certain features (Corona et al. 2013). Most previous descriptions of the manus of Scelidotherium deal with specimens from the Pampean Region of Argentina (e.g., Burmeister 1881; Schulthess 1920; McDonald 1987; Cuenca Anaya 1995), but Esteban et al. (1992) described the manus of S. leptocephalum from the Province of Salta in northwestern Argentina. A problem is that the lack of detailed specimen-based descriptions (with explicit provenances for the specimens) in the previous literature precludes identifying geographical variations even if intraspecific differences are described in mani of Scelidotherium.

Hoping to contribute knowledge useful to resolve these problems related to systematics, here we describe bones from scelidotheriine hands from three different individuals collected from the late Pleistocene of the Province of Córdoba, in central Argentina, intermediate in location between the aforementioned specimens of Scelidotherium. Haro et al. (2016) referred all of these to Scelidotherium, on the basis of the lack of contact between metacarpal V and the unciform. One of these, previously referred to the species S. leptocephalum (Krapovickas et al. 2017), represents a remarkably complete manus. However, considering that some specimens of C. cuvieri also lack contact between metacarpal V and the unciform (Winge 1915), the basis for such referral is insufficient. In this work we test the referral of the specimens from the Province of Córdoba to the genus Scelidotherium, and document the implications of the characters from the manus of that material to the study of variation and phylogenetic interrelationships of that taxon among the Scelidotheriinae.

Materials and Methods

The studied materials belong to three specimens. CORD PZ 4464 is an articulated complete left thoracic member (i.e., forelimb). CORD PZ 4586 includes a partially preserved left manus, including scaphoid, lunar, cuneiform, co-ossified trapezoid and magnum, unciform, co-ossified trapezium and metacarpal I, metacarpals II and IV, proximal phalanx of digit I, ungual phalanx of digit II, and all phalanges of digit III. It is associated with the caudal part of the skull, a third metatarsal, other thoracic limb bones, and a few other unprepared bones. CORD PZ 11293 consists of an ulna, magnum, unciform, and many rib fragments.

Anatomical terminology follows the conventions of the Nomina Anatomica Veterinaria (International Committee on Veterinary Gross Anatomical Nomenclature 2005), in agreement with Harris (2004). We follow the phylogenetic definitions of the Mylodontinae Gill, 1872, and the Scelidotheriinae Ameghino, 1904, proposed in Haro et al. (2016). Here we propose further phylogenetic definitions for Mylodontidae Gill, 1872, and Mylodontoidea Gill, 1872. We define Mylodontidae as the least inclusive node-based clade including Mylodon darwinii Owen, 1839, Scelidotherium leptocephalum Owen, 1839, their common ancestor, and all its descendants. We define Mylodontoidea as the stem-based clade encompassing all taxa more closely related to M. darwinii than to Megatherium americanum Cuvier, 1796, and Megalonyx jeffersonii Desmarest, 1822.

Comparisons were made principally with other members of the Scelidotheriinae. These comparisons based on traits noted in descriptions of the manus in the scelidotheriine taxa listed in Table 1. Further comparisons are made with the specimen PVUNS 198, referred to Proscelidodon Bordas, 1935, by Aramayo (1988) and to Scelidotherium by Esteban et al. (1992). We assembled those comparisons into character-taxon matrices (Supplementary Data S1–S4) containing 177 characters (Supplementary Data S5), of which 24 belong to the skeleton of the manus. A few characters from the manus are continuous, and are treated as such (sensu Goloboff et al. 2006). Because of the error of scaling based on total range of values for a character (Farris 1990), each continuous character is scaled so that the mean weight of all its parsimony-informative transformations equals one (the procedure is explained and justified in Supplementary Data S5). The characters are based on the studies of Winge (1915), Kraglievich (1923), Ortega (1967), McDonald (1987), Scillato-Yané and Carlini (1998), McDonald and Perea (2002), Gaudin (2004), and Miño-Boilini (2012), and also include new ones proposed in this study. Only those characters that are parsimony-informative for the sampled taxa were included. Winge (1915) provided most of the manus traits we have included. Among characters provided by Gaudin (2004), all those being parsimony-informative for the included taxa were included. Incorporation of characters in Gaudin (2004) was used instead of constraints to recover the relative position of the included outgroups (whose interrelationships are not the focus of our work) supported by large morphology-based analyses (i.e., Gaudin 2004; Boscaini et al. 2019).

Table 1 Taxa and material used for comparisons. Collection numbers refer to specimens studied first hand. Bibliographical citations on the right column are sources of descriptions

We performed a phylogenetic analysis including all characters, using the presently described specimens as independent terminals (Analysis 1), in order to test their referral to the genus Scelidotherium. Use of the criterion of parsimony with specimens as independent terminals in phylogenetic analyses informs alpha-taxonomic assignments (Vrana and Wheeler 1992; Yates 2003; Tschopp et al. 2015). On the grounds of parsimony, the hypothesis of referral to the genus Scelidotherium can be rejected if the specimens are not recovered most closely related to species of Scelidotherium than to other species. Reference to the species S. leptocephalum would be only possible if, unlike in previous analyses, the genus turns out to be not monophyletic, because the only other undisputed species of Scelidotherium, S. parodii, is mostly known by the facial region of the skull (Kraglievich 1923), which is lacking in the specimens here reported. We performed a second analysis, in which we combined all the information from S. leptocephalum, S. parodii, and all the specimens more closely related to these than to species of genera different from Scelidotherium (according to the results of the previous analysis), within a single terminal representing the genus Scelidotherium (Analysis 2). This analysis, incorporating the data from the presently described specimens within a taxon, would better inform on the relationships of that taxon. For the sake of comparison with the complete dataset, we performed two other phylogenetic analyses using the same terminals as the previous one: one only including the characters belonging to the manus (Analysis 3) and one including all characters except for those of the manus (Analysis 4).

The data matrix was assembled using Microsoft Excel 2013. We performed the phylogenetic analyses with the program TNT (Goloboff et al. 2003; Goloboff and Catalano 2016). The analyses were based on equal-weights parsimony and searches were exact. No aprioristic constraints were used. We used the ‘Pruned trees’ command in TNT to identify wildcard terminals once the analysis was performed. This identification was used to hide such terminals from the strict consensus, in order to detect resolution hidden by their presence. No analyses were run excluding wildcard terminals. We obtained the strict consensus and the reduced strict consensus trees. We calculated jackknife frequencies (p: 0.36; 10,000 replicates) and Bremer support values on the strict consensus trees. We only used unambiguous optimization for reconstruction of the character state evolution because TNT does not provide ACCTRAN or DELTRAN reconstructions.

The ingroup includes all of the scelidotheriines included in the comparisons indicated above plus S. parodii, Procelidodon rothi (Ameghino, 1908), and Sibyllotherium guenguelianum Scillato-Yané and Carlini, 1998. The present taxon list includes as terminals those specimens here described, as well as the one studied in Aramayo (1988), PVUNS 198, to test its referral to Scelidotherium by Esteban et al. (1992). The choice of the ingroup terminals was made attempting to include all possible scelidotheriine taxa. This implies including the terminals used by previous phylogenetic works (McDonald and Perea 2002; Cartelle et al. 2009; Miño-Boilini 2012) and manual material of uncertain affiliation (Aramayo 1988). We selected five sloth taxa as outgroups based on a previous large phylogenetic study (Gaudin 2004), and did not enforce outgroup topology a priori. These five taxa are M. darwinii, Pseudoprepotherium confusum Hirschfeld, 1985, Hapalops longiceps Scott, 1904, the extant genus Choloepus Illiger, 1811, and the more incompletely known genus Nematherium Ameghino, 1887. Use of the genus Nematherium as outgroup follows its usage in previous phylogenetic studies of scelidotheriines (e.g., McDonald and Perea 2002; Miño-Boilini 2012). The selection of Ps. confusum and H. longiceps as outgroups was based on their currently accepted status as early representatives of mylodontines and megatherians, respectively, for which several bones of the hand skeleton are known and described (Scott 1904; Hirschfeld 1985). The selection of M. darwinii as a further outgroup is based on both the amount of data on the structure of the manus available, and comparatively studied (Haro et al. 2016), and its status as a member of the Mylodontinae, the closest relatives to scelidotheriines (according to Gaudin 2004). The selection of the genus Choloepus as a further outgroup is based on its recognition as a mylodontoid in recent molecular analyses (Delsuc et al. 2019; Presslee et al. 2019). Source data are available in Table 1.

Availability of Data and Material

All data generated or analyzed during this study are included in this published article [and its supplementary information files].

Institutional Abbreviations: CORD PZ, Colección de Paleozoología, Museo de Paleontología, Universidad Nacional de Córdoba, Córdoba, Argentina; DMNS, Denver Museum of Nature and Science, Colorado, U.S.A.; PMU, Paleontologiska Museet Uppsala Universitet, Uppsala, Sweden; PVUNS, Paleontología de Vertebrados, Universidad Nacional del Sur, Bahía Blanca, Argentina; ZMK 1/1845 (previously referred to ZMUC), Lund’s collection, Zoologisk Museum, Statens Naturistoriske Museum, Københavns Universitet, Copenhagen, Denmark.

Systematic Paleontology

Tardigrada Latham and Davies, 1795

Mylodontidae Gill, 1872

Scelidotheriinae Ameghino, 1904

Scelidotherium Owen, 1839

Type Species

Scelidotherium leptocephalum Owen, 1839, from Punta Alta, Bahía Blanca, Province of Buenos Aires, Argentina.

Included Species

The species Scelidotherium leptocephalum, the youngest and largest species in body size, was distinguished from Scelidotherium bravardi Lydekker, 1886, by Lydekker (1886) and more recently by Miño-Boilini et al. (2014), and from Scelidotherium parodii by Kraglievich (1923). McDonald (1987) considered S. bravardi a junior synonym of S. leptocephalum and that the slight difference in size reflects its older age as there is a trend for the species to increase through time.

Diagnosis

The following diagnosis is based on McDonald (1987), with additions from Miño-Boilini et al. (2014). Skull elongate, narrow; nasal passage narrower than in Catonyx Ameghino, 1891, or Proscelidodon, and not inflated. Dorsal profile of braincase flat, lacking concavity of Catonyx or dorsal swellings of Proscelidodon. Occiput sloped posteroventrally with occipital condyles projecting posteriorly. Temporal cristae parallel and not diverging to the supraorbital processes until anterior of frontoparietal suture (usually approximately level with postorbital constriction); no sagittal crest. Dorsal inclination of predental part of maxilla weak, 0 to 10° relative to the tooth-bearing part of the maxilla. Ventral portion of maxilla between toothrows straight, lacking median grooves. Premaxilla with constant dorsoventral thickness (not tapering rostrally as in Catonyx and Proscelidodon), lacking step for contact with the anterior edge of the premaxillary process of the maxilla. Lateral process of nasal extends anteriorly beyond the edge of the maxilla. Paroccipital process mediolaterally flattened and projecting ventrally to the entotympanic. Tympanohyal with posterior expansion between mastoid and paroccipital processes. Teeth small in proportion to size of the skull compared to Catonyx and Proscelidodon. First upper tooth narrower relative to length than in Catonyx. Second to fourth upper teeth forming elongated triangles lacking buccal furrows, corners not forming distinct lobes in occlusal aspect. Mandibular symphysis forming an elongated spout, squared anteriorly and flattened; internal surface shallow; angle of spout to tooth row small, average 17°; symphysis lacking ventral keel. Posterior lobe of lower fourth tooth parallel to anterior lobe and without lingual curve. Digit I of manus reduced and without ungual phalanx. Scaphoid with proximal and distal articular surfaces well separated on the dorsum. Metacarpal IV with articular facets for metatarsal III and unciform that meet at an obtuse angle. Femur relatively short with a strong medial slope of the proximal end relative to the distal end; medial surface of femoral shaft concave.

Scelidotherium sp.

Referred Material

Materials listed in McDonald (1987) and Miño-Boilini et al. (2014), as well as specimens CORD PZ 4586, 4464, and 11293.

Hypodigm

Specimens CORD PZ 4586, 4464, and 11293.

Localities and Horizons

CORD PZ 4464 was collected in Pampa de Oláen (Province of Córdoba), Vaca Corral Formation, late Pleistocene (Lujanian SALMA) (31°9′11.2”S; 64°35′31.3”W; 1138 m.a.s.l.; Krapovickas et al. 2017). CORD PZ 4586 comes from the vicinity of Noetinger, on the Pampean plains, age is uncertain, unattributed loess deposits (Leguizamón et al. 2000). CORD PZ 11293 was collected in Sierra de Las Peñas (32°34′39”S, 64°20′14”W), east of the nearby town of Elena, Province of Córdoba (Fig. 1), La Invernada Formation, late Pleistocene (Lujanian SALMA).

Fig. 1

Maps indicating provenances of the Scelidotherium Owen, 1839, specimens CORD PZ 4464, 4586, and 11293 from the Province of Córdoba, Argentina: a, map of South America (dark grey) indicating placement of the Province of Córdoba (light gray) within Argentina (white); b, close-up on the Province of Córdoba. Black dots in b represent urban centers, pentagons with numbers within represent highway routes, crossed-bone symbol indicates the location of the materials

Description

The following description focuses on characters showing variation among scelidotheriines. Measurements are provided in Table 2 and Supplementary Data S5 (Tables S2 to S5). In CORD PZ 4464 carbonate concretions make it difficult to disarticulate most bones in order to describe the articular surfaces (Fig. 2) or obtain measurements.

Table 2 Phylogenetically useful measurements (in millimeters) of hand bones of scelidotheriine specimens from the Province of Córdoba

Fig. 2

Articulated block containing most bones of the manus (without the proximal carpal row) of the Scelidotherium Owen, 1839, specimen CORD PZ 4464 in dorsal view. Abbreviations: id, intermediate and/or distal phalanx of digit IV; m, magnum; mII–V, metacarpals II–V; p, proximal phalanx of digit V; t, trapezoid; u, unciform. Stippled areas represent sedimentary matrix; hatched areas represent breakage. Scale bar represents 20 mm

Brief Notes on the Non-manual Skeleton

The skull of CORD PZ 4586 presents temporal lines quite separate from the midline, caudal borders of the temporal fossae not parallel to the borders of the occiput, a trough between the paroccipital and mastoid processes obstructed by the tympanohyal, and well-developed fossae for the M. rectus capitis ventralis on the basioccipital. All these features are shared with Scelidotherium, but not Catonyx (McDonald 1987). The humerus of CORD PZ 4464 presents an entepicondylar foramen, as in S. leptocephalum and most other scelidotheriines, but not Catonyx cuvieri (Winge 1915). The radius of CORD PZ 4464 presents a prominent pronator crest of the radius, as in S. leptocephalum, but not C. cuvieri (Winge 1915). The ulna of CORD PZ 4464 and 11293 presents a relatively narrow proximocranial projection of the olecranon and an anconeal process of the ulna that is relatively not prominent, as in S. leptocephalum, but not C. cuvieri (Winge 1915). The ulna in both specimens presents a relatively blunt coronoid process, as in S. leptocephalum, but not Catonyx tarijensis Gervais and Ameghino, 1880, (Miño-Boilini 2016).

Carpus

Scaphoid

CORD PZ 4586 presents a clearly distinct non-articular interval on the dorsal aspect that separates the proximal from distal articular surfaces. This is similar to some other Scelidotherium and Valgipes Gervais, 1874, specimens in ZMK 1/1845 (Winge 1915), but dissimilar to C. chiliense (PMU M4394; Sefve 1915) or the C. cuvieri material in ZMK 1/1845 (Winge 1915) in which the articular surfaces are much closer to one another or in contact. The medial part of the proximal articular surface presents a dorsopalmarly oriented concavity. The palmar border of the articular surface for the radius is just a little less distally extended on the palmar surface than the dorsal border is on the dorsal surface. In both CORD PZ 4464 and 4586 the scaphoid contacts the lunar by two surfaces completely separated by a recessed non-articular surface. This surface is proximodistally wide, and the facets are proximodistally narrow and dorsopalmarly elongate. The laterally facing proximal articular facet for the lunar, which contacts the articular surface for the radius at an edge, is slightly convex dorsopalmarly and straight proximodistally, and more closely resembles the condition of C. cuvieri than that of V. bucklandi described by Winge (1915). In both CORD PZ 4464 and 4586 the distal articular surface for the lunar contacts the articular surface for the magnum. In CORD PZ 4586 the distal articular surface for the lunar is flat, but in CORD PZ 4464 it is dorsopalmarly sinuous. The dorsopalmar depth of the bone at its lateral side, relative to its mediolateral breadth, is much higher than in the specimens of S. leptocephalum, C. cuvieri, and V. bucklandi measured by Winge (1915), although it more nearly approaches the specimens of the first species (Table 2). In CORD PZ 4464, in distal view, the long axis of the trapezial process is oblique relative to the lateral border of the articular facet for the magnum. The articular surface for the trapezium-metacarpal I is not clearly preserved. The subtriangular trapezoid articular facet presents a well-developed palmar angle, which in CORD PZ 4586 reaches the palmar half of the distal surface. In both CORD PZ 4464 and 4586, the articular facet for the magnum is single, as in C. chiliense and C. cuvieri (Sefve 1915; Winge 1915). This contrasts with the subdivided condition found in V. bucklandi (Winge 1915). In both CORD PZ 4464 and 4586, the articular surface is deeply concave and elongate in the dorsopalmar direction. It narrows mediolaterally at mid-length due to the presence of a notch in the medial border. This gives the facet a bilobate shape, with the palmar lobe mediolaterally wider. In CORD PZ 4586 (Fig. 3a), the lateral border of the articular facet for the magnum is straight. By contrast, in CORD PZ 4464 (Fig. 3b) that magnum facet border is sinuous and has a dorsal concavity and palmar convexity.

Fig. 3

Scaphoid and lunar of the Scelidotherium Owen, 1839, specimens CORD PZ 4586 and 4464. a, distal view of scaphoid in CORD PZ 4586; b, distal view of scaphoid in CORD PZ 4464; c, proximomedial view of lunar in CORD PZ 4586. Abbreviations: m, articular surface for magnum; r, articular surface for radius; t, articular surface for trapezoid; s, articular surface for scaphoid; tp, trapezial process. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; hatched areas represent breakage. Scale bar equals 10 mm

Lunar

In both CORD PZ 4464 and 4586 the articular surfaces for the radius and cuneiform do not meet, unlike in Proscelidodon gracillimus Rovereto, 1914 (Ortega 1967). In CORD PZ 4586, the articular surface for the cuneiform forms an acute angle with the radial articular surface, in contrast to the straight angle present in Pr. gracillimus (Ortega 1967). Thus, the articular facet faces laterodistally. In CORD PZ 4586 the scaphoid articular facets do not occupy the entire dorsal border in medial aspect. This contrasts with the more extensive facet in Pr. gracillimus (McDonald 1987). In CORD PZ 4586, the proximal and distal articular surfaces for the scaphoid are continuous with the articular surfaces for the radius and the magnum, respectively. In CORD PZ 4586, the proximal part of the articular surface for the scaphoid is dorsopalmarly slightly concave and proximodistally roughly straight. In CORD PZ 4586, the dorsal part of the articular surface for the magnum presents a dorsopalmarly directed convexity (Fig. 3c), as in the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992), and C. cuvieri, but in contrast with the thoroughly concave condition in V. bucklandi (Winge 1915). This convexity relates to the presence of a prominent oblique ridge within the surface (Winge 1915), which is lacking in Pr. gracillimus (Ortega 1967). The ridge is closer to the dorsal border than to the palmar one, as in C. cuvieri (Winge 1915), in contrast to the ridge closer to the palmar border reported in unspecified ‘Scelidotherium’ materials by Ortega (1967). In CORD PZ 4586, the dorsal border of the articular surface for the magnum is wider than the palmar one, as in the scelidotheriine material studied by Cuenca Anaya (1995) and in Pr. gracillimus (Ortega 1967), but in contrast to the condition of unspecified ‘Scelidotherium’ materials reported by Ortega (1967). In CORD PZ 4586, the articular surface for the unciform is dorsopalmarly extensive, reaching the dorsal border of the distal surface. This contrasts with the palmarly restricted condition of V. bucklandi (Winge 1915).

Cuneiform

In CORD PZ 4586, the cuneiform’s lateral side is proximodistally higher relative to the ulnar facet width (ratio: 0.98; Supplementary Data S5, Table S2). The ratio is included within the range of proportions of C. cuvieri (range of ratio: 0.83–1) but not in the range of other specimens of Scelidotherium (range of ratio: 0.78–0.83), and very far from those of V. bucklandi (range of ratio: 0.5–0.63), according to measurements in Winge (1915). In CORD PZ 4464 and 4586, the ulnar articular surface is subtriangular (Fig. 4a, b), unlike the subcircular facet found in the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). The ulnar facet is more elongate in CORD PZ 4586 than in CORD PZ 4464. In CORD PZ 4464, the proximal articular surface extends a little into the medial side of the bone (Fig. 4c), as in the scelidotheriine material described by Cuenca Anaya (1995) and probably Esteban et al. (1992), if their ‘articular surface for the radius’ corresponds to this extension. On the other hand, in CORD PZ 4586 such an extension is absent (Fig. 4d). In CORD PZ 4586, the articular surface for the ulna is nearly flat, as in C. cuvieri (Lund 1842) and the Scelidotherium specimen from Rosario de La Frontera (Esteban et al. 1992) and PVUNS 198 (Aramayo 1988). On the other hand, in CORD PZ 4464 the surface is slightly convex along all axes, even without considering its medial extension. In both CORD PZ 4586 and CORD PZ 4464 the articular surface for the pisiform is largely flat, except for its proximodistally convex proximal region. In CORD PZ 4586, the articular surface for the pisiform is continuous with that for the ulna. By contrast, in CORD PZ 4464 it is partially separated by a rough surface. Both articular surfaces may be connected or not in C. cuvieri (Winge 1915). In CORD PZ 4464, the articular surface for the pisiform is elliptical in outline (Fig. 4e), unlike the subcircular facet of the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). In CORD PZ 4586, the outline of the facet has a subcircular main part (Fig. 4f) and a smaller, parallelogram-shaped part proximally that connects the subcircular part of the pisiform facet with the ulnar facet. CORD PZ 4586 bears a curved ridge medial to the pisiform facet whose position, orientation, and curvature resemble part of the pisiform facet border in CORD PZ 4464 (Fig. 4e, f). In CORD PZ 4586, this ridge, together with the border of the pisiform facet, limits an area very similar in shape to that occupied by the pisiform facet in CORD PZ 4464. The articular surface for the lunar contacts the articular surface for the unciform. In CORD PZ 4464 and 4586, the medial part of the articular surface for the unciform is saddle-shaped, with a dorsopalmar convexity and a mediolateral concavity, unlike the concave facet of V. bucklandi (Winge 1915). In CORD PZ 4464, the lateral part is slightly mediolaterally concave, but in CORD PZ 4586 it is slightly mediolaterally convex.

Fig. 4

Cuneiform of the Scelidotherium Owen, 1839, specimens CORD PZ 4464 and 4586. a, proximal view of bone in CORD PZ 4464; b, proximal view of bone in CORD PZ 4586; c, proximomedial view of bone in CORD PZ 4464;, proximomedial view of bone in CORD PZ 4586; e, palmar view of bone in CORD PZ 4464; f, palmar view of bone in CORD PZ 4586. Abbreviations: l, articular surface for lunar; m, medially facing part of proximal surface; p, articular surface for pisiform; r, ridge probably related to pisiform facet ontogeny; u, articular surface for ulna; un, articular surface for unciform. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; stippled areas represent sedimentary matrix; hatched areas represent breakage. Scale bar equals 10 mm

Pisiform

It is walnut-shaped. The relatively large articular surface for the cuneiform is adjacent to but separated by an obtuse edge from the relatively small articular surface for the ulna. The cuneiform facet is concave on the direction of the lesser axis, but mostly straight along the direction of the major axis. A concavity is also present in this facet in other scelidotheriines (Winge 1915; McDonald 1987; Cuenca Anaya 1995).

Trapezoid

In CORD PZ 4586, the trapezoid is fused with the magnum (Fig. 5a–f), as in the C. cuvieri specimen ZMK 1/1845:831 (Lund 1842; Winge 1915). Fusion is not clear for CORD PZ 4464. In both CORD PZ 4464 and CORD PZ 4586 the trapezoid is proximodistally flattened (Fig. 5a–e) and approximately triangular in proximal/distal views. Both CORD PZ 4464 and 4586 have a palmar process, which enhances the triangular shape of these bones in proximal view, as in most specimens of Scelidotherium (McDonald 1987), and some of C. chiliense (McDonald 1987). This contrasts with its absence in most, but not all, specimens of C. tarijensis and C. cuvieri, and at least one of V. bucklandi (Winge 1915; McDonald 1987). In both CORD PZ 4464 and 4586 the trapezoid is dorsopalmarly deeper than mediolaterally wide (ratios are 1.04 and 1.1, respectively; Table 2), unlike scelidotheriine specimens lacking a palmar process (McDonald 1987). They are relatively slightly shallower than in an unspecified Scelidotherium specimen in ZMK 1/1845 (ratio: 1.13), but slightly relatively deeper than in C. cuvieri (range of ratio: 0.89–0.95), and significantly deeper than in V. bucklandi (ratio: 0.73), according to measurements in Winge (1915). In CORD PZ 4586, the dorsal surface of the trapezoid presents an obtuse angle between its medial and distomedial borders (Fig. 5a), as in the scelidotheriine material described by Cuenca Anaya (1995), but in CORD PZ 4464 the borders form an acute angle. In CORD PZ 4586, the trapezoid is quite low at the medial border. The ratio between its maximum height at the medial border and its maximum mediolateral width (0.3; Table 2) more closely approaches the one of V. bucklandi (0.44), than that of C. cuvieri (range: 0.5–0.61), or Scelidotherium (0.76), according to the measurements in Winge (1915). In CORD PZ 4586, the trapezoid presents a large articular surface for the trapezium-metacarpal I medially. That articular surface is dorsopalmarly elongate, and occupies the central part of the medial surface. It is strongly concave dorsopalmarly, but not proximodistally. In CORD PZ 4586, the articular surface for the metacarpal II is convex dorsally and concave palmarly. It presents a clear, prominent, blunt, and oblique edge on its dorsal part, which separates regions of the facet with different orientations (Fig. 5b). This relatively sharp inflection is shared with the Montehermosan (Miocene–Pliocene) specimen PVUNS 198 (Aramayo 1988), and the specimens of C. cuvieri described by Winge (1915); it is unlike the even curvature of this articular facet in V. bucklandi (Winge 1915).

Fig. 5

Trapezoid and magnum of the Scelidotherium Owen, 1839, specimens CORD PZ 4586 and 11293. a, dorsal view of co-ossified bones in CORD PZ 4586; b, distal view of co-ossified bones in CORD PZ 4586; c, proximomedial view of co-ossified bones in CORD PZ 4586; d, proximal view of co-ossified bones in CORD PZ 4586; e, palmar and slightly distal view of co-ossified bones in CORD PZ 4586; f, palmodistal view of co-ossified bones in CORD PZ 4586; g, proximomedial view of magnum in CORD PZ 11293; h, lateral view of magnum in CORD PZ 4586; i, lateral view of magnum in CORD PZ 11293; j, proximal view of magnum in CORD PZ 11293; k, distal view of magnum in CORD PZ 11293; l, palmodistal view of magnum in CORD PZ 11293, arrow indicates mediodistal corner of trapezoid dorsal surface. Abbreviations: l, articular surface for lunar; mII, articular surface for metacarpal II; mIII, articular surface for metacarpal III on magnum; r, ridge on articular surface for metacarpal II on trapezoid; s, articular surface for scaphoid; t, articular surface for trapezoid; u, articular surface for unciform. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; hatched areas represent breakage. Scale bar equals 10 mm

Magnum

In CORD PZ 11293, this bone is clearly free from fusion to the trapezoid (Fig. 5g), as in PVUNS 198 (Aramayo 1988). In CORD PZ 4586 and 11293, the articular surface for the scaphoid is not subdivided (Fig. 5c, g), as in C. chiliense (Sefve 1915) and C. cuvieri, but not V. bucklandi (Winge 1915). In CORD PZ 4586 and 11293, the scaphoid articular facet resembles a sinuous band in shape (Fig. 5c, g), unlike the crescentic shape of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586 and 11293, the dorsal end of the articular surface does not taper, but does in the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586, the dorsal portion of the articular surface for the scaphoid is nearly as wide as the palmar one, whereas in CORD PZ 11293 the scaphoid facet narrows dorsally (Fig. 5c, g). In both CORD PZ 4586 and 11293, this facet is mostly dorsopalmarly convex, but its dorsal portion is dorsopalmarly concave. Such a dorsal concave part most closely agrees with the description of Ps. confusum by Hirschfeld (1985) than with previous descriptions of the condition as convex in other late Pleistocene scelidotheriines (e.g., C. chiliense, C. cuvieri, S. leptocephalum, V. bucklandi, Sefve 1915; Winge 1915; Esteban et al. 1992). In both CORD PZ 4586 and 11293, the palmar portion of the scaphoid facet is dorsopalmarly convex, as in the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992), unlike the flat facet of V. bucklandi (Winge 1915). In CORD PZ 11293 the ridge between the articular surfaces for scaphoid and lunar is sharper than in CORD PZ 4586. In both CORD PZ 4586 and 11293, the dorsal portion of the articular surface for the lunar is dorsopalmarly concave (Fig. 5h, i), but convex in V. bucklandi (Winge 1915). CORD PZ 11293 presents a single articular facet for the trapezoid, which is undivided by edges or non-articular tracts (Fig. 5g). This contrasts with the presence of two distinguishable facets in V. bucklandi (Winge 1915) and C. chiliense (Sefve 1915). The articular facet for the unciform is dorsopalmarly elongated and proximodistally narrow (Fig. 5h, i). In CORD PZ 11293, the facet is strongly dorsopalmarly concave, except at its dorsal end (Fig. 5j), as in the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). In CORD PZ 4586, the dorsopalmar concavity is nearly absent (Fig. 5d). In both specimens, the dorsal end of the facet is dorsopalmarly convex. In CORD PZ 11293, the unciform facet does not contact the articular surface for the metacarpal III (Fig. 5i), unlike in C. chiliense (Sefve 1915). In both CORD PZ 4586 and 11293, the articular surface for the metacarpal II is well developed on the dorsal portion of the bone distal surface (Fig. 5b, k), as in C. chiliense (Sefve 1915), but not V. bucklandi (Winge 1915). In CORD PZ 11293, this facet is dorsopalmarly longer than mediolaterally wide, and extends along the dorsal two-thirds of the bone distal aspect (Fig. 5k). This facet is dorsopalmarly convex on its dorsal portion but dorsopalmarly concave at the palmar one. In both CORD PZ 4586 and 11293, the articular surfaces for metacarpals II and III contact dorsally (Fig. 5 k). In both CORD PZ 4586 and 11293, they do not contact at the middle (Fig. 5b, k), as in C. cuvieri, but not V. bucklandi (Winge 1915). In CORD PZ 4586, there is no contact palmar to that point (Fig. 5b), as in a specimen of C. cuvieri (Winge 1915). In both CORD PZ 4586 and 11293, the articular surface for metacarpal III is dorsopalmarly longer than mediolaterally wide (Fig. 5b, k), and reaches the palmar end of the bone. In CORD PZ 4586 and 11293, the articular surface for metacarpal III is not remarkably narrow dorsally (Fig. 5b, k), unlike in V. bucklandi (Winge 1915). The dorsal portion of the metacarpal III articular surface is dorsopalmarly convex, and the palmar one is dorsopalmarly concave. In CObut contrasts with the shortened,RD PZ 4586, this dorsopalmar concavity is much deeper than in CORD PZ 11293 (Fig. 5h, i). In this, therefore, CORD PZ 4586 more closely resembles the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). In both CORD PZ 4586 and 11293, the palmar portion of the articular surface for the metacarpal III is mediolaterally convex, but more so in the latter (Fig. 5e–l). In CORD PZ 4586, the dorsal portion of the metacarpal III articular surface presents a marked, nearly transversely oriented edge that divides the medial part of the facet into two areas; in CORD PZ 11293 it is fainter (Fig. 5b, k). In CORD PZ 4586, the portion of the metacarpal III articular surface dorsal to this edge is slightly mediolaterally concave, but in CORD PZ 11293 it is slightly mediolaterally convex (Fig. 5f, l).

Unciform

The proximal aspect of the bone presents a prominent bump at its mediopalmar region, and forms part of the articular facets for the lunar and cuneiform. These facets are convex at the bump (both mediolaterally and dorsopalmarly). In both CORD PZ 4586 and 11293, the articular surface for the lunar exposes on the proximal aspect (Fig. 6a, b), unlike in the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4464, 4586, and 11293, the lunar articular facet is dorsopalmarly elongated, reaches the dorsal border of the proximal surface (Fig. 6a, b), and separates the articular facets for the magnum and cuneiform. That dorsal extension is shared with the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992) and C. cuvieri (Winge 1915), but contrasts with the shortened, palmarly restricted facet of V. bucklandi (Winge 1915). In CORD PZ 4586 and 11293, the facet is roughly a skewed parallelogram in shape, unlike the subrectangular shape of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586 and 11293, it does not extend on the medial surface to encroach the articular facet for the magnum, unlike in some specimens of C. cuvieri (Winge 1915). In CORD PZ 4586 and 11293, the articular surface for the cuneiform is irregularly reniform in proximal view, unlike the subrectangular shape of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586 and 11293, the surface is approximately saddle-shaped. In CORD PZ 11293, the lateral portion of the cuneiform articular surface is dorsopalmarly slightly concave (Fig. 6c), as in the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586, it is less concave (Fig. 6d), and in CORD PZ 4464, it is slightly convex. In CORD PZ 4464, 4586, and 11293, the cuneiform articular surface is dorsopalmarly concave at its mediodorsal region and dorsopalmarly convex at its mediopalmar region, unlike the flat condition of the medial portion of the facet of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 11293, on the medial aspect, the articular facet for the magnum is dorsopalmarly elongate, and irregular in shape, unlike the subtriangular shape of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In both CORD PZ 4586 and 11293, the metacarpal III articular facet faces mostly distally. In both specimens it is saddle-shaped, dorsopalmarly convex and a mediolaterally concave, unlike the flat or pit-like relief of C. cuvieri or the concave relief of V. bucklandi (Winge 1915). In both CORD PZ 4586 and 11293, the shape of the articular surface for the metacarpal III is approximately quadrilateral, with rounded medial angles (Fig. 6e, f), instead of subcircular as in the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In both CORD PZ 4586 and 11293, the metacarpal IV articular surface faces mostly distally and is roughly saddle-shaped in its dorsal portion, specifically dorsopalmarly convex and mediolaterally concave. In CORD PZ 11293, the middle and palmar portions are convex to flat. The facet is not undulating, unlike in some specimens of C. cuvieri (Winge 1915). In CORD PZ 11293, the outline of the articular facet for metacarpal IV is dorsopalmarly elongated, unlike the subcircular shape of the S. leptocephalum specimen from Rosario de la Frontera (Esteban et al. 1992). In CORD PZ 4586 and 11293, the articular surface for metacarpal IV presents a distinct dorsal border, continuing the lateral and medial borders at clear angles. In CORD PZ 4464, 4586, and 11293, an articular facet for metacarpal V is lacking, as in most other described specimens of Scelidotherium and a single C. cuvieri specimen (Winge 1915; McDonald 1987; Esteban et al. 1992). This contrasts with the presence of the facet in V. bucklandi, two Scelidotherium specimens, and most C. cuvieri specimens (Winge 1915; McDonald 1987).

Fig. 6

Unciform of the Scelidotherium Owen, 1839, specimens 4586 and 11293. a, proximal view of bone in CORD PZ 11293; b, proximal view of bone in CORD PZ 4586; c, lateral view of bone in CORD PZ 11293; d, lateral view of bone in CORD PZ 4586; e, distal view of bone in CORD PZ 11293; f, distal view of bone in CORD PZ 4586. Abbreviations: c, articular surface for cuneiform; l, articular surface for lunar; mIII, articular surface for metacarpal III; mIV, articular surface for metacarpal IV. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; hatched areas represent breakage. Scale bar equals 10 mm

Trapezium-Metacarpal I

In CORD PZ 4464 and 4586, the trapezium and metacarpal I are co-ossified and their contact is indicated by a rugose, raised ridge on the palmar surface of the bone. In CORD PZ 4586, the ridge continues on the dorsal surface (Fig. 7a), but in CORD PZ 4464 it does not (Fig. 7b). In both CORD PZ 4464 and 4586, the element is axioabaxially wider proximally. In CORD PZ 4464, abaxially to the articular facet for the trapezoid, the proximal surface of the bone is saddle-shaped (axioabaxially concave and a dorsopalmarly convex) (Fig. 7c). In CORD PZ 4586, the surface is similar, but just concave at its axial portion (Fig. 7d). In CORD PZ 4464, the surface bears a smooth articular facet at its abaxial portion for the scaphoid, and the rest of the surface is rugose and badly preserved (Fig. 7c). In CORD PZ 4586, the aforementioned smooth articular facet is not present, and the rugosity is more strongly marked (Fig. 7d). In CORD PZ 4586, the articular surface for the trapezoid is a little smaller than the articular surface for the metacarpal II. In CORD PZ 4586, the articular surface for the trapezoid is proximal and adjacent to the articular surface for metacarpal II; both facets met at an obtuse angle. In both CORD PZ 4464 and 4586, the axial knob on the proximal end of the bone is prominent, as in C. cuvieri but not V. bucklandi (Winge 1915). In CORD PZ 4586, distal articular surface does not reach the abaxial side of the distal end, as in most specimens of Scelidotherium, but not the genus Catonyx (McDonald 1987). In CORD PZ 4586, the distal articular surface presents a keel on its palmar part. In CORD PZ 4586, a conspicuous articular facet is present on the dorsoaxial surface of the distal end (Fig. 7a), but in CORD PZ 4464 a facet is absent in this position (Fig. 7b).

Fig. 7

Co-ossified trapezium-metacarpal I of the Scelidotherium Owen, 1839, specimens CORD PZ 4464 and 4586. a, dorsal view of bone in CORD PZ 4586; b, distodorsoabaxial view of bone in CORD PZ 4464; c, proximal view of bone in CORD PZ 4464; d, proximoabaxial view of bone in CORD PZ 4586. Abbreviations: p, proximal surface; r, ridge probably representing suture between trapezium and metacarpal I; s, articular surface for possible sesamoid. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; hatched areas represent breakage. Scale bar equals 10 mm

Metacarpals

Only metacarpals II–V are described in this section. Metacarpal I was described above with the carpal bones because it is co-ossified with the trapezium.

Metacarpal II

In CORD PZ 4586, the bone is relatively short compared with its axioabaxial width at the proximal end, yielding a ratio of 0.67 (Table 2), which is larger than those previously reported for other scelidotheriines. Among these, it more nearly approaches C. cuvieri (range of ratio: 0.56–0.64); then C. chiliense (range of ratio: 0.57–0.59); then Scelidotherium (range of ratio: 0.56–0.57), and finally V. bucklandi (range of ratio: 0.52–0.54), according to published measurements (Sefve 1915; Winge 1915; Pujos 2000). In CORD PZ 4586, the proximal aspect of the bone is approximately triangular and bears the articular surfaces for the trapezoid and magnum. The articular surface for the trapezoid occupied most of the proximal aspect, and reaches both its dorsal and palmar borders. The facet is dorsally concave and palmarly convex. In CORD PZ 4586, the articular facet for the magnum is an axioabaxially narrow band axial and adjacent to the articular surface for the trapezoid. In CORD PZ 4586, it extends into the dorsal part of the proximal surface and thus precludes the facet for the trapezoid to reach the axial border of the proximal surface, unlike in V. bucklandi (Winge 1915). In CORD PZ 4586, the articular surface for the magnum is dorsopalmarly excavated as an acute angle, but axioabaxially nearly straight. In CORD PZ 4586, the articular surface for the trapezium-metacarpal I is slightly concave dorsopalmarly, and slightly convex proximodistally. In CORD PZ 4586, the facet is dorsopalmarly elongate, dorsally expanded and palmarly narrow, whereas it is subrectangular in the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). In CORD PZ 4586, the axial aspect presents a proximodistally extended and dorsopalmarly narrow flange, whose tip points distally (Fig. 8a). This process is adjacent to part of the badly preserved articular surface for metacarpal III. The portion just distal and dorsal to this process is proximodistally straight or slightly convex, and dorsopalmarly concave. In CORD PZ 4586, there is no evidence that a sulcus divided the articular surface for the metacarpal III, unlike in the S. leptocephalum specimen from Rosario de La Frontera (Esteban et al. 1992). The distal articular surface is dorsopalmarly elongated. The midline carina on the distal articular surface is thicker dorsally and thinner palmarly (Fig. 8b), as in C. cuvieri, but not V. bucklandi (Winge 1915). The carina is palmarly sharper. In CORD PZ 4586, the border of the midline carina is nearly straight in axial/abaxial views, unlike the convex profile of several specimens of V. bucklandi (Winge 1915). In CORD PZ 4586, the carina is slightly bowed axialwards (Fig. 8b). In CORD PZ 4586, the axial condyle presents a notch at its axial border (Fig. 8b), as in C. cuvieri, but not V. bucklandi (Winge 1915). In both CORD PZ 4464 and 4586, the proximal part of the axial border of the dorsal surface is distinctively prominent dorsally but does not overlap the abaxial margin of metacarpal III, unlike in some, but not all, Catonyx specimens (Winge 1915; McDonald 1987).

Fig. 8

Metacarpals II and IV of the Scelidotherium Owen, 1839, specimens CORD PZ 4464 and 4586. a, dorsoaxial view of metacarpal II in CORD PZ 4586; b, distal view of metacarpal II in CORD PZ 4586; c, proximoabaxial view of metacarpal IV in CORD PZ 4586; d, palmar view of distal end of metacarpal IV in CORD PZ 4586; e, palmar view of metacarpal IV in CORD PZ 4464. Abbreviations: c, carina; cs, bulge probably representing co-ossified sesamoid; f, projecting flange on axial surface of bone; mV, articular surface for metacarpal V; n, notch in the border of the axial condyle of the distal end; r, rugose area; s, articular surface for sesamoid. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; stippled areas represent sedimentary matrix; hatched areas represent breakage. Scale bar equals 20 mm

Metacarpal III

In CORD PZ 4464, the bone narrows dorsopalmarly at midlength. The space separating the articular surfaces for magnum and unciform does not expose on the dorsal surface of the metacarpal when articulated with these two carpals (Fig. 2), as in other scelidotheriine specimens referred to the genera Scelidotherium and Catonyx (pers. obs.), but unlike V. bucklandi (pers. obs.). The axial contours of the bone in dorsal/palmar views are nearly parallel to the long axis of the bone, suggesting an axially facing facet for metacarpal IV. A notch is present on the abaxial border of the distal articular surface.

Metacarpal IV

In both CORD PZ 4464 and 4586, the proximal and distal ends are dorsopalmarly deeper than the shaft (Fig. 8c). The proximally facing articular surface for the unciform is dorsopalmarly sinuous: concave at its dorsal part and slightly convex at its palmar part. In CORD PZ 4586, the articular facets for metacarpal III and unciform form an obtuse angle, as in Scelidotherium but not Catonyx (McDonald 1987). In CORD PZ 4586, the articular facet for the metacarpal V is not in contact with the facet for the unciform, unlike in Pr. gracillimus, V. bucklandi, and most, but not all, C. cuvieri specimens (Winge 1915; Ortega 1967). In CORD PZ 4586, the articular facet for metacarpal III is dorsopalmarly sinuous, slightly concave dorsally and slightly convex palmarly, and proximodistally slightly convex dorsally, but roughly straight palmarly, unlike the simply concave facet of V. bucklandi (Winge 1915). In CORD PZ 4586, on the proximal end of the abaxial surface, the articular surface for metacarpal V is dorsopalmarly concave and proximodistally sinuous, proximally convex and distally concave. In both CORD PZ 4464 and 4586, the shaft is axioabaxially wider than dorsopalmarly deep, as in other Scelidotherium specimens (McDonald 1987). In CORD PZ 4586, the distal articular keel is clearly marked and relatively less expanded dorsopalmarly than in Pr. gracillimus (Ortega 1967). Only the axial condyle is present on the distal articular surface, and the border of the keel midline protrudes axiodistally. In CORD PZ 4586, two articular facets for the sesamoid bones are present palmar to the distal articular surface (Fig. 8d). In CORD PZ 4464, articular facets for sesamoids are not visible, but a bulge is present in the place corresponding to the axial sesamoid (Fig. 8e).

Metacarpal V

In CORD PZ 4464, the proximal surface does not have an articular surface for the unciform, as has been previously described in other individuals of the genus (Schulthess 1920; McDonald 1987). The palmar surface presents three prominent oblique crests.

Phalanges

Proximal Phalanx of Digit I

In both specimens CORD PZ 4464 and 4586, this phalanx is present, unlike in some Scelidotherium specimens (Burmeister 1881). In CORD PZ 4464 and 4568, the phalanx presents a sulcus on its proximal surface for the carina of the trapezium-metacarpal I. The sulcus in the latter specimen is very extensive dorsopalmarly, reaching the dorsal half of the facet. In both specimens, on the palmar aspect, the sulcus corresponds to an indentation of the proximal border, unlike in most described specimens of C. cuvieri (Winge 1915). In both CORD PZ 4464 and 4586, the distal end is badly preserved, and the presence or absence of a distal articular facet cannot be determined. In CORD PZ 4586, the palmar part of the distal end presents a raised surface that has a low sagittally oriented ridge in the middle. In both CORD PZ 4464 and CORD PZ 4586, the bone lacks any evidence of being the result of fusion of two originally distinct phalanges.

Proximal Phalanx of Digit II

This phalanx, recovered only in CORD PZ 4464, is proximodistally flattened, and in dorsal view presents a centered notch that marks the dorsal limit of the groove for the articular carina of the metacarpal.

Intermediate Phalanx of Digit II

This phalanx, recovered only in CORD PZ 4464, presents a width among the distal articular condyles that is slightly lesser than that of the proximal end.

Distal Phalanx of Digit II

In both CORD PZ 4464 and 4586, the phalanx is about as large as the distal phalanx of the digit III of the manus, unlike the relatively smaller phalanx in C. cuvieri and V. bucklandi (Winge 1915). In CORD PZ 4586, the proximal articular surface presents two concave surfaces for the distal condyles of the intermediate phalanx separated by a wide sagittal ridge. In CORD PZ 4586, the ungual process is dorsopalmarly depressed (Fig. 9a, b), as in C. cuvieri, but unlike the axioabaxially compressed phalanx of V. bucklandi (Winge 1915). Prominent acute longitudinal edges are present on the axial and abaxial sides of the ungual process. In CORD PZ 4586, the process presents an axial side flatter than the abaxial, as in C. cuvieri, but unlike the more symmetrical flattening of the sides in V. bucklandi (Winge 1915).

Fig. 9

Distal phalanx of digit II of the manus of the Scelidotherium Owen, 1839, specimen CORD PZ 4586. a, abaxial view; b, palmar view. Abbreviations: u, ungual process. Each interpretative drawing is just below the corresponding photograph. On interpretative drawings, grey areas represent articular surfaces; hatched areas represent breakage. Scale bar equals 20 mm

Proximal Phalanx of Digit III

In both CORD PZ 4464 and 4586, this bone is preserved articulated to phalanx III-2, with sediment interposed in the peripheral parts of their contact. In CORD PZ 4586, the abaxial palmar process of phalanx III-1 is preserved relatively closer to that of phalanx III-2 than in CORD PZ 4464. In both specimens, as much as exposed, phalanx III-1 is proximodistally shorter than axioabaxially wide or dorsopalmarly deep. In CORD PZ 4586, the groove on the proximal articular surface for the distal carina of the metacarpal clearly narrows palmarwards, unlike in V. bucklandi (Winge 1915). Axially to the sulcus for the carina on the proximal articular surface, the articulation is narrow, reduced, and closer to the palmar border than to the dorsal, as in C. cuvieri, but unlike the sizable facet of V. bucklandi (Winge 1915). In CORD PZ 4464 and CORD PZ 4586, an articular surface for a sesamoid is not clear, unlike the clear facet present in V. bucklandi (Winge 1915).

Intermediate Phalanx of Digit III

In both CORD PZ 4464 and CORD PZ 4586, the distal articular surface is a trochlea. In both CORD PZ 4464 and CORD PZ 4586, the distal articular surface faces partially axially (more so in the former), as in C. cuvieri, but not V. bucklandi (Winge 1915).

Distal Phalanx of Digit III

In both CORD PZ 4464 and 4586, the proximal end narrows less in dorsal view than in the distal phalanx of digit II. The distal end of the phalanx is clearly pointed in CORD PZ 4586, contrasting with the rectangular condition of the distal end in other scelidotheriines (McDonald 1987).

Proximal Phalanx of Digit IV

In CORD PZ 4464, this phalanx presents a deep depression on its palmar surface. The region of the distal articular surface near the palmoaxial corner is approximately flat. In palmar view, the distal end is more axially set than the proximal.

Intermediate(?) Phalanx of Digit IV

In CORD PZ 4464, the most distal preserved element of digit IV presents on the dorsal surface a prominence (Fig. 2). The distal end of the phalanx is unpreserved. In palmar view, the phalanx presents its distal end more axially positioned than the proximal.

Proximal Phalanx of Digit V

The distal articular surface faces distoaxially, and the abaxial surface of the bone is very convex proximodistally.

Intermediate(?) Phalanx of Digit V

In CORD PZ 4464, this phalanx is free from the proximal phalanx of the fifth digit, as in some, but not all, specimens of Scelidotherium (Winge 1915; Cuenca Anaya 1995). It is the shortest and smallest preserved phalanx in CORD PZ 4464. The phalanx narrows distally, but its shape is incompletely exposed.

Other Manual Bones

Palmar Sesamoid (Falciform)

In CORD PZ 4464, this dorsopalmarly flattened bone is nearly rectangular, as in C. cuvieri, but unlike the rounded bone of V. bucklandi (Winge 1915).

Abaxial Sesamoid of Digit II

As far as we know, no published work has described this element in other scelidotheriine specimens. In CORD PZ 4464, this bone is characterized by a very prominent palmar keel that turns to extend into the proximal surface.

Abaxial Sesamoid of Digit III

As far as we know, no published work has described this element in other scelidotheriine specimens. In CORD PZ 4464, this bone is stouter than in mylodontines such as Paramylodon harlani (Owen, 1839) and M. darwinii (Stock 1925; Haro et al. 2016), as measured with its axioabaxial width vs. proximodistal length ratio (0.77). The sesamoid presents a prominent oblique ridge on its abaxial side that contrasts with the lesser and more longitudinal structure of M. darwinii. The ridge on the abaxial side of the palmar surface is relatively more proximodistally extended than in M. darwinii. The articular surface for the abaxial condyle is mediolaterally more concave than in M. darwinii. The bone is irregularly pentagonal in cross-section, instead of square as in M. darwinii (Haro et al. 2016). The bone is deeper than wide, unlike the depressed element of Paramylodon harlani (Stock 1925).

Axial Sesamoid of Digit III

As far as we know, no published work described this element in other scelidotheriine specimens. It is quite incomplete, and the preserved parts do not allow confirming there are differences from the condition described for Paramylodon harlani by Stock (1925).

Abaxial Sesamoid of Digit IV

As far as we know, no published work described this element in other scelidotheriine specimens. This incomplete sesamoid bears a concave articular surface for metacarpal IV and a convex palmar surface. The preserved parts do not allow confirming there are differences from the condition described for Paramylodon harlani by Stock (1925).

Phylogenetic Analysis

The phylogenetic analysis of the entire data matrix including S. leptocephalum, S. parodii, and the specimens PVUNS 198, CORD PZ 4464, 4586, and 11293 as different terminals (Analysis 1) produced 15 MPTs (L: 340.454; CI: 0.624; RI: 0.598), provided in Supplementary Data 1. The strict consensus (Fig. 10a) recovered all the specimens here reported, together with S. leptocephalum and S. parodii, within a group. Another recovered clade is a trichotomy between C. cuvieri, C. tarijensis, and C. chiliense. Sibyllotherium guenguelianum is recovered as the most basal taxon among the Scelidotheriinae. Neonematherium flabellatum Ameghino, 1904, is recovered as most closely related to the remaining scelidotheriines than Sibyllotherium guenguelianum. These remaining scelidotheriines form a large polytomy in which the only recognized groups are those corresponding to the genera Scelidotherium and Catonyx. Included outgroups to the Scelidotheriinae present the same topology as in Gaudin (2004). The apparent lack of resolution vanishes when hiding in the consensus the position of the species Pr. rothi and Proscelidodon patrius (Ameghino, 1888) (Fig. 10b), which are wildcard taxa. Hiding Pr. rothi alone exposes a monophyletic clade exclusively formed by the genera Catonyx and Scelidotherium, here informally called ‘clade A’. Hiding the position of Pr. patrius alone reveals a clade including V. bucklandi, the specimen PVUNS 198, Pr. rothi, and the genera Catonyx and Scelidotherium, which is here informally called ‘clade B’. Hiding both Pr. patrius and Pr. rothi, another clade that includes clade A and the specimen PVUNS 198 emerges, in addition to the formerly mentioned clades. Proscelidodon rothi is alternatively recovered as the sister taxon to the genus Catonyx or to V. bucklandi. Proscelidodon patrius is recovered either as: a) part of a trichotomy with the specimen PVUNS 198 and the clade A; b) part of a trichotomy with Pr. gracillimus and the clade B; or c) as the sister taxon of Pr. gracillimus. The unambiguous apomorphies of all the groups reported in this analysis and the others are provided in Supplementary Data S5. Bremer support and Jackknife frequency values are generally low. When the scores for the included specimens, together with those of S. parodii and S. leptocephalum, are used to form a single terminal corresponding to the genus Scelidotherium, considering all of these form a monophyletic group in the previous analysis, three MPTs (L: 333.925; CI: 0.658; RI: 0.594) are recovered from the analysis (here referred to as Analysis 2). The MPTs are provided in Supplementary Data 2. The strict consensus presents exactly the same groups as in Analysis 1 (Fig. 10c). The reduced strict consensus, after hiding the position of Pr. patrius and Pr. rothi, is also the same as in Analysis 1 (Fig. 10d). Jackknife frequency and Bremer support values increase compared to Analysis 1, but are still generally low.

Fig. 10

Phylogenetic relationships of the Scelidotheriinae, incorporating Scelidotherium Owen, 1839, specimens CORD PZ 4464, 4586, and 11293. a, Strict consensus of analysis including all terminals and all characters (Analysis 1 in text); b, part of reduced strict consensus of Analysis 1, obtained hiding the position of wildcard taxa Proscelidodon rothi and Proscelidodon patrius; c as in a, but merging Scelidotherium leptocephalum, Scelidotherium parodii, and CORD PZ 4464, 4586, and 11293 into a single terminal, Scelidotherium (Analysis 2 in text); d, part of reduced strict consensus of Analysis 2, obtained hiding the position of wildcard taxa Pr. rothi and Pr. patrius; e, as in c, but excluding all the characters belonging to the manus (Analysis 3 in text); f, as in c, but only including the characters belonging to the manus (Analysis 4 in text). Above branches are the respective Bremer support values; below the branches are the Jackknife frequencies

The analysis of the data matrix excluding the characters from the manus (here referred to as Analysis 3) produced a single MPT (L: 253; CI: 0.617; RI: 0.601). This tree is provided in Supplementary Data 3. The tree (Fig. 10e) presents the relationships of the basal part, including the scelidotheriine outgroups, Sibyllotherium guenguelianum and N. flabellatum, identical to the strict consensus of Analysis 2. On the other hand, except in the recovery of the clade represented by a trichotomy between C. chiliense, C. cuvieri, and C. tarijensis, it largely disagrees from the strict consensus of the other analyses regarding the relationships of post-Friasian scelidotheriines. A clade formed by Pr. rothi and V. bucklandi is recovered in this analysis as the sister taxon to the Catonyx group. The clade formed by Pr. rothi, V. bucklandi, and the species of the genus Catonyx presents Pr. patrius, Pr. gracillimus, and the genus Scelidotherium as successive outgroups. Jackknife frequency and Bremer support values are generally larger than in the previous analyses, although they are relatively low for the clades that include species of the genus Proscelidodon. The analysis of the data matrix only including the characters from the manus (here referred to as Analysis 4) recovers eight MPTs (L: 75.516; CI: 0.844; RI: 0.694), provided in Supplementary Data 4. The strict consensus (Fig. 10f) recovers a clade formed by Scelidotherium and the species of the genus Catonyx, and another formed by that clade and the specimen PVUNS 198. Hiding Pr. patrius and Pr. gracillimus from the strict consensus exposes a clade including M. darwinii, Ps. confusum, V. bucklandi, and the clade formed by Scelidotherium, the Catonyx species, and the specimen PVUNS 198. Hiding Pr. gracillimus, V. bucklandi, and M. darwinii from the strict consensus exposes Pr. patrius as basal to clade including Ps. confusum and the clade formed by Scelidotherium, the Catonyx species, and the specimen PVUNS 198. Bremer supports and Jackknife frequencies are rather low.

Discussion

Comparative Remarks

The previous comparisons clearly indicate that specimens CORD PZ 4464, 4586, and 11293 more closely resemble other specimens of Scelidotherium and C. cuvieri than V. bucklandi. At least the specimens CORD PZ 4464 and 11293 most closely resemble Scelidotherium than Catonyx. CORD PZ 4464 more closely resembles other specimens of Scelidotherium than Catonyx in the relatively deep scaphoid. CORD PZ 11293 shares with Scelidotherium, but not C. chiliense, an undivided trapezoid articular surface on the magnum. The specimen CORD PZ 4586, on the other hand, most closely resembles Scelidotherium in some traits, but Catonyx in others. CORD PZ 4586 more closely resembles other specimens of Scelidotherium than Catonyx in the radial and trapezoid articular facets separated by a considerable interval in the scaphoid, the facets for metacarpal III and unciform on metacarpal IV meeting at an obtuse angle, and the restricted distal articular surface on trapezoid-metacarpal I. On the other hand, the specimen CORD PZ 4586 more closely approaches the range of variation documented in Catonyx than that of previously described specimens of Scelidotherium in several proportions. Such proportions include the proximodistal height/ulnar facet width in the cuneiform, height of medial part/width in trapezoid, and proximal width/length in metacarpal II. These similarities may relate to the greater number of measurements provided for C. cuvieri than for Scelidotherium in the comparative work by Winge (1915), which increase the range of variation in the former. In addition, CORD PZ 4586 more closely resembles a specimen of C. cuvieri than any one described for Scelidotherium in the fusion between magnum and trapezoid. Finally, the three specimens here described share an unciform lacking the articular facet for metacarpal V with most known specimens of Scelidotherium, but not most known specimens of Catonyx.

Interestingly, many features of the Scelidotherium specimens here described differ from those in the S. leptocephalum specimen from Rosario de La Frontera reported by Esteban et al. (1992). As previously noted, CORD PZ 4464, 4586, and 11293 all differ from the Rosario de la Frontera specimen in having an unciform with the cuneiform articular facet presenting a dorsopalmarly directed concavity in its mediodorsal region rather than a flat condition. CORD PZ 4586 and 11293 most closely resemble each other than the Rosario de la Frontera specimen, because of having the articular surface for the scaphoid on the magnum band-shaped and sinuous instead of crescentic, with the dorsal part not tapering; and the unciform with the lunar articular surface exposed on the proximal aspect, the cuneiform articular facet irregularly reniform instead of rectangular, the articular surface for metacarpal III quadrangular instead of subcircular, and a non-rectangular parallelogram-shape. CORD PZ 4464 and 4586 share the articular surface for the ulna on the cuneiform roughly triangular, rather than subcircular as in the Rosario de la Frontera specimen. These differences may imply that the specimens here described are taxonomically distinct from the Rosario de la Frontera specimen, and would therefore do not belong to the species S. leptocephalum. Indeed, there is no evidence supporting the referral of CORD PZ 4464, 4586, and 11293 to that species, even if their Pleistocene age is suggestive—such evidence will only emerge in the form of overlapping material of S. parodii and/or further study on the variation in overlapping material of S. leptocephalum. Alternatively, the species S. leptocephalum would present a high variability in hand structure. That the latter is the case is consistent with the variability in manual structure described for C. cuvieri and V. bucklandi by Winge (1915), and in species of the genera Catonyx and Scelidotherium by McDonald (1987). The phenotypic variation found between the specimens here reported and the Rosario de La Frontera one may indicate the later belongs to a different geographical subspecies, which is possible considering all the specimens here described come from localities in the Córdoba Province that are at least 590 km from Rosario de la Frontera.

Although the presence of geographical variation has been considered in the cases of differences between specimens assigned to the same species that come from different regions, in the case of C. chiliense by McDonald (1987) and in the case of C. cuvieri by Corona et al. (2013), these studies did not go beyond proposing the possibility. The possible reason is that larger samples are required to find statistically significative differences between populations; at least this holds true for the present case. A larger wealth of material of Scelidotherium coming from Rosario de la Frontera, the Córdoba Province, and anywhere in between, is required to test the hypothesis of geographical variation in Scelidotherium.

Phylogenetic Considerations

The position of the specimens CORD PZ 4464, 4586, and 11293, closer to the species in the genus Scelidotherium than to those in the genus Catonyx in Analysis 1, is compatible with referring these specimens to the genus Scelidotherium (as in Haro et al. 2016; Krapovickas et al. 2017), and not with referral to Catonyx or any other previously recognized genus, on the grounds of parsimony. Their position is also compatible with the possibility that these specimens represent new genera, but we consider it more conservative to refer them to Scelidotherium given that the manus is unknown in one of the two species of the genus. The inclusion of new manual characters in Analysis 2 that supports a phylogenetic scheme in which Scelidotherium and Catonyx are more closely related to each other than either is to with V. bucklandi contradicts the results of Cartelle et al. (2009), Miño-Boilini (2012), and Miño-Boilini et al. (2014), but matches the pre-cladistic taxonomic scheme of Winge (1915), as well as the phylogenetic ideas of Kraglievich (1923), mostly based on upper tooth shape. When all the parsimony-informative characters, except for those belonging to the skeleton of the hand, are included, the interrelationships between the genera Scelidotherium, Catonyx, and Valgipes inferred by Cartelle et al. (2009) are supported. The analysis of hand characters alone supported the close relationship between Scelidotherium and Catonyx, and the basal position of Valgipes, as in the complete dataset, and unlike the other incomplete dataset. However, unlike all the other analyses, it recovered Ps. confusum closer to the genera Scelidotherium and Catonyx than Pr. patrius. Thus, both partial analyses do share only some of the groups recovered in the complete analysis, and contradict it regarding some groups. It is apparent both partial datasets contribute resolution for different groups in the complete analysis, and the result of their inclusion is not superfluous.

The results of the complete analyses allowed supporting the assignment of the genera Valgipes, Neonematherium Ameghino, 1904, Sibyllotherium Scillato-Yané and Carlini, 1998, and Proscelidodon, but not Nematherium, to the Scelidotheriinae, by using H. longiceps, Choloepus, and the mylodontines M. darwinii and Ps. confusum as additional outgroups. Our results indicate the use of the added outgroups, in which the osteology of the manus is better known than in the genus Nematherium, has been of fundamental importance in recognizing the derived nature of the similarities shared by the genera Scelidotherium and Catonyx but not V. bucklandi. With regard to the position of N. flabellatum, our analyses 1 and 3 support the results of Miño-Boilini (2012, 2014) regarding its basal position, which is compatible with its comparatively early geological age, over the hypothesis of close relationships to Scelidotherium supported by McDonald and Perea (2002) and Cartelle et al. (2009). The non-monophyletic relationships of the species of the genus Proscelidodon are supported in all the present and aforementioned previously published works. The scarcity of material from the skeleton of the manus in the species of that polyphyletic genus is probably responsible for their ambiguous position in our analysis, in face of the change in position of the scelidotheriines for which the hand characters are better known.

The present results allowed testing the alternative generic assignments of the specimen PVUNS 198, either to the genus Proscelidodon, as proposed by Aramayo (1988), or to the genus Scelidotherium, as proposed by Esteban et al. (1992), on the basis of the lack of articulation between unciform and metacarpal V. The latter hypothesis presents the problem that on the same ground the specimen can be referred to C. cuvieri. The position of that specimen basal to the dichotomy between the genera Scelidotherium and Catonyx in Analyses 1, 2, and 4 (the specimen is not included in analysis 3) indicates rejection of the Esteban et al.’s (1992) assignment. The assignment to Proscelidodon cannot be rejected or supported, however, given the variable position of the two species of the genus, Pr. patrius and Pr. rothi. In no MPT produced by the Analyses 1, 2, or 4 is the specimen PVUNS 198 recovered as more closely related to some species of the genus than to other scelidotheriine taxa. However, lack of resolution at the trichotomy between the specimen PVUNS 198, Pr. patrius, and the clade A admits as a possible resolution the grouping of the former two terminals, so the reference to that genus and species is not rejected nor supported. It is thus possible that some specimens of Pr. patrius lacked the facet, as in S. leptocephalum and C. cuvieri. Considering there are specimens with the facet in the genera Scelidotherium and Catonyx (Winge 1915; McDonald 1987), and others without it in the latter two genera, the possibility of variation in this regard on the genus Proscelidodon would imply that the feature is variable in the three main genera of the Scelidotheriinae, and thus of minor importance for systematics. On stratigraphic grounds, the possibility that PVUNS 198 belongs to Pr. patrius, which is the only scelidotheriine, and quite abundant, in the same beds (Ameghino 1888; Taglioretti et al. 2014), seems high. A thorough comparative study on the osteology of the manus of the reported Montehermosan scelidotheriine material, for which published data are scarce and insufficient, will be required to address this point. Our results indicate the Scelidotherium-Catonyx divergence seems to be at least Chapadmalalan in age according to our results, considering the age of the oldest unequivocal member of these clades, namely S. parodii, and the oldest known possible member, namely Pr. rothi. The Scelidotherium-Valgipes divergence would be at least Montehermosan in age, given the age of the oldest unequivocal member of these clades, namely specimen PVUNS 198. Our data indicate the Scelidotherium-Catonyx divergence would not have been posterior to the Chapadmalalan, given the age of the oldest unequivocal member of these clades, S. parodii.

Our results clearly differ from previous scelidotheriine phylogenies, which presented fewer characters from the skeleton of the manus, in the position of the genus Scelidotherium and the more resolved position of Pr. gracillimus. Other phylogenetic studies incorporating postcranial characters into data matrices dominated by cranial, mandibular, and dental characters also change the relationships inferred by the cranial, mandibular, and dental characters alone. For example, the analysis of Haro et al. (2017), which incorporated postcranial characters to the large data matrix of Gaudin (2004), contradicted the results of the latter in the position of the genera Lestodon Gervais, 1855, and Thinobadistes Hay, 1919. The analysis of Boscaini et al. (2019) contradicted the results of Gaudin (2004) in the position of Paramylodon harlani and M. darwinii, in addition to that of Lestodon and Thinobadistes. The analysis of Amson et al. (2017) on megatherian sloths contradicted the results of Gaudin (2004) regarding the position of basal megatherian genera. All of this indicates that the incorporation of further postcranial data into data matrices allows testing hypotheses based on cranial, mandibular, and dental data and surveying the postcranium for characters that vary between taxa is therefore a worthy enterprise. As in previous phylogenetic analyses based on morphology, such as Gaudin (2004), our analysis fails to recover Choloepus as a mylodontine, contrasting with recent paleoproteomic results (Presslee et al. 2019). Despite this, and the very different aspect of the hand of Choloepus compared with that of mylodontines (Owen 1842), our comparisons include a similarity between this genus and mylodontines more derived than Ps. confusum, namely the dorsopalmarly convex dorsal part of the articular surface for the scaphoid on the magnum (CORD PZ 4570; DMNS 6458; pers. obs.). Choloepus also resembles mylodontids more than other sloths in the lack of the ungual phalanx of digit IV (Schulthess 1920). Finally, and surprisingly, many features of the manus of Choloepus resemble those of scelidotheriines close to Scelidotherium. These include the lack of concavity in the articular surface for the ulna on the cuneiform, at least half of the articular facet for metacarpal II positioned in the dorsal part of the magnum, the absence of an articular facet for the metacarpal V on the unciform (through by metacarpal loss in Choloepus; Schulthess 1920), and the presence of a notch on the axial border of the distal articular surface of metacarpal II. Further comparative study of the manus of Choloepus seems therefore promising to provide useful tests to currently contradicting phylogenetic hypotheses based mostly on the head skeleton or molecules.

Anatomical Interpretations

Two strange features in the lunar of unspecified ‘Scelidotherium’ materials were described by Ortega (1967): (1) facet for the magnum wider palmarly than dorsally (unlike CORD PZ 4586 and other scelidotheriine material); and (2) ridge on that facet closer to the palmar border than to the dorsal (also unlike CORD PZ 4586 and C. cuvieri). The condition of that material would be the same as that of other scelidotheriines if the orientation of the lunar in the unspecified ‘Scelidotherium’ materials was misinterpreted by Ortega (1967), confusing dorsal for palmar. Fewer assumptions are required if Ortega (1967) misoriented the bone, than by inferring two autapomorphies in the unspecified ‘Scelidotherium’ materials he examined. On the cuneiform, the similarity in contour of the pisiform articular facet in CORD PZ 4464 and the facet plus adjacent curved ridge in CORD PZ 4586 suggests the ridge corresponds to the border of the articular facet in some moment of the ontogeny (either previous or later to the preserved stage).

Regarding the trapezium-metacarpal I, the rugose, raised ridge on the palmar surface of the specimens CORD PZ 4464 and 4586, and on the dorsal surface of the former, likely represents the boundary of the contact between the fused trapezium and metacarpal I. This is apparent because its orientation is similar to that of the unfused parts of the suture in a specimen of M. darwinii (CORD PZ 4570; pers. obs.). Therefore, it represents further evidence of this fusion in mylodontids, as interpreted for other mylodontid species by most authors (e.g., Lund 1842; Winge 1915; Stock 1925; McDonald 1987), despite exceptions (Owen 1842; Cuenca Anaya 1995). On the distal end of the trapezium-metacarpal I, the articular facet on the dorsoaxial surface in CORD PZ 4586 likely corresponds to a sesamoid on the M. adductor digiti I longus tendon.

On metacarpal IV, the articular facets for the sesamoid bones of CORD PZ 4586 indicate the sesamoids were not fused to the metacarpal, as in the C. cuvieri specimens described by Winge (1915), but unlike the S. leptocephalum specimens described in the same work (Winge 1915). The lack of an articular facet for the axial sesamoid on the metacarpal IV of CORD PZ 4464 and the presence in the same region of a bulge suggest fusion of the metacarpal IV to the axial sesamoid. Fusion between proximal and intermediate phalanges in the digit I of the manus has been suggested for V. bucklandi by Cartelle et al. (2009) and previously for Paramylodon harlani by Stock (1925). These hypotheses are likely incorrect, because the first digit has no more than two phalanges (including the ungual) in other mammals (Webb 1989) and most other tetrapods (Noble 1931; Romer 1956). When a single non-ungual phalanx is known, as in many specimens of Scelidotherium (and no evidence of fusion is present, as in CORD PZ 4464 and 4586), the most parsimonious interpretation is to infer loss of the ungual phalanx, rather to infer fusion between both phalanges and then loss of the ungual process in the distal one. More assumptions are required to suppose a fused neomorphic intermediate phalanx is present. Phalanges III-1 and III-2 likely did not fuse in CORD PZ 4586, given that the abaxial palmar processes of both phalanges are closer together in that specimen than in CORD PZ 4464. This suggests the former specimen departs in this feature from V. bucklandi (Winge 1915; Cartelle et al. 2009). On the distal element of digit IV in CORD PZ 4464, the prominence on the dorsal surface closely resembles the extensor process of a distal phalanx more than any other feature on the dorsal surface of other intermediate phalanges. This suggests that the distalmost phalanx of digit IV of the manus formed through a fusion of the originally intermediate and distal phalanges, resembling the condition suggested for C. cuvieri by Winge (1915), and differing from the unfused condition of C. tarijensis (Cuenca Anaya 1995).

Conclusions

Currently reviewed evidence supports referral of the presently described specimens CORD PZ 4464, 4586, and 11293 to the genus Scelidotherium. Variation is present within the genus Scelidotherium, as indicated by the differences seen in the material from Rosario de la Frontera and the Province of Córdoba. The anatomy of the sloth manus provides evidence useful for understanding the relationships of scelidotheriines, as it does with mylodontines. These characters include a metacarpal II wider at the proximal end than 55% its proximodistal length, a sharp ridge on the articular surface for metacarpal II in the trapezoid, an articular facet for metacarpal II on the magnum largely on the dorsal region of the bone, a notch on the axial border of the distal articular surface of metacarpal II, and a markedly flattened distal phalanx in digit II of the manus, which indicate a closer relationship between the genera Scelidotherium and Catonyx than with Valgipes. The proposed phylogenetic scheme coming from inclusion of data from the structure of the skeleton of the manus only applies to the genera Scelidotherium, Catonyx, and Valgipes, and the species Pr. gracillimus. The position of the other species of the genus Proscelidodon is however less certain, probably in connection with their less well-known skeleton of the manus.