Megantereon, Croizet & Jobert, 1828
publication ID |
https://doi.org/ 10.1111/j.1096-3642.2007.00333.x |
persistent identifier |
https://treatment.plazi.org/id/039FFA63-A835-FFC8-FC36-31E2E889D221 |
treatment provided by |
Felipe |
scientific name |
Megantereon |
status |
|
OSTEOLOGY OF MEGANTEREON
SKULL AND UPPER DENTITION
The skull of Megantereon cultridens SE 311 ( Fig. 1 View Figure 1 ; Table 1) has been figured in several publications over the years, e.g. Piveteau (1961), Kurtén (1968) and, most recently, Palmquist et al. (2007). It differs from other sabrecats, with the sole exception of some specimens of Smilodon populator , in having a nearly vertical occiput relative to the long axis of the skull. Unfortunately, much of the skull is heavily restored, as also pointed out by Antón & Werdelin (1998), and the area behind the coronal suture on the left side, the entire right side, most of the palatal region and the entire basicranial region are restored ( Figs 1 View Figure 1 , 2 View Figure 2 ). Other skulls of Megantereon sp. , which we concur with Turner & Antón (1997) are also referable to M. cultridens , indicate that the occipital region was probably posteriorly inclined ( Vekua, 1995; Antón & Werdelin, 1998), for instance MNHNPer2001a (Museum national d’Histoire Naturelle, Paris; pers. obs). The dentition in SE311 is, however, excellently preserved. The resemblance to S. populator in the restored areas is probably not accidental. In his brief report, Schaub (1925: 255–256) states that ‘Zum vergleich steht mir neben rezenten Feliden der Gipsabguss des Skeletts von Smilodon neogaeus [= S. populator ] Lund... zur Verfügung’ (‘For com- *As reconstructed.
†Excluding the protocone.
‡Anterior part of symphysis restored.
§See Emerson & Radinsky (1980).
Abbreviations: ant, anterior to; AP, anteroposterior diameter; cor, height of coronoid process above jaw condyle; D, dorsoventral depth; Di P 3 –C 1, diastema between third upper premolar and upper canine; Jj–C 1, anteroposterior distance from centre of jaw cotyle to centre of upper canine; Jj–I 3, anteroposterior distance from centre of jaw cotyle to centre of third upper incisor; Jj–P 4, anteroposterior distance from centre of jaw cotyle to upper carnassial paracone; L, anteroposterior length; LM, labiolingual diameter; MAM, inferred moment arm of the masseter about the temporomandibular joint; MAT, inferred moment arm of the temporalis about the temporomandibular joint; post, posterior to; ZW, width across zygomatic arches.
parative purposes, besides recent felids, I have a plaster cast of the skeleton of Smilodon neogaeus [= S. populator ] Lund... at my disposal’). It thus appears likely that Smilodon was used as a template for reconstruction the skull of SE311 ( Fig. 2A–D View Figure 2 ).
The left premaxilla is virtually complete with the exception of the labial part in front of the I 1 alveolus. Of the right premaxilla, only the area between I 1, I 2 and the incisive foramen is preserved ( Fig. 1A, B, E View Figure 1 ). The left premaxilla is partly preserved from the premaxillary suture to the sutures of the nasal and frontal bones, along with nearly all of the ventral part. The left palate is only preserved adjacent to the left M 1 alveolus. The left jugal is preserved in the area surrounding the preorbital foramen and the upper part adjacent to the sutures of the remaining part of the zygomatic arch, including the postorbital process.
The zygomatic arch is distinctly antereroposteriorly shorter than in comparably sized extant large felids, and is also distinctly taller dorsoventrally, which resembles the condition in other derived sabrecats, such as Smilodon . Medially, there are distinct muscle scars for the origin of the m. masseter (both pars profunda and pars superficialis) The preorbital foramen is slightly larger than in extant large felids of similar size, but not quite as hypertrophied as in Smilodon and Homotherium . The area immediately anteroventrally to the foramen is not sculptured, and, accordingly, lacks any distinctive signs of having had muscular attachments. Naples & Martin (2000) suggested that the hypertrophied preorbital foramina of derived machairodonts and nimravids were indicative of a condition known as hysticomorphy, where the anterior masseter passes through the foramen, as in certain extant rodents. This appears not to have been the case in Megantereon .
Only the lowermost part of the left temporal (squamous portion) is preserved, extending caudally to the squamous suture, and a small part of the occipital crest is also preserved. The majority of the left glenoid fossa is preserved, with the exception of the innermost part adjacent to the orbitosphenoid. Both the pre- and the post-glenoid processes are preserved; the post-glenoid process is distinctly ventrally deflected, which is a condition common to all derived sabrecats and nimravids, and is probably an adaptation for attaining a large gape ( Emerson & Radinsky, 1980; Christiansen, 2006; see also Fig. 29 View Figure 29 , below). Around half of the left frontal is preserved, including most of the anterior postorbital process and orbital foramen, adjacent to the palatine, along with the ethmoid and lacrimal. The left nasal is preserved along the sutures towards the maxilla and the frontal, but unfortunately not the portion that was to reach the internasal suture. The facial part of the skull appears to display a less posterior inclination of the facial part relative to the basicranial part of the skull than reported for Homotherium and Smilodon ( Emerson & Radinsky, 1980; Christiansen, 2006), which is a further adaptation for attaining a large gape, but the marked degree of restoration makes this tentative.
Although the palate is only partly preserved, it is evident that Megantereon did not possess the palatal ridges that are present in Smilodon ( Fig. 1B, D View Figure 1 ). The nasal aperture appears to have been distinctly proportionally larger than in extant felids, and broadly comparable with the equally hypertrophied apertures in Smilodon and Homotherium . The rostrum is comparable with the Dminisi specimen, described by Vekua (1995) as M. megantereon (= M. cultridens ) and to the Perrier specimen MNHNPer2001a, although M. ‘ megantereon ’ lacks the nasals. The temporal fenestra are incompletely preserved ( Fig. 1B, C View Figure 1 ), but it is fairly obvious that the size of the temporal musculature must have been distinctly less than in extant large felids ( Fig. 2 View Figure 2 ), and the bite force probably correspondingly lower.
All incisors are preserved, with the exception of the right I 1, where the crown is broken off, and the left I 3, where the tip of the main cusp is broken off. The preserved incisors are large, pointed and tricuspid, with one main cusp and two posterior accessory cusps, as commonly seen in other derived machairodonts. This condition differs from that of modern felids, which all have proportionally smaller, transversely spatulate I 1 and I 2, and where only I 3 is reminiscent of the shape observed in Megantereon and other derived sabrecats ( Fig. 1B, E View Figure 1 ). The incisor arcade is procumbent and rounded, as in other derived sabrecats ( Biknevicius, Van Valkenburgh & Walker, 1996), creating a conspicuously large diastema to the upper canine ( Table 1), but is less anteriorly protruding than in Homotherium .
Both canines are well preserved, and only the tips were broken off but have subsequently been glued back together. They are enormously elongate and blade-like ( Table 1), and entirely different from those of any extant carnivoran, and among sabrecats, only Smilodon has upper canines of comparable length. The cemento-enamel junction is located far from the alveolar margin, but less so at the mesial and distal crests than at the vestibular and lingual crests, as also found in Smilodon ( Riviere & Wheeler, 2005; our pers. observ.). The canines, along with the rest of the dentition, lack crenulations, and appear largely unworn. In Smilodon , the canines tend to be slightly crenulated, and the incisors may also have slight crenulations along the cutting edges. In Homotherium , however, the entire dentition tends to be more or less serrated, especially the canines.
SE311 lacks a P 2, which is commonly found in extant lynxes ( Werdelin, 1987), whereas most other modern felids have just a vestigial P 2. In Smilodon and Homotherium , it is absent. Both the left and the right P 3 are preserved, and are laterally compressed with a small parastyle, a reduced paracone and a small metastyle ( Fig. 3 View Figure 3 ). The cingulum is distally distinct, as in extant Acinonyx . The size of the parastyle and metastyle are not unusual compared with pantherine cats, but the paracone is distinctly smaller than observed in modern felids. Accordingly, the tooth as such is only slightly reduced, but not to the extent seen in Smilodon or Homotherium , in which the tooth is vestigial.
Both carnassials (P 4) are preserved, and are very large and mesiodistally elongate ( Figs 1A, B View Figure 1 , 2A, B View Figure 2 , 3A View Figure 3 ; Table 1) with a distinct parastyle, a paracone, a metastyle blade and a small protocone. The presence of a cusped protocone differs from the condition observed in several other, derived sabrecats, such as Smilodon and Homotherium , but also from advanced forms of Machairodus , where this cusp is absent. No ectoparastyle is present, contrasting with several other machairodonts. Owing to the better developed parastyle, with distinct sheering crests, and the small protocone, the tooth overall has a more blade-like appearance than in any pantherine, bearing a closer resemblance to the carnassial of Acinonyx jubatus , but it is not as advanced as observed in Smilodon or Homotherium . Both M 1 are preserved and are much reduced, but no more than often seen in extant felids.
MANDIBLE AND LOWER DENTITION
The left hemimandible is nearly complete with the exception of the labial part at the canine alveolus ( Fig. 4A, B View Figure 4 ; Table 1). The right hemimandible ( Fig. 4C, D View Figure 4 ) is much more fragmentary. Labially, the anterior portion is preserved, although not the area below the teeth, and lingually the ramus is preserved to just posterior to the carnassial, although with some restoration around the symphysis. The posterior part of the hemimandible, the angular process, condyle and the coronoid process is entirely restored, and looks similar to preserved portions of its left counterpart, although the coronoid process has been made larger (compare Fig. 4A, C View Figure 4 ). The dental formula for Megantereon SE 311 (3/3 1/1 2/2 1/1) differs from those of extant pantherines (3/3 1/1 3/3 1/1). The mandible of the Dmanisi Megantereon ( Vekua, 1995) is similar to that of SE311, with the exception of P 3 being more reduced in the former, and clearly offset from P 4. This could merely be intraspecific variation, and not nec- essarily an argument for species differentiation. In MNHNPer2001b, P 3 is only slightly more reduced than in SE311, but is no more offset from P 4.
The large, rather robust incisors form a broad, protruding, subcircular battery, which includes the lower canine. Immediately posterior to the canine, the mandibular corpus curves lingually. An unusually large and distinct, but labiolingually thin mandibular flange is present, and it is distinctly larger than observed in Homotherium or Smilodon gracilis , resembling the condition commonly observed in certain nimravids, such as Hoplophoneus and Eusmilus . This morphology is entirely absent in any extant carnivoran. The symphyseal plates are very tall and nearly vertical, but rather shallow. The plates display a high degree of rugosity, as commonly seen in large felids, but lack the distal depth of extant large felids. A single, large mental foramen is present.
Viewed dorsally ( Fig. 4E View Figure 4 ), the mandibular corpus curves distinctly medially around the premolars and carnassial, a condition known as carnassial rotation ( Figs 4E View Figure 4 , 5C, G, K View Figure 5 ; Bryant & Russell, 1995), resulting in a vestibular outwards turning of the teeth. The mandibular fossa is deep, but lower than observed in pantherines, owing to the much reduced coronoid process. The fossa reaches far anteriorly and protrudes below M 1, which is unusual in pantherines, but common in other derived sabrecats. The coronoid process is about the same height relative to the length of the ramus ( Table 1) as in Smilodon and Homotherium , but is relatively shorter in the anteroposterior direction, rendering it more gracile in appearance than in the latter forms ( Fig. 5A, E View Figure 5 ). The condylar process is transversely shorter and stouter than observed in pantherines, but not as stout as in Smilodon and Homotherium . The angular process is ventrally deflected and lingually directed, resulting in the jaw condyle protruding posteriorly to the angular process, a condition also found in other derived machairodonts, but unusual in pantherines ( Figs 4A, B View Figure 4 , 5I, J View Figure 5 ).
All teeth are preserved and no empty alveoli are present. The incisors are pointed and procumbent, as in other derived sabrecats ( Biknevicius et al., 1996), with a main cusp and two posterior accessory cusps. I 3 is the largest, followed by I 2 and I 1 ( Table 1). The canine is much reduced in height, comparable with the condition in some species of Machairodus , but not completely incisiform as observed in Smilodon or Homotherium , as the canine is substantially larger than I 3 ( Table 1; Fig. 5A, E View Figure 5 ). The teeth lack crenulations. P 3 is reduced in size, but distinctly larger than the P 3 of the Dmanisi specimen of Megantereon ( Vekua, 1995; see also Dzaparidze et al., 1992). It is also directly adjacent to P 4, whereas in the former the two are separated by a large diastema. The tooth has a minute parastyle, a paracone and a small metastyle. This is similar to MNHNPer 2001b.
P 4 has a distinct parastyle, a paracone and a small metastyle. The cusps are more pointed than in extant pantherines ( Fig. 3A, E, F View Figure 3 ) and are posteriorly inclined, but less so than in Smilodon , where the cusps of P 4 overlap with those of M 1, thus in effect creating one long shearing blade. The cheetah also has pointed cusps ( Fig. 3D View Figure 3 ), but these are more bladelike, unlike the condition in Megantereon and other machairodonts. Contrary to the condition in Smilodon and Homotherium , the tooth is unworn ( Fig. 3A–C View Figure 3 ). M 1 has a distinct paraconid and protoconid, but no discernible talonid. The tooth displays distinct wear facets ( Fig. 3A View Figure 3 ), and a deep, worn, carnassial notch, distinctly deeper than usually observed in Smilodon , but not comparable to the extreme wear frequently present in Homotherium ( Fig. 3C View Figure 3 ) where the general shape of the tooth is often entirely altered by attrition (rather than abrasion, as indicated by the clearly visible traces of the P 4 cusps). This could indicate that Meganteron, like Smilodon , processed carcasses in a different manner from Homotherium . The wear on the carnassials in machairodonts is not comparable to the condition of the spotted hyena ( Crocuta crocuta ), where the carnassial cusps are worn down by attrition, whereas wear in the former is always distinctly angular.
AXIAL SKELETON
The presacral vertebral column in Megantereon cultridens SE 311 is almost complete ( Fig. 6 View Figure 6 ; Table 2), and the vertebral pattern of extant felids of seven The measurements given are as preserved, and in several cases the actual value should have been slightly higher, as indicated by a plus symbol.
Abbreviations: Azw, width across prezygapophyses; C, cervical vertebra; Ch, posterior height of centrum; L, lumbar vertebra; Lc, length of centrum; Lz, length of centrum from anterior margin of prezygapophysis to posterior margin of postzygapophysis; Ns, height of neural spine along its longitudinal axis; Pzw, width across postzygapophyses; T, thoracal vertebra; Tp, width across transverse processes.
cervical, 13 thoracic, seven lumbar and three sacral vertebrae is also present in Megantereon . As in other felids, the centra are generally amphiplatyan, albeit with gentle opisthocoeli in the cervicals. The apparent completeness of the presacral vertebral column of SE311 tends, however, to conceal the fact that several vertebrae are less well preserved, and in some cases key features are not discernible. Fortunately, most vertebrae are well preserved, and although several have been restored, this usually only amounts to cosmetic restoration. In several specimens of Smilodon fatalis , there is evidence of an extra thoracic dorsal or, alternatively, of one fewer lumbar dorsal vertebra ( Merriam & Stock, 1932), but this is not the case in Megantereon , despite controversy about the identity of one of the dorsals, as noted below.
The vertebral column of well-known sabretoothed felids appears to differ from the morphology observed in large felines, such as extant and extinct Panthera , especially with regards to proportions and morphology of the cervical vertebrae ( Schaub, 1925; Merriam & Stock, 1932; Rawn-Schatzinger, 1992; Turner & Antón, 1997; Antón & Galobart, 1999). However, a detailed comparison of vertebral proportions with extant large felids has not previously been made for any of the few sabrecats for which such a comparison is possible. For the purpose of this study, we measured the centrum length, posterior centrum height, width across the transverse processes (where applicable) and height of the neural spine of all 27 presacral vertebrae in 18 specimens of extant large felids (five Panthera leo ; three P. onca, three P. pardus, four P. tigris and three Puma concolor ), to which the vertebral proportions of Megantereon were then compared. Despite the frequent divergence of the vertebral proportions in Megantereon from extant large felids, especially with respect to the cervicals, Megantereon was built much along the lines of extant large felines, as shown by the uniformity of pattern of vertebral proportions along the presacral vertebral column ( Fig. 7 View Figure 7 ).
Cervical vertebrae
Schaub (1925) was the first to note the apparent hypertrophy of the anterior cervicals of M. cultridens SE 311. All seven cervical vertebrae are present, and in general they are well preserved, although minor restoration has been carried out in several places. It is apparent that the cervicals, and not merely the anterior ones, are indeed proportionally larger than corresponding vertebrae in extant felids ( Fig. 7A View Figure 7 ). In Megantereon , the great length of the cervical centra also makes them appear slightly gracile compared with extant large felids when comparing centrum length to posterior centrum height ( Fig. 7B View Figure 7 ), or the height of the neural spine compared with centrum length ( Fig. 7C View Figure 7 ). However, the transverse processes, when preserved, are well developed, and even fall within the range of extant large felids, despite the size of the vertebral centra ( Fig. 7D View Figure 7 ). The proportional size differences and divergent morphology of sabretoothed felid cervicals from extant felids has been attributed to the differences in the functional morphology of the upper cervical region, which is inferred to have resulted in different bite mechanics, as discussed under the palaeobiology section below.
Atlas (C1)
The atlas of Megantereon is a large, heavily constructed bone, and the vertebral body is distinctly more dorsoventrally massive than in extant large felids, but bears a close resemblance to the morphology in Smilodon ( Merriam & Stock, 1932) and Homotherium ( Rawn-Schatzinger, 1992; see also Antón & Galobart, 1999). Unfortunately, the atlas in Megantereon is one of the least well preserved vertebrae. The corpus is large and massive and the medial tuberosity is long, low and massive, as in Smilodon ( Merriam & Stock, 1932) . The vertebral arch is much less deeply intended than in Homotherium (Rawn- Schatzinger, 1992), again resembling the condition in Smilodon ( Merriam & Stock, 1932) . The articular facets for the occipital condyles are deeply concave and are overhung by a wall of bone along the dorsal margin, as in other felids. The neural canal is subcircular, as in other felids. As in other felids, the entrance to the neural canal of the vertebrarterial canal is situated anterolaterally on the dorsal side of the atlas corpus and exits posterolaterally on the ventral side of the corpus. Ventrally there is a posterior lip for articulation with a notch basally on the axis odontoid process, as in other felids.
The most distinguishing feature of the atlas in derived sabretoothed felids, and the character in which they display the greatest divergence from the condition in extant felids, are the hypertrophied atlas wings. Unfortunately, the allegedly preserved left atlas wing of Megantereon , which resembles those of Smilodon ( Figs 6 View Figure 6 , 8A, F View Figure 8 ) is entirely reconstructed. From the broken bases along the atlas corpus it is, however, evident that the transverse possesses must have been large, and the morphology of the closely related Smilodon and Homotherium , along with inferences of upper cervical function in derived sabretoothed felids ( Turner & Antón, 1997; Antón & Galobart, 1999; Antón et al., 2004; Salesa et al., 2005), makes it likely that the processes were turned posteriorly, unlike the condition in extant felids. Strictly speaking, however, this is not discernible from the preserved specimen. As noted above, the similarity of non-preserved characters in Megantereon to corresponding characters in Smilodon is probably not accidental. In extant felids, the atlas wings rarely extend much posterior to the posteriormost extent of the articulating facets for the axis ( Fig. 8K, P View Figure 8 ).
Axis (C2)
The axis of Megantereon is a very large and rather solidly constructed bone, as in other derived sabrecats. At a total length of nearly 85 mm, including the odontoid process ( Table 2), it is larger than the largest specimens of Homotherium from the Friesenhahn Cave ( Rawn-Schatzinger, 1992), despite this sabrecat being larger than Megantereon ( Turner & Antón, 1997) , and it is within the low range of Smilodon fatalis specimens from Rancho La Brea ( Merriam & Stock, 1932). Among extant felids, it is far larger than the axis in pumas (52.5–57.9 mm), leopards (49.2–59.5 mm) and jaguars (58.1–63.4 mm), and is comparable with male lions (e.g. CN1440: 85.5 mm, and CN6043: 86.9 mm) and almost the size of Siberian tigers (e.g. CN5698: 90.6 mm). The ventral keel along the centrum is well developed compared with extant large felids, indicating a strong attachment for the m. longus colli ( Barone, 1967; Crouch, 1969; Done et al., 2000). Among extant large cats, only Acinonyx has prominent ventral keels on the cervical centra.
As in other felids, the odontoid process is long and pointed at the anterior tip, but it forms a greater inclination to the long axis of the centrum than in Smilodon and extant large felids ( Fig. 8B, G, L, Q View Figure 8 ), indicating a slightly more inclined position of the axis relative to the atlas. The anterior articulating surfaces for the atlas are very large and extend distinctly further dorsally than in extant felids, resembling the condition in Smilodon ( Merriam & Stock, 1932) . Unlike Smilodon and extant felids, the dorsal part of the articulating surface is somewhat posteriorly inclined, resembling the condition in Homotherium ( Rawn-Schatzinger, 1992) . The postzygapophyses are very large and more elongate than in extant felids, resembling the condition in Smilodon ( Merriam & Stock, 1932) , but are slightly more inclined, and distinctly more so than is usually the case in extant felids. In Smilodon , there is a large foramen for the vertebrarterial canal on the posterolateral face of the axis, but this is absent in Megantereon . In extant pantherines, the foramen is frequently present in lions and jaguars but often absent in leopards ( Fig. 8B, G, L, Q View Figure 8 ). The transverse processes are preserved only as a fragment on the left side. The neural canal is taller relative to the transverse width than was the case in the atlas.
The neural spine appears very large, tall and winglike, but unfortunately the entire anterior part is reconstructed, and again resembles the condition of Smilodon ( Fig. 8B, G View Figure 8 ). As in Smilodon , the posterior neural spine greatly overhangs the postzygapophyses, albeit less extremely. Merriam & Stock (1932) stated that this great posterior overhang in Smilodon ‘presents a structure totally unlike that in the true felines’ but this is only partly correct, since it is absent in the extinct Panthera atrox ( Merriam & Stock, 1932) and in the extant lion, but is often present in jaguars, leopards and tigers ( Fig. 8L, Q View Figure 8 ), albeit not developed to the extreme observed in Smilodon or Megantereon . Viewed dorsally, the posterior part of the neural spine in Megantereon becomes wide and triangular, although markedly less so than in Smilodon , and the posterior border is straighter than the sinusoid border of Smilodon .
Third cervical
The third cervical is well preserved. As in other felids, the centrum and its respective central articulating facets are greatly inclined. The most noticeable deviation from extant felids lies in the enormous development of the transverse processes, distinctly more so than in extant large felids and Smilodon fatalis ( Fig. 8C, H, M, R View Figure 8 ). It is, however, comparable with the condition in the larger and more robust Smilodon populator . As in Smilodon , the transverse processes extend well posterior to the posterior face of the centrum, unlike the condition in extant felids. The diapophysial facet is markedly expanded compared with the process ramus, as in other felids, albeit more so than is usually the case in extant large felids. The hypertrophied transverse processes indicate particularly powerful m. longissimus cervicis and m. intertransversalis ( Barone, 1967; Crouch, 1969; Done et al., 2000).
As in Smilodon , the posterior hyperapophyses are well developed and project posterodorsally as two elongate prongs. In extant felids they are more knoblike and less well developed. The neural spine is low but elongate, and is less well developed than in some specimens of Smilodon ( Merriam & Stock, 1932) and most extant large felids. The neural canal is smaller and more rounded than is the case in the axis. A slit-like vertebrarterial foramen runs medial to a wall of bone along the dorsolateral side of the centrum, as in other felids. The ventral keel is also well developed. In dorsal view, C3 is markedly squarish in outline, but is less quadratic than in extant felids, owing to its proportionally greater centrum length ( Fig. 7A View Figure 7 ). The prezygapophyses are large and oval, and are more inclined than often seen in extant felids, resembling the condition in Smilodon ( Merriam & Stock, 1932) . This is also the case for the postzygapophyses. A round nutrient foramen extends into the base of the postzygapophysial pedicel, and in this position is often also found a foramen in extant felids, albeit much smaller.
Fourth cervical
The fourth cervical vertebra is also rather well preserved, although the right transverse process is incomplete. The central articulating facets are inclined relative to the long axis of the centrum, as in other felids. The neural spine is distinctly better developed than is the case in C3 and less anteriorly projecting than seen in Smilodon and most extant felids ( Fig. 8D, I, N, S View Figure 8 ). The transverse processes are less well developed than was the case in C3. This is a deviation from the condition in Smilodon , in which the transverse processes of C4 are comparably developed with C3 ( Merriam & Stock, 1932). Also unlike Smilodon , the diapophysial facet is not bifurcate. In extant felids a bifurcate condition is frequently seen, albeit less extremely so than in Smilodon .
Ventrally, the inferior lamina is well developed, more so than in Smilodon . In extant felids, the inferior lamina is also usually well developed, albeit less so than in Megantereon . This indicates a strong development of the m. intertransversalis for twisting the cervicals about their long axis ( Barone, 1967; Crouch, 1969; Done et al., 2000). The ventral keel is also well developed. The hyperapophyses are well developed, and resemble those of C3, although slightly thinner. The neural canal is more oval than is the case in C3. The prezygapophyses and postzygapophyses are more inclined than seen in extant large felids, and resemble the condition in Smilodon . The angular difference in inclination from the condition in extant large felids is, however, less than seen in C3.
Fifth cervical
Cervical five is quite well preserved, although the right transverse process is fragmentary and the neural spine has lost much of its height. It was, however, clearly well developed, as seen from its anteroposterior extension and lateromedial width. The centrum has inclined articulating facets, as in other felids. The lateral face of the centrum has undergone cosmetic restoration, and it is unknown whether a depression was present, as in Smilodon ( Merriam & Stock, 1932) . The left transverse process is less well developed than in C4, and distinctly less so than in Smilodon . The diapophysial facet is also not birfurcate, as in Smilodon . The dorsal transverse process lamella is not preserved, but since it is present in Smilodon and extant felids it can be inferred for Megantereon as well. The inferior lamella appears to have been rather well developed, but is not preserved in its entity on either side of the centrum. The above indicates a proportionally stronger development of the m scalenus primae costae than m. scalenus supracostalis ( Barone, 1967; Crouch, 1969; Done et al., 2000). The prezygapophyses and postzygapophyses are slightly more inclined than in C4. The hyperapophyses are distinctly smaller than in C3 and C4.
Sixth cervical
Cervical six is well preserved and has only undergone cosmetic restoration along the inferior lamellae and transverse processes. It bears substantial resemblance to that of Smilodon ( Fig. 8E, J View Figure 8 ), but the neural spine is less well developed and the postzygapophysis is proportionally larger and less steeply inclined. The transverse processes are well developed, but are slightly more gracile than the condition in Smilodon . In extant felids, the hyperapophyses are usually small, wing-like extensions dorsally to the postzyga- pophyses, but in both Megantereon and Smilodon they are distinctly less well developed on C6 ( Fig. 8E, J, O, T View Figure 8 ). The vascular foramen along the centrum is distinctly larger than in C4–5. The greatest divergence from extant felids is found in the enormous development of the inferior lamina ( Fig. 8E View Figure 8 ). This indicates a more powerful development of the m scalenus primae costae than the m. scalenus supracostalis ( Barone, 1967; Crouch, 1969; Done et al., 2000). The lamella is less triangular than in Smilodon and does not extend as markedly posteroventrally.
Seventh cervical
Cervical seven is well preserved and only the tip of the neural spine and smaller sections of the left transverse process are missing. As in other felids, the centrum articulating facets are less steeply inclined compared with the long axis of the centrum than is the case in the preceding cervicals. Also, as in other felids, the neural spine is distinctly larger compared with centrum length than in the preceding cervicals ( Fig. 7C View Figure 7 ). The postzygapophyses are less steeply inclined than in the preceding cervicals, unlike the condition often seen in extant felids, where there is less difference in inclination. The transverse processes are less well developed compared with centrum length than in Smilodon ( Merriam & Stock, 1932) , but comparable with the condition in extant large felids ( Fig. 7D View Figure 7 ). There is no indication of an inferior lamella. The hyperapophyses are very small, and distinctly less well developed than in extant large felids. The neural canal is lateromedially wider than in the preceding cervicals. There is no vertebrarterial foramen along the sides of the centrum, as in extant large felids. In some specimens of Smilodon this feature may, however, occasionally be present ( Merriam & Stock, 1932).
Dorsal vertebrae
In general, the dorsal vertebral column is well preserved, although several vertebrae have undergone cosmetic restoration. In some cases, for instance T2–7, a more extensive restoration around the neural arches has taken place, resulting in a peculiar anterior inclination of the neural spine of T2 ( Figs 6 View Figure 6 , 9B View Figure 9 ). Generally, the dorsal vertebral column differs less in Megantereon from the proportions and morphology observed in extant large felids than do the cervicals. The centra are, however, often slightly more massive in comparison with their length, and the transverse processes are well developed compared with the length of the centrum ( Fig. 7B, D View Figure 7 ). The neural spines appear well developed in Megantereon , but relative to centrum lengths they are, in fact, comparable with those of extant large felids, and in the anterior dorsals they are actually less well developed ( Fig. 7C View Figure 7 ). In Smilodon , the thoracic vertebrae usually have very well-developed neural spines ( Fig. 9G–J View Figure 9 ; Merriam & Stock, 1932), but in Homotherium ( Rawn-Schatzinger, 1992) they appear to be distinctly lower than in extant felids, and also Megantereon .
First thoracic
The first thoracic vertebra is solidly constructed; more so than in extant felids, and the neural spine, although lower than in extant felids, is very wide anteroposteriorly. It is, however, more lateromedially slender than the condition in Smilodon and the upper section appears to have been slightly less lateromedially dilated, although the apex is missing. The upper part of the neural spine of Megantereon is triangular in cross-section, resembling the condition in Homotherium ( Rawn-Schatzinger, 1992) . The transverse processes are short and massive, and are of similar width relative to centrum length to those of extant large felids ( Fig. 7D View Figure 7 ), but are slightly less well developed than in Smilodon . The concave and elongate diapophysial facet is well developed and is almost the length of the centrum, indicating that the tuberculum of the first dorsal rib was large. The diapophysis in Megantereon is proportionally larger than those of extant felids, although tigers and lions sometimes also have very large diaphysial facets ( Fig. 9A, M View Figure 9 ), and also than that of Smilodon ( Fig. 9G View Figure 9 ). The parapophysis is situated anteroventrally at the basis of the transverse process adjacent to the anterior articulating facet of the centrum, as in other felids, but is less well demarcated than in extant large felids and Smilodon , probably owing to postmortem abrasion. In Smilodon , there is a distinct ventral keel on T1 ( Merriam & Stock, 1932), and this is frequently also seen in extant large felids, albeit less well developed. In Megantereon , a low ridge runs along the ventral part of the centrum, but it is less well developed than in Smilodon . The prezygapophyses are inclined to around a 45° angle, as in other felids. The postzygapophyses are, however, slightly more steeply inclined than in Smilodon and extant large felids. The neural canal is distinctly lateromedially oval, as in Smilodon and extant felids, and less circular than in Homotherium ( Rawn-Schatzinger, 1992) .
Second thoracic
The second thoracic vertebra is almost complete but for a small portion of the right transverse process. In Homotherium , T1 and T2 are very similar (Rawn- Schatzinger, 1992). This is not the case in Megantereon , Smilodon and extant large felids, where the neural spine of T2 is not just larger in absolute measure, but is distinctly larger proportionally to centrum length, and the transverse processes are proportionally distinctly more narrow compared with centrum length than is the case in T1 ( Fig. 7C, D View Figure 7 ). In Megantereon , the difference in neural spine height relative to the length of the centrum is readily apparent ( Figs 6 View Figure 6 , 9A, B View Figure 9 ), although the apex is missing in T1. This is both owing to the increased height of the spine itself relative to the spine of T1, as in other felids, but also owing to the centrum of T2 being distinctly longer relative to T1. Compared with the condition in extant felids, this is, however, not owing to a proportionally small T2 but a proportionally large T1 ( Fig. 7A View Figure 7 ). In extant felids, the centrum of T2 can be both smaller, of similar size and larger than T1, but usually by no more than 1–2 mm, whereas the difference is> 5 mm in Megantereon . Smilodon and Homotherium appear to follow the condition observed in extant felids ( Merriam & Stock, 1932; Rawn-Schatzinger, 1992).
The diapophysial facet is distinctly smaller than in T1, and more resembles the condition of other felids and faces more laterally than on T1, as in other felids. Also, the parapophysis has migrated to the anterolateral face of the centrum and is situated around the middle of the centrum, which is slightly lower than in Smilodon and extant large felids ( Fig. 9B, H, N, T View Figure 9 ). The posterior demifacet is markedly larger on T2 than on T1.
Thoracic 3–10
The pattern along thoracic vertebrae 3 to 10 in Megantereon follows that of other felids. The parapophysis is situated at the upper, anterolateral part of the centrum on T3, and remains approximately in this position in the succeeding thoracics, as in other felids. The posterior demifacet tends to be situated slightly higher than on T2, but also remains in approximately the same position and is of a similar size throughout T3–10. The percentile length of the centra of T3–10 of the entire presacral vertebral column follows the patterns observed in extant large felids closely ( Fig. 7A View Figure 7 ). The centra do, however, appear somewhat heavier in build than in extant felids, as seen from their greater dorsoventral diameters relative to length along T3–10 ( Fig. 7B View Figure 7 ). The pre- and postzygapophyses are horizontally orientated and gently slope ventolaterally, and the postzygapophyses are usually slightly more inclined. This is similar to the condition in other felids.
The height of the neural spines along T3–10 relative to centrum length follows the pattern observed in extant large felids closely ( Fig. 7C View Figure 7 ), although they are somewhat shorter in T3–5. As in other felids, the neural spines become progressively more posteriorly inclined towards the rear, having an inclination of around 35–40° to the long axis of the centrum in T8–10 ( Fig. 9C, I, O, U View Figure 9 ). Where preserved ( Fig. 6 View Figure 6 ), the transverse processes are proportionally wider and more heavy in overall build than in extant large felids ( Fig. 7D View Figure 7 ), and they bear substantial resemblance to corresponding vertebrae in Smilodon ( Merriam & Stock, 1932) . The neural canal gradually becomes less oval further posteriorly and also decreases in absolute size.
The diapophyses show slightly varying degrees of posterior and lateral inclination along the thoracic vertebral column, but this is clearly both owing to some actual variation of the facets along the vertebral column, as seen in other sabrecats such as Smilodon ( Merriam & Stock, 1932) and Homotherium (Rawn- Schatzinger, 1992), and extant large felids, but is in some cases also due to restoration, for instance on T6 ( Fig. 6 View Figure 6 ). The metapophyses are fairly prominent on T3–10, and in general become progressively larger towards the rear, as in Smilodon and extant large felids. This does not necessarily occur linearly, however, as the metapophysis is large on T7 and T10, but is less well developed on T8 and T9 (see Fig. 9C View Figure 9 ). Such variation is also seen in other felids. As in other felids there are no anapophyses prior to T11.
Eleventh thoracic
This vertebra is unfortunately not well preserved, and has been restored with peculiar-looking postzygapophyses situated on thin pedicels ( Fig. 6 View Figure 6 ). This imperfect preservation results in a morphology that is quite different from a posterior thoracic or lumbar vertebra in any felid. In the mounted skeleton of Megantereon (see Fig. 27 View Figure 27 , further below), it was placed as the third lumbar, but this is evidently incorrect. Although lacking the entire upper section above the neural arches, thus making identification difficult, the centrum bears a distinct parapophysis and shows evidence of having had rather thick transverse processes situated on the neural pedicels and not the lateral face of the centrum, and is thus congruent with T11 and not with a lumbar, or, for that matter T12 or T13. The vertebra is proportionally slightly shorter than in extant large felids ( Fig. 7A View Figure 7 ) and is heavy in build ( Fig. 7B View Figure 7 ). The posterior-most dorsals and lumbars also show the characteristic size increase compared with the thoracals observed in other felids.
According to Rawn-Schatzinger (1992), the transition from anterior-mid to posterior dorsals in Homotherium occurs at T12, where the prezygapophyses are orientated almost horizontally, as in the preceding dorsals, but the postzygapophyses are orientated sharply laterally, as in the succeeding dorsals and lumbars. This is an unusual condition. In Megantereon , Smilodon and extant felids, this transition occurs at T11. The transition is also evident in the position of the zygapophyses, in that the postzygapophyses are situated on elongate pedicels well above the level of the prezygapophyses. In extant large felids, the transition is also evident in the great reduction of the neural spine compared with both preceding and succeeding vertebrae ( Fig. 7C View Figure 7 ). In Smilodon , two morphs are present, one lacking and one possessing a well-developed neural spine on T11 ( Merriam & Stock, 1932), but the state of preservation in Megantereon does not allow detailed comparison with either.
Twelfth thoracic
Thoracic vertebra 12 is well preserved, and bears resemblance to that of other felids ( Figs 6 View Figure 6 , 9D, J, P, V View Figure 9 ). The metapophysis is well developed and rises to two-thirds of the height of the neural spine. In extant large felids, it often almost reaches the top of the neural spines. In extant felids, a wide, wing-like lamina extends from the anterior part of the centrum up along the metapophysis, and this is also the case in Megantereon . Although small portions are missing, it would appear that the lamina is slightly less well developed than in Smilodon ( Fig. 9D, J View Figure 9 ) and comparable with the condition in extant large felids. The apex of the metapophysis is distinctly thickened, as in other felids. The anapophysis is also well developed, and appears better developed than in Smilodon ( Fig. 9D, J View Figure 9 ) given that it virtually overlaps the postzygapophysis, as commonly seen in extant large felids ( Fig. 9V View Figure 9 ). Viewed dorsally, the anapophyses appear proportionally slightly more slender than in Smilodon ( Merriam & Stock, 1932) , and broadly comparable with those of extant large felids. The apex is slightly thickened and rugose, as in other felids.
As in other felids, the intervertebral notch of T 12 in Megantereon is distinctly V-shaped, and comparisons with Smilodon , Homotherium and extant felids imply that this was also the case for the notch on T11, although this section is not preserved. Unlike extant felids, the apex of the neural spine is not thickened. The parapophysis is well developed but the diapophysis is indistinct. This is also frequently the case in extant felids. In Smilodon , the facet may be present on both sides of the centrum, one side only, or absent, which appears to be the most common condition ( Merriam & Stock, 1932).
Thirteenth thoracic
The last thoracic vertebra is also well preserved and bears great resemblance to T12. In contrast to the condition in extant felids, where the centrum of T13 is often 1–2 mm longer than in T12, the two are almost the same size in Megantereon ( Table 2). As in other felids, the metapophyses are larger than on T12, as are the anapophyses. The neural spine is slightly more anteriorly inclined and the parapophysis is situated lower on the anterior face of the centrum, as in other felids. The zygapophyses, metapophyses, anapophyses and neural spine appear slightly less well developed relative to centrum length than in Smilodon ( Merriam & Stock, 1932) .
Lumbar 2–7
The first lumbar vertebra is not preserved. The transverse processes are absent in L2, L4 and L5, and in L3 the apex of the right transverse process is missing. In L6, the apex of the right transverse process is missing, whereas almost the entire process is missing from L7. The metapophyses and anapophyses are fully preserved only on L3. On L2, most of the anapophyses are not preserved, and in L4–7 the anapophyses were evidently greatly reduced, perhaps even more so than in extant felids, where they merely constitute thin prongs. In contrast, the metapophyses of Megantereon remain well developed on all lumbars, although imperfectly preserved in L4–6. This is similar to the condition in other felids.
Apart from this, the vertebrae are generally well preserved and show the same overall pattern as other felids in becoming progressively larger from L2 to L6, and then sharply decreasing in centrum length at L7 ( Table 2, Fig. 7A View Figure 7 ). The length of the centra relative to the entire presacral vertebral column, and the height of the posterior face of the centrum relative to centrum length closely follow the patterns observed in extant large felids. As in other felids, the transverse processes curve distinctly anteriorly and also become progressively larger towards the rear. In general, they appear comparably developed to the condition in Smilodon , and they are less anteriorly inclined than in extant felids, and more laterally orientated ( Fig. 6 View Figure 6 ). Overall, the lumbar transverse processes are distinctly less well developed in Megantereon , Smilodon and Homotherium relative to centrum size than in extant large felids ( Figs 7C View Figure 7 , 9F, L, R, X View Figure 9 ). The proportionally greater size of the lumbars ( Fig. 7A View Figure 7 ) and markedly anterior inclination of the transverse processes ( Fig. 9R, X View Figure 9 ) in extant large felids, however, result in their actually appearing more narrow relative to centrum size than in Megantereon ( Fig. 7D View Figure 7 ).
The neural canal is distinctly oval in the anterior lumbars, which is accentuated towards the rear, where the neural canal is almost squarish, as in other felids. As in other felids, the zygapophyses are sharply inclined to approximately 60–70°, and increase in size from L2 to L4, after which they decrease slightly again. In general, they appear less well developed than in Smilodon ( Merriam & Stock, 1932) . The height of the neural spines increases towards the rear ( Table 2), as in extant felids, but because the centra also become longer the ratio of neural spine height to centrum length changes less towards the rear. Megantereon follows the pattern observed in extant felids closely, but its neural spines appear rather short, with the exception of L6 ( Fig. 7C View Figure 7 ). As in other felids, the last lumbar is markedly shorter than L6 and has a much greater width across the postzygapophyses, providing strength and stability to the sacroiliac joint.
Sacrum
The three sacrals are firmly co-ossified to each other and to the lateral face of the iliac wings. The sacroiliac joint is very powerfully constructed (see also Schaub, 1925). The left sacroiliac joint has, however, undergone cosmetic restoration, as have parts of the surface and neural spines ( Fig. 10B View Figure 10 ). The anterior sacral connects to the ilia with very strong transverse processes, and the neural spine is missing. The prezygapophyses are angled at around 50°, comparable with the condition frequently seen in other felids. They are more anteroposteriorly elongate than in Smilodon . The sacrum rapidly tapers posteriorly and the neural spines are moderately tall, less so than in Smilodon ( Merriam & Stock, 1932) , but like the condition in Smilodon , they are more massively built than in extant large felids. Ventrally, the anterior sacral centrum is gently convex, and as in other felids, sacrals two and three develop a prominent keel. Unfortunately, the posterior part of the last sacral is weathered, and the postzygapophyses are not preserved.
No caudals are preserved, but the tail of Megantereon was presumably short, owing to the strong posterior tapering of the sacrum, more so than in extant large felids and resembling the condition in Smilodon ( Merriam & Stock, 1932) and Homotherium (Rawn- Schatzinger, 1992), and the fact that the tails also appear to have been short in these two closely related forms.
RIBS
There are a number of ribs preserved ( Fig. 11 View Figure 11 ), both from the anterior and the mid-posterior part of the ribcage, although a positive assignment of the elements is not possible, other than the first rib. The first rib was much straighter than the following and very short and heavy with a massive rugosity ventrally for the costal cartilage. Overall the ribs bear substantial resemblance to those of other felids.
SHOULDER GIRDLE AND STERNUM
The sternum is represented by the complete manubrium and four mesosternal elements ( Fig. 12C–E View Figure 12 ). No parts of the xiphisternum are preserved. The anterior, triangular portion of the manubrium was originally broken away from the posterior, almost rectangular section, but has subsequently been glued back together. The bone has not been restored. Proportionally, it bears great resemblance to that of Smilodon ( Merriam & Stock, 1932) , except that it is slightly more slender. Typically of derived machairodont felids it is, however, much more heavy in build than in extant large felids, even large tigers, and only large, older male lions occasionally have manubria that approach, although do not reach, such heavy proportions.
The dorsal surface is bevelled with a prominent median ridge extending to the extreme apex. The ventral surface is gently concave. The posterior facet for the mesosternum is cut off squarely, and the rugose facets for the first costal cartilage are large and heavy. As in Smilodon and extant felids, the apex is rugose and incompletely formed, indicating cartilaginous covering in life. At a length of 77.1 mm, it is distinctly smaller than Smilodon manubria, which range from 100 to 141 mm ( Merriam & Stock, 1932). It is comparable with those of lions (N = 7; 73.8– 101.1 mm) and tigers (N = 12; 66.9–93.6 mm), and is much larger than in leopards (N = 4; 47.6–60.7 mm) and jaguars (N = 4; 61.7–69.5 mm). The mesosternal elements resemble those of Smilodon and other large felids in being square and heavy, and in having a gently convex dorsal and gently concave ventral surface. The end facets for other sternal elements are rugose. The size (length ¥ width at middle) of the four mesosternal elements are: 35.5 ¥ 16.8 mm; 35.5 ¥ 19.0 mm; 37.7 ¥ 15.8 mm; and 38.3 ¥ 19.4 mm.
Both scapulae are present and generally well preserved, although the left scapula is the better preserved of the two ( Fig. 12A, B View Figure 12 ). The left scapula is virtually complete, although minor parts of the supraspinous and infraspinous fossae and scapular spine have been superficially restored. The dorsoventral length in straight line from the superior border to the lower margin of the glenoid fossa is 221.6 mm. The right scapula has undergone more extensive restoration, and the distal part of the scapular spine and acromion process are absent. The supraspinous fossa has been entirely restored along the anterodorsal border. As preserved, its dorsoventral length is 217.8 mm. Scapula length in Megantereon is comparable with lions (N = 17; ♀: 175.7–242.1 mm; ♂: 210.2–288.4 mm) and Bengal tigers (N = 15; ♀: 190.7–241.7 mm; ♂: 199.1–241.3 mm), and far larger than in leopards (N = 16; ♀: 123.2–150.8 mm; ♂: 144.6–170.3 mm), and jaguars (N = 8; ♀: 152.1– 158.2 mm; ♂: 163.3–172.6 mm). Scapula length in Smilodon is 266–358 mm ( Merriam & Stock, 1932), far exceeding the scapula size in even Siberian tigers (N = 3; ♂: 270.5–284.8 mm), despite Smilodon being lion-sized overall ( Kurtén & Anderson, 1980; Akersten, 1985), albeit substantially more massive in build ( Christiansen & Harris, 2005). This indicates that derived sabrecats had large scapulae for their body size.
The scapula is typically felid in overall morphology, but there are noticeable differences from both extant large felids and Smilodon . The cranial border is strongly curved, the caudal border is virtually straight from the caudal angle to the glenoid fossa and not sinusoid as in Smilodon , and the caudal angle is cut at an angle of ~104°. In all those characters, Megantereon bears a substantially closer resemblance to the condition in extant large felids than in Smilodon ( Merriam & Stock, 1932) . As in Smilodon , the spina scapula is rather straight and does not curve across the infraspinous fossa, as commonly seen in extant felids. The spina scapula reaches a maximum height of 36.9 mm in the left scapula. This is proportionally greater relative to scapula length than in Smilodon , and is similar to the proportions in Neofelis nebulosa (P = 0.378) and Panthera tigris (P = 0.517), but is less than in Puma concolor (P = 0.001), Panthera leo (P = 0.022), P. onca (P = 0.018) and P. pardus (P = 0.043). Accordingly, Megantereon has a proportionally lower spina scapula than several of the big cats ( Fig. 13A View Figure 13 ).
Megantereon has a distinct, short and very heavy metacromion process, which constitutes a difference from Smilodon , in which this process is absent ( Merriam & Stock, 1932). The acromion process is very large and heavy, distinctly more so than in Smilodon and extant large felids, indicating that the m. acromiodeltoideus ( Barone, 1967; Crouch, 1969; Done et al., 2000) was powerfully developed. The scapular notch is more clearly demarked than in Smilodon , resembling the condition in extant large felids. The supraglenoid tubercle is similar to those of other large felids. The scar for the m. biceps brachii is well developed, as in other large felids.
The scapula of Smilodon is markedly narrower than is the case in extant felids. The scapula of Megantereon is distinctly wider relative to overall length than in Smilodon , resembling the condition in extant large felids more closely. The infraspinous fossa in Megantereon appears comparably developed with the condition observed in Smilodon , but the supraspinous fossa is distinctly larger, indicating that Megantereon had more powerfully developed m. supraspinatus ( Barone, 1967; Crouch, 1969; Done et al., 2000) than did Smilodon . The scar for the m. teres major along the caudal border is distinctly larger in Megantereon than in Smilodon , and the scar for the m. triceps brachii is also large.
The above proportional differences are corroborated by analyses of fossae widths to scapula length in Megantereon and extant big cats. The infraspinous fossa is narrower relative to scapula length in Megantereon ( Fig. 13B View Figure 13 ) than in Puma concolor (P = 0.001) and all the pantherines (P <0.001), but not in Neofelis nebulosa (P = 0.062). However, the supraspinous fossa is wider relative to scapula length in Megantereon ( Fig. 13C View Figure 13 ) than in Panthera leo (P = 0.014) and P. pardus (P = 0.029), but not in P. onca (P = 0.583), P. tigris (P = 0.330), Puma concolor (P = 0.092) and Neofelis nebulosa (P = 0.971). This results in the ratio of the infraspinous to supraspinous fossa width being significantly lower in Megantereon than in Puma concolor (P = 0.014), Neofelis nebulosa (P = 0.006) and the pantherines (P <0.001). This indicates that Megantereon had a more powerfully developed m. supraspinatus relative to the m. infraspinatus than in extant big cats, and also than in Smilodon .
FORELIMB
The forelimbs are almost complete ( Fig. 14 View Figure 14 ), although several smaller elements are missing from the left carpus and the preservation is often better in the right forelimb.
Humerus
Both humeri are well preserved, especially the right one ( Fig. 14A–D View Figure 14 ), but in several places the surface of the left humerus is slightly weathered, as are parts of the distal trochlea. Humerus length ( Table 3) is within the lower size range of lions (N = 17: 222.2– 367.0 mm) and tigers (N = 19: 236.6–372.7 mm), and is far larger than in jaguars (N = 8: 195.8–238.7 mm), leopards (N = 16: 176.4–222.6 mm) and pumas (N = 6: 187.4–231.9 mm). It is also larger than in Smilodon gracilis (N = 4; 232.9–259.9 mm), but well below S. fatalis (N = 7: 352.5–381.6), and although smaller adult specimens are known (see Merriam & Stock, 1932), few appear to be as small as Megantereon . It is of course also below the values in the even larger sabrecat S. populator (N = 4; 333.5–387.5 mm), but only slightly smaller than small specimens of the otherwise large Homotherium serum ( Turner & Antón, 1997) , but as pointed out by Rawn-Schatzinger (1992), the humerus is short in this species compared with the epi- and metapodials.
As in other derived sabrecats, the humerus of Megantereon is very similar to those of extant big cats, albeit with exaggerated robustness overall. The ratio of humerus least circumference of the diaphysis relative to articular length in Megantereon (0.356; Fig. 15A View Figure 15 ) is distinctly higher than in Panthera leo (0.318 ± 0.005; P <0.001), P. onca (0.322 ± 0.008; P <0.001), P. pardus (0.290 ± 0.005; P <0.001), P. tigris (0.303 ± 0.005; P <0.001), Neofelis nebulosa (0.298 ± 0.010; P = 0.009) and Puma concolor (0.291 ± 0.009; P <0.001), and even than in Smilodon *Proximal articulating surface weathered.
Diaphysial diameters are taken at the site of circumference, which is least circumference in the humerus, but are taken at midshaft in radius and ulna.
Abbreviations: AP, anteroposterior diameter; LM, lateromedial diameter.
gracilis (0.330 ± 0.011; P = 0.047) and S. fatalis (0.337 ± 0.008; P = 0.027). Only the very robust S. populator has a higher circumference/length ratio than Megantereon (0.401 ± 0.011; P = 0.003).
The humeral head is relatively wider than in extant large felids, as in Smilodon fatalis and S. populator , and the greater tuberosity is prominent, and much lateromedially wider than in extant large felids ( Fig. 14B View Figure 14 ), even more so than is frequently the case in S. fatalis , but less than often seen in S. populator . The dorsal profile of the greater tuberosity in S. fatalis is less curved with a distinct median notch, whereas it is more evenly rounded in extant large felids. In Megantereon , the dorsal profile is also curved without a medial notch, but the tuberosity is distinctly larger than in extant large felids, and the anterior edge is less ventrally curving than is often the case in extant large felids. Viewed anteriorly ( Fig. 14A View Figure 14 ), the tuberosity gradually descends from the level of insertion for m. supraspinatus towards the deltopectoral crista, and in this respect it bears a closer resemblance to the condition in extant large felids than Smilodon fatalis , where the proximal edge is more horizontal and abruptly descends towards the deltopectoral crista. Viewed anteriorly, the greater tuberosity crest is markedly thickened, distinctly more so than in extant felids and even S. fatalis , resembling the condition in S. populator . This indicates a particularly massive insertion for the m. deltoideus acromialis and m. pectoralis ( Barone, 1967; Crouch, 1969; Done et al., 2000), congruent with scapula morphology (above).
The facet for insertion of the m. supraspinatus ( Fig. 14D View Figure 14 ) is large and appears distinctly more elongate than in extant large felids, but it is also frequently rather elongate in Smilodon fatalis and S. populator . The adjacent facets for insertion of m. infraspinatus and m. teres major are also distinctly larger than in extant large felids. The bicepital groove is large and wide, as in other large felids. The lesser tuberosity is better developed than in extant large felids ( Fig. 14A View Figure 14 ), and is proportionally larger than in some, but not all, examined specimens of Smilodon fatalis . Curiously, the tubercle is frequently also very well developed in S. gracilis , and, as expected, in S. populator . This indicates a well-developed m. subscapularis in Megantereon , in accordance with scapula morphology (above).
The deltopectoral crista is very massive in Megantereon , and its length relative to the humerus has been used to infer locomotor behaviour in mammals (see Elissamburu & Vizcaino, 2004). However, the distal extension of the crista to the length of the humerus (DLH; Fig. 16 View Figure 16 ) in Megantereon (0.549) is similar to the values in most other large felids, both extant and extinct [ Panthera leo (P = 0.054), P. onca (P = 0.544), P. pardus (P = 0.329), P. tigris (P = 0.055), Neofelis nebulosa (P = 0.416), Smilodon fatalis (P = 0.064) and S. populator (P = 0.965)], but DHL is lower in Puma concolor (P = 0.022) and Smilodon gracilis (P = 0.015). Accordingly, the deltopectoral crista does not extend further down the diaphysis than is the case in most other large felids (contra Schaub, 1925), probably for biomechanical reasons, as a distal extension provides a larger inlever and thus greater inforce, but also reduces the outvelocity and amplifies torsional moments ( Hildebrand, 1988).
The distal end of the humerus is very wide. The distal width of the humerus relative to the articular length in Megantereon (0.282; Fig. 15B View Figure 15 ) is significantly higher than in Panthera leo (0.264 ± 0.014; P <0.001), P. pardus (0.236 ± 0.012; P <0.001), P. tigris (0.266 ± 0.013; P <0.001) and Puma concolor (0.238 ± 0.023; P = 0.006), but is not significantly different from Panthera onca (0.273 ± 0.017; P = 0.197), Neofelis nebulosa (0.267 ± 0.030; P = 0.343) or S milodon gracilis (0.270 ± 0.018; P = 0.283). Distal humeral width is proportionally greater in S. fatalis (0.307 ± 0.009; P <0.001) and S. populator (0.332 ± 0.012; P = 0.003) than in Megantereon .
All felids have proportionally well-developed medial epicondyles, and, accordingly, powerfully developed m. flexor carpi radialis, digitorum profundus and superficialis, and m. pronator teres ( Barone, 1967; Crouch, 1969; Done et al., 2000), as the forelimbs are very important not only for climbing (see e.g. Taylor, 1974) but also for catching prey. Megantereon is no exception, but the medial epicondyle appears unusually well developed, distinctly more so than in even jaguars or tigers. In this respect, it bears resemblance to powerfully built specimens of Smilodon fatalis and, in particular, S. populator . The medial epicondyle is also very powerfully developed relative to extant large felids ( Fig. 14B View Figure 14 ), indicating powerful m. extensor carpi ulnaris, digitorum communis, indicis, pollicis longus and minimi digitis. Overall, muscle attachments of the entire humerus appear powerfully developed compared with even jaguars, lions and tigers, resembling the condition in Smilodon fatalis and S. populator .
The olecranon fossa is proportionally wider and lower than in extant large felids, and its lateral face is distinctly flaring, unlike the condition in extant large felids, but this is also present in some specimens of Smilodon fatalis and S. populator . The radial fossa is wide and low, as in other large felids, and appears proportionally larger than in most extant large felids owing to the great width of the distal humerus. The coronoid fossa is distinctly taller than wide, unlike the condition in extant felids, resembling the condition in Smilodon sp. There is an elongate, slit-like entepicondylar foramen, but it is distinctly smaller than usually seen in extant large felids, resembling the condition often seen in the three species of Smilodon .
The articulating facets for the antebrachium are large, and the capitulum is heavy and subcylindrical. The trochlea is narrower ( Table 3), and projects distinctly beyond the capitulum. The long axis of the trochlea forms a moderately wide angle (33.7°) to the capitulum ( Fig. 14A View Figure 14 ; see Heinrich & Rose, 1997). The angle is frequently shallower among extant large felids, but varies substantially intraspecifically, and the value in Megantereon is within the variation of extant large felids. In some extant felids, e.g. jaguars, the trochlea frequently extends only slightly ventrally to the capitulum, whereas in tigers it often extends distinctly beyond the capitulum. In all three species of Smilodon this variation is also present, although the trochlea of S. fatalis and S. populator usually extends further below the capitulum than in jaguars. There is no reason to infer differences in elbow function or limb posture in Megantereon from other large felids.
Radius
The radius differs from those of extant felids primarily in being distinctly heavier in build ( Fig. 14H, I View Figure 14 ; Table 3), a feature which it has in common with Smilodon fatalis ( Merriam & Stock, 1932) , S. gracilis ( Berta, 1987) and S. populator (our pers. obsserv.). The radial head is less flaring than in extant large felids, resembling the condition in Smilodon fatalis ( Merriam & Stock, 1932) , S. gracilis ( Berta, 1987) , S. populator and Homotherium ( Rawn-Schatzinger, 1992) . The biceps tuberosity is very prominent, distinctly more so than in Homotherium and S. gracilis , resembling the condition in S. fatalis and extant large felids and indicating a powerful m. biceps brachii ( Barone, 1967; Crouch, 1969; Done et al., 2000). The insertion scar for m. pronator teres is well developed, as in other felids, and is situated lower on the shaft than is frequently the case in tigers and lions, more resembling the condition in leopards and jaguars. The scar for the origin (not insertion, as stated in Berta, 1987) of m. abductor pollicis longus along the posterolateral side of the shaft is elongate and distinct, as in other large felids. The scar for the origin (not insertion, as stated in Berta, 1987) of m. flexor digitorum profundus is large and elongate, tapering towards the medial side of the posterior shaft, but is rather indistinct. This could be owing to slight postmortem weathering, in particular along the left radius, but the scar is frequently also indistinct in extant large felids.
The styloid process is distinctly shorter, thicker and blunter than in extant felids, resembling the condition in Smilodon fatalis ( Merriam & Stock, 1932) , and differing from the condition in S. gracilis ( Berta, 1987) and Homotherium ( Rawn-Schatzinger, 1992) . The distal articulating facet is concave with clearly demarcated edges, and is wider laterally than medially, as in other felids. The area for the m. extensor carpi radialis in Megantereon appears to have been wider than in Smilodon fatalis ( Merriam & Stock, 1932) , S. gracilis ( Berta, 1987) and Homotherium ( Rawn-Schatzinger, 1992) , but the margins of the scar are less well preserved.
Ulna
The left ulna ( Fig. 14E–G View Figure 14 ) is completely preserved but the coronoid lip for the humeral trochlea has broken away from the proximal articulating cotyle in the right ulna. As with the radius, the ulna in Megantereon is powerfully built compared with those of extant large felids, but interspecific differences are present. The ratio of anteroposterior diameter of the ulnar diaphysis at midshaft to overall length (including the olecranon process) in Megantereon (0.098; Fig. 17A View Figure 17 ) is not significantly different from the proportions in Smilodon gracilis (0.102 ± 0.009; P = 0.176), Panthera tigris (0.098 ± 0.002; P = 0.845) and P. onca (0.103 ± 0.003; P = 0.069), but is significantly higher than in P. leo (0.086 ± 0.002; P <0.001), P. pardus (0.086 ± 0.002; P = 0.001), Neofelis nebulosa (0.085 ± 0.004; P = 0.024) and Puma concolor (0.081 ± 0.004; P = 0.001). Smilodon fatalis has a significantly thicker ulna than Megantereon (0.112 ± 0.003; P = 0.026), but despite S. populator having the most massive ulnae of all (0.123 ± 0.006), low sample size (N = 2) prevents assumptions of significance from Megantereon (P = 0.412). The single included specimen of Machairodus giganteus was evidently extremely large, but has a markedly thinner ulna (0.088), and proportionally appears more similar to Homotherium ( Rawn-Schatzinger, 1992) or a lion (see also Turner & Antón, 1997).
The olecranon process of Megantereon is well developed, but not unusually so, and the posterior face of the ulna is markedly straight, more so than often seen in extant felids. The length of the olecranon process from the centre of rotation in the cotyle to the tip compared with total ulna length in Megantereon (0.209; Fig. 17B View Figure 17 ) is not significantly different from the averages in Smilodon gracilis (0.221 ± 0.009; P = 0.144), S. populator (0.220 ± 0.009; P = 0.642), Panthera onca (0.208 ± 0.004; P = 0.803),and P. tigris (0.211 ± 0.003; P = 0.672), and is significantly higher than in P. leo (0.191 ± 0.003; P <0.001), P. pardus (0.180 ± 0.003; P <0.001), Neofelis nebulosa (0.176 ± 0.005; P <0.001), and Puma concolor (0.183 ± 0.005; P <0.001), but lower than in Smilodon fatalis (0.235 ± 0.005; P = 0.003). The single specimen of Machairodus giganteus again appears very gracile with a proportionally small olecranon (0.163), resembling Homotherium ( Rawn-Schatzinger, 1992) . Although the olecranon is not markedly elongate, its transverse width, markedly greater than in extant felids, Machairodus and Homotherium , and similar to Smilodon , indicates that it constituted a very large area for insertion of m. triceps and was designed for withstanding substantial moments about the joint.
In Megantereon , the olecranon process is relatively straight, whereas it is angled somewhat more posteriorly in Smilodon gracilis ( Fig. 18 View Figure 18 ). The angle for the insertion of m. triceps, in this analysis the inferred centroid of the muscular inforce, is also rather low in Megantereon , albeit close to the angles in the other species. Van Valkenburgh (1987) found that olecranon angle to the long axis of the diaphysis was greater in terrestrial than scansorial and arboreal species, and that the angle increases with body size. However, the included felids are all large animals, and it is doubtful if Megantereon was scansorial or arboreal to any greater extent than extant big cats. Rather, the differences are most likely attributed to intraspecific variation; among extant large felids the olecranon process angle to the long axis of the ulna often shows substantial intraspecific variation.
The radial notch is deeply concave, as in other felids, and on the left ulna the wide coronoid lip for the humeral trochlea is completely preserved. It is less medially flaring than is frequently the case in extant large felids, and is less distally orientated than in extant large felids, more closely resembling the condition in Smilodon , especially S. gracilis ( Berta, 1987) . The anconeal process is well developed and deeply U-shaped, as in other felids, but unlike the condition in extant large felids, the medial part is not markedly inclined with respect to the lateral part. In this latter respect, Megantereon resembles all three species of Smilodon . The scar for the m. brachialis and m. biceps brachii is very well developed, as in other large felids, but the insertion of the m. anconeus and origin of m. flexor carpi ulnaris ( Barone, 1967; Crouch, 1969; Done et al., 2000) are indistinct, as is frequently the case in extant large felids. The styloid process is rather long and massive, and curves gently anteriorly.
Manus
The right manus is virtually complete and well preserved ( Figs 14J, K View Figure 14 , 19 View Figure 19 , 20 View Figure 20 ), with only some cosmetic restoration of the proximal part of the palmar surface of metacarpal III. The right manus consists of the full carpus, metacaparpals I–V, and the proximal phalanx of digit I, the proximal and middle phalanx and claw sheath of digit II, the complete proximal and incomplete distal phalanx and claw sheath of digit III, the proximal and distal phalanx of digit IV and the claw sheath of digit V. The left manus is less complete, consisting of the trapezoid, scapholunar and unciform, and metacarpals II–V, which are also well preserved with only cosmetic restoration in a few places. Note that in Schaub (1925: fig. 4) the articulated left manus (‘Manus sin.’ [= sinistre]) is shown in its entity, albeit excluding phalanges, but this is probably the left manus with missing elements restored, based on the preserved elements of the right manus
Scapholunar
The scapholunar (or Os scaphoideum and Os lunatum; or radiale and intermedium) is similar to those in other large felids. In proximal view ( Fig. 19B View Figure 19 ) the large, squarish and gently convex articular facet for the radius takes up most of the surface, and it is wider than in extant large felids, resembling the condition in Smilodon ( Merriam & Stock, 1932; Berta, 1987). The proximal process is large with a large, lateral sesamoid facet, and curves more upwards than in Smilodon fatalis ( Merriam & Stock, 1932) , resembling S. gracilis ( Berta, 1987) and extant large felids. The process is distinctly larger than in Homotherium ( Rawn-Schatzinger, 1992) . In distal view ( Fig. 19A View Figure 19 ), the large articulating facets for the distal carpals extend across the entire corpus. Towards the ulnar side, the elongate facet for the unciform is distinctly wider than in extant large felids, and the facet for articulation with the magnum is proximally wider and tapers less distally than in extant large felids. The adjacent facets for the trapezoid distally and trapezium dorsally are broadly similar to those of Smilodon fatalis ( Merriam & Stock, 1932) , and the trapezoid facet is wider than in S. gracilis ( Berta, 1987) . The bevelled ridge separating the two facets is less demarked than in extant felids.
Cuneiform
The cuneiform (or ulnare; or Os triquetrum) in Megantereon resembles that of other large felids, but is slightly more triangular in outline than in extant large felids and also than Smilodon ( Merriam & Stock, 1932) , owing to a large lip, but this is frequently also present in Smilodon and in extant large felids, albeit usually more pronounced in tigers than in lions. In medial view ( Fig. 19C View Figure 19 ), the bowl-shaped articulating facet for the unciform is apparent, and it is distinctly oval in outline, as opposed to a more triangular outline, as in Smilodon , or a rectangular outline, as often seen in lions. Among extant felids the facet is, however, morphologically quite variable. In distal view ( Fig. 19D View Figure 19 ), the two large articulating facets for the pisiform and ulna are visible. As in other felids, they are separated by a raised ridge and slope gently towards the sides, giving the distal face a triangular profile.
Pisiform
The pisiform (or Os pisiforme) of Megantereon is a long, stout bone, as in other large felids, but appears heavier in build than in leopards and jaguars, more resembling the condition in lions. It appears proportionally longer than in Smilodon and Homotherium ( Rawn-Schatzinger, 1992) . In distal view ( Fig. 19E View Figure 19 ) the compact build of the bone is readily apparent. The facet for the cuneiform on the radial face is narrower than in extant tigers and lions, more resembling the condition often seen in leopards, and is less keyholeshaped than in Smilodon . The articulating facet for the ulna on the ulnar face is markedly less distally directed and flaring than in extant big cats, resembling the condition in Smilodon ( Merriam & Stock, 1932) . The proximal face of the bone ( Fig. 19F View Figure 19 ) is unfortunately somewhat weathered.
Trapezium
The trapezium (or distal carpal I; or Os trapezium) is very stout, more so than even in Smilodon and large male lions, and the outer edge is more rounded and less triangular than in Smilodon , Panthera atrox ( Merriam & Stock, 1932) and extant lions, resembling the frequent condition in tigers and jaguars. In Megantereon and Smilodon ( Merriam & Stock, 1932) , the outer facets for the trapezoid and metacarpal II are largely confluent, and the metacarpal facet is elongate. As figured in Merriam & Stock (1932), the trapezoid facet is relatively wider in Panthera atrox , but in extant lions and tigers it is usually more elongate and rectangular, whereas the metacarpal II facet is oval and sometimes almost round, and thus very different from the condition in P. atrox . Facet morphology is, however, subject to substantial intraspecific variation among extant large felids. In inner distal view ( Fig. 19G View Figure 19 ), the large facet for metacarpal I is distinct, and it bears great resemblance to those of other large felids. In proximal view ( Fig. 19H View Figure 19 ) are two large, rectangular–oval facets for the scapholunar. In Smilodon ( Merriam & Stock, 1932) , the facets are confluent and deeply U-shaped, but as also noted by these authors, the facets are very variable in P. atrox (compare their fig. 80D, H). This is also the case in extant large felids. In lions, there may be two large, oval–triangular facets, or a U-shaped facet, as figured for Smilodon in Merriam & Stock (1932), whereas tigers and jaguars often have just one, saddle-shaped, rectangular facet.
Trapezoid
The trapezoid (or distal carpal II; or Os trapezoideum) is distinctly triangular in outline, as in Smilodon . Proximally ( Fig. 19I View Figure 19 ), the saddle-shaped facet for the scapholunar takes up most of the surface, as in other felids. In Megantereon , the deeply concave portion of the facet is more pentagonal than in Smilodon ( Merriam & Stock, 1932) , whereas the smaller, weakly concave portion is oval–triangular, as in Smilodon and extant large felids. In Panthera atrox ( Merriam & Stock, 1932) and extant pantherines, the deeply concave portion is more rectangular, and the weakly concave portion is proportionally larger than in Megantereon and Smilodon . Along the outer rim, the facet for the magnum is visible in Megantereon , but not in Smilodon . Occasionally, this is also the case in some lions, leopards, jaguars and tigers, although not as markedly as in Megantereon . The magnum facet is, however, more exposed in distal perspective, but owing to it being twisted about its long axis, it can occasionally extend onto the dorsal face also. In distal view ( Fig. 19J View Figure 19 ), the facet for metacarpal II is triangular, as in other large felids, and as in Smilodon ( Merriam & Stock, 1932) , there is a notch in the facet. In Panthera atrox and extant pantherines the notch is either absent, or present but fairly small, as often seen in tigers and jaguars. Along the edge, the dis- tally turned portion of the magnum facet is visible, and it is relatively larger than most often seen in extant pantherines. In inner view, the facet for the trapezium is subrectangular, resembling the condition in Smilodon , whereas it is often triangular and S-shaped in pantherines.
Magnum
The magnum (or distal carpal III; or Os capitatum) is very similar to those of extant large felids, and is robust, more so than in leopards, and more comparable with those of lions, jaguars and tigers. The inner side ( Fig. 19K View Figure 19 ) is unfortunately slightly weathered, but the large round to drop-shaped facet for metacarpal II is similar in proportion and placement along the palmar side to those in other felids. The ligament pit is proportionally larger than often seen in extant felids. The facet for the trapezoid is very similar to other large felids, in being a raised, rectangular ridge, which is gently concave on top. The facet for the scapholunar extends along the proximal edge, and is fairly narrow, as in extant large felids. The facet for metacarpal III extends as a narrow, concave band along the distal edge. Along the outer side ( Fig. 19L View Figure 19 ), the facet for the unciform is, unfortunately, not well preserved, but is appears to have been large, as in other large felids. Along the distal edge, the facet for metacarpal III flares more dorsally than is the case in Smilodon ( Merriam & Stock, 1932) . In Panthera atrox ( Merriam & Stock, 1932) and extant pantherines, the facet is virtually unexposed along the outer face of the magnum, and is large, rectangular and distinctly concave in distal view. In Megantereon , it is more bowl-shaped, and the inner part of the facet is offset at a steeper angle to the distal portion of the facet.
Unciform
The unciform (or distal carpal 4 and 5; or Os hamatum) also bears a close resemblance to those of other large felids. In medial view ( Fig. 19M View Figure 19 ), the facet for the magnum is wider than in Smilodon ( Merriam & Stock, 1932) and lions and tigers, but some jaguars also have a wide facet. As in Smilodon and Homotherium ( Rawn-Schatzinger, 1992) , the facet for the scapholunar extends well onto the medial side, unlike the condition in extant large felids. In Megantereon and Smilodon , the lower edge of the facet is curved, whereas it is straighter in Homotherium (Rawn- Schatzinger, 1992) and extant large felids. In lateral view ( Fig. 19N View Figure 19 ), the facet for the cuneiform is very wide, distinctly more so than in Smilodon ( Merriam & Stock, 1932) , Homotherium ( Rawn-Schatzinger, 1992) and extant felids. The gently concave ligament fossa is very similar to that of Smilodon with a downward-curving, distal portion, and is not square as in Homotherium and extant felids. As in Smilodon and Homotherium , the fossa is much larger than in extant big cats.
Metacarpals
The relative length of the metacarpals is I–V–II–IV– III ( Table 3). In extant felids, metacarpal IV is frequently slightly longer than II, as is the case in Smilodon ( Merriam & Stock, 1932) , but in Homotherium serum metacarpal IV is the longest (Rawn- Schatzinger, 1992), an unusual condition among felids (the notion by Rawn-Schatzinger that metacarpal IV is longest in felids in general is incorrect). In Homotherium latidens metacarpal III is also the longest, however ( Ballesio, 1963). The metacarpals of Megantereon are distinctly heavier in overall build than in extant large felids and Homotherium , although not quite as robust as in Smilodon . At a greatest length of around 89 mm, metacarpal III approaches the low size range of lions (N = 15: 89.3–125.7 mm) and tigers (N = 19: 88.8–124.3 mm), and is distinctly larger than in jaguars (N = 8: 59.4–75.9 mm) and leopards (N = 16: 57.6–80.1 mm). It is comparable with the upper size range in pumas (N = 6: 71.3–85.9 mm), which, however, have much more slender metacarpals than Megantereon . Metacarpal III length in Megantereon is also far less than in Homotherium ( Rawn-Schatzinger, 1992: table 18: 112.0– 127.5 mm). All metacarpals in Megantereon are, however, well within the size ranges of Smilodon fatalis ( Merriam & Stock, 1932) , indicating that Megantereon had proportionally longer metacarpals than did Smilodon , as the latter was a distinctly larger cat ( Turner & Antón, 1997; Christiansen & Harris, 2005).
Metacarpal I bears substantial resemblance to that of Smilodon ( Merriam & Stock, 1932) , and is much more robust than in extant large felids and Homotherium ( Rawn-Schatzinger, 1992) . The dorsal facet for the trapezium is wide and appears slightly less concave than in Smilodon and extant large felids, and extends more distally than in either. As in Smilodon , there is a distinct dorsal ridge extending obliquely from the medial border of the trapezium facet, which is absent in extant felids. The distal tubercle is well developed, and the articulating facet for the first phalanx is wider and slightly heavier than in Smilodon , and distinctly larger than in Homotherium and extant big cats. The median keel of the articulation facet is also well developed.
Metacarpal II is the thickest metacarpal, and has a less triangular articulating facet for the trapezoid than in extant felids, as in Smilodon ( Merriam & Stock, 1932; Berta, 1987), but the facet is concave, like in those taxa, and unlike the condition in Homotherium ( Rawn-Schatzinger, 1992) . As in other felids, there are heavy, rugose areas on the palmar face for insertion of the flexor musculature, and the projection for articulation with metacarpal III is heavier than in extant large felids. The proximomedial tubercle for the m. extensor carpi radialis is more pronounced than in extant felids, and the adjacent, trench-like groove for the radial artery is very pronounced, more so than in extant felids, Smilodon ( Merriam & Stock, 1932; Berta, 1987) and Homotherium ( Rawn-Schatzinger, 1992) . The distal articulating facet for the proximal phalanx is medially asymmetrical and bears a very pronounced keel on the palmar surface, as in other felids.
Metacarpal III has a distinctly more pronounced articulating facet for metacarpal IV than does Smilodon ( Merriam & Stock, 1932; Berta, 1987), Homotherium ( Rawn-Schatzinger, 1992) and extant large felids, indicating a strong unison of the two. The medial facet for metacarpal II is less steeply inclined than in Smilodon and extant felids, and is not as dorsoventrally long either. The proximal articulating facet for the magnum is deeply concave with a gently raised median ridge. The proximolateral facet for the unciform is less triangular and more elongate than in extant large felids. The dorsopalmar, rugose area for insertion of the flexor muscles is less raised and triangular than is often the case in extant felids. The distal articulating facet bears a strong keel, and the facet is bulbous and taller than in extant felids and Smilodon .
Metacarpal IV is decidedly more slender than II and III. The articulating facet for metacarpal III is markedly more dorsally situated than is the case in Smilodon ( Merriam & Stock, 1932; Berta, 1987), Homotherium ( Rawn-Schatzinger, 1992) and extant large felids, probably owing to the pronounced facet on metacarpal III. The lateral articulating facet for metacarpal V is similar to that of Smilodon , and less smoothly rounded than in extant large felids. The proximal articulating facet for the unciform is smoothly convex, as in other felids, and unlike the condition in Smilodon , the facet forms a less steep angle with the metacarpal III facet, but is not as continuous with it as is the case in extant felids. The ventral, dorsal edge of the facet is distinctly more raised than in extant felids. Berta (1987) states that the posterior side of the proximal end of metacarpal IV not being broad is a characteristic of Smilodon , but this is also present in Megantereon . As in other felids, there is an elongate, rugose scar on the lateral face for insertion of the m. interosseus.
Metacarpal V has a less well-developed medial articulating tubercle for metacarpal IV than in extant large felids, resembling the condition in Smilodon ( Merriam & Stock, 1932; Berta, 1987). The convex articulating facet for the unciform is triangular in dorsal view, but is not as broad towards the palmar surface compared with the dorsal surface as in extant felids. The proximal, lateral facets for the m. extensor carpi ulnaris are more raised than in extant felids. As in other felids, the distal articulating facet is markedly medially skewed, and the keel is only pronounced on the palmar face of the facet.
Phalanges
The phalanges of the right hand are almost completely preserved ( Fig. 20 View Figure 20 ), albeit lacking digit V. Morphologically, they bear great resemblance to those of extant felids. The most noticeable feature lies in the great development of the terminal phalanges, indicating that Megantereon had very powerfully developed claws on the forelimbs. We could not, however, affirm that the terminal phalanges had been correctly assigned to their respective digits, as the articulating facet of the proposed phalanx II of digit I was weathered, although their traditional placement appears likely (see below). Phalanx III of digit II appeared, however, to fit well onto the distal articulating facet of phalanx II.
The proximal phalanx of digit I is very similar to that of Smilodon fatalis ( Merriam & Stock, 1932) , and is very short, thick and heavy in build. The median ridge of the proximal articulating facet for the metacarpal is indistinct, as in extant large felids, and unlike Smilodon , where it is better developed. The distal articulating facet is wide and subcylindrical, and is wider dorsally than palmarly, as in other felids. The asymmetrical articulating facets in metacarpal I and the proximal phalanx are similar to those of extant felids, indicating that the digit functioned in a similar manner. The terminal phalanx assigned to this digit, although weathered, is probably the right one, as it is the largest of the terminal phalanges, as in other sabrecats where articulated mani have been recovered, such as Smilodon ( Méndez-Alzola, 1941; Cox & Jefferson, 1988), Homotherium ( Antón, Galobart & Turner, 2005) , Machairodus ( Gaudry, 1862) and Lokotunailurus ( Antón, 2003; Werdelin, 2003). Sabrecats, accordingly, appear to have had very powerfully developed dewclaws.
Phalanges I are present in digits II–IV, and are morphologically very similar to those of extant large felids. Proximally, the articulating facet is deeply concave with a distinct median notch along the raised palmar face, and distally the facets are distinctly ginglymoid. This strongly indicates that motion was restricted to a dorso-palmar plane, lacking significant mediolateral or torsional components, and that the phalanges were built to withstand substantial force.
Two phalanges II from digits II and IV are completely preserved, and the majority, lacking only the proximal end, is preserved in digit III. Noticeable differences from extant felids are the very massive proportions of the phalanges, the pronounced development of the ligament scars on the proximal part of the dorsal surface, and the less pronounced dorsal lip of the proximal articulating surface. Significantly, the very wide distal articulating surfaces are strongly laterally skewed and asymmetrical, in that their lateral parts are ventrally turned compared with the medial part. This morphology is very similar to that of extant felids, and is a key feature in the claw retraction mechanism ( Gonyea & Ashworth, 1975), strongly suggesting that this was also present in Megantereon .
PELVIS
The pelvis ( Fig. 10 View Figure 10 ) is well preserved and only the cranial ventral iliac spine on the right-hand side is missing. The area around the ischiadic tubers appears weathered, albeit only mildly so. The pelvis has undergone cosmetic restoration in relatively few places. More extensive restoration has taken place along the dorsal face of the left sacroiliac joint, the left tuber coxae and parts of the iliac crest, and the left ischium posterior to the acetabulum. As those parts are preserved in the right-hand side of the pelvis, this side had been used for modelling of missing parts. As preserved, the greatest length from the iliac crest to the ischiadic tuber is 260.8 mm, and external width is 167.7 mm (across tuber sacrale), 117.3 mm (across the acetabula) and 115.1 mm (across the ischiadic tubers). Pelvic length in Megantereon is comparable with the lower range in lions (N = 11; 245.4–333.2 mm) and tigers (N = 13; 256.5– 345.3 mm) and is distinctly larger than in leopards (N = 8; 165.4–218.2 mm) and jaguars (N = 7; 200.5– 226.9 mm). It is markedly less than in Smilodon fatalis (283–368 mm; Merriam & Stock, 1932), and Homotherium (322 mm; Rawn-Schatzinger, 1992).
The pelvis is typical for a large felid, but differs in several respects from those of extant felids. One difference lies in the overall proportions, as the pelvis of Megantereon is proportionally wider than in Smilodon ( Merriam & Stock, 1932) and extant large felids. Overall, the pelvis also appears strongly built. The iliac wings are elongate and concave externally, as in other felids, and are longer compared with the puboischiadic part of the pelvis than in Smilodon , and the anterior iliac crest is less pointed than in Smilodon ( Merriam & Stock, 1932) , resembling the condition in extant large felids more closely. The cranial dorsal iliac spine appears slightly more flaring than often seen in extant large felids, resembling the condition in Smilodon . The caudodorsal iliac spine is distinctly less clearly demarcated than in Smilodon , again resembling the condition in extant large felids. Medially, the iliac crest for the last sacral is less well developed than often seen in extant large felids. The iliac tubercle is well developed, elongate and somewhat crest-like, resembling the condition in extant large felids more closely than in Smilodon , where it is wider and thicker ( Merriam & Stock, 1932).
The lesser ischiadic notch is indistinct, and the ischiadic spine is less distinct than is often the case in extant large felids, resembling the condition in Smilodon ( Merriam & Stock, 1932) . Despite slight weathering, the ischiadic tubers were evidently not nearly as posteriorly inclined as in Smilodon and extant felids, resulting in the ischiadic arch being only gently concave ( Fig. 10B View Figure 10 ), an unusual condition in a large felid, and not deeply U-shaped, as in Homotherium ( Rawn-Schatzinger, 1992) , or V- shaped, as in Smilodon ( Merriam & Stock, 1932) and extant felids. The ischiadic tuberosity is well developed, extending down the ischiadic arc, as in other felids.
The pubic pecten (brim) is straighter and shallower than in extant large felids, resembling the condition in Smilodon . The main difference from extant felids lies in the proportional length of the pubic symphysis. As in Smilodon ( Merriam & Stock, 1932) and Homotherium ( Rawn-Schatzinger, 1992) , the symphysis in Megantereon is proportionally markedly shorter than in extant large felids, and the anterior face of the pubic brim terminates at around mid-level of the acetabulum, whereas in extant felids it terminates around the anterior rim of the acetabulum. The symphysis is also more anteriorly rounded than in extant felids, resembling the condition in Smilodon ( Merriam & Stock, 1932) . The obturator foramen is more elongate than in Smilodon and extant felids, although its morphology frequently varies, and measures 50.2 ¥ 20.0 mm (length ¥ height) on the left-hand side. The acetabulum is deeply concave and almost cylindrical, measuring 35.5 ¥ 35.3 mm (length ¥ height).
HIND LIMBS
Both hind limbs are largely complete, albeit without the fibulae, most of the smaller tarsal bones and phalanges. The hind limb long bones generally appear less heavy in build compared with those of extant large felids than was the case in the forelimb (but see below), in particular the humerus and radius. The hind limb appears to have been more gracile than the forelimb, and the metapodials proportionally rather elongate and slender. A right patella was present in the mounted skeleton, but could not be found. No evidence of fabellae is present.
Femur
Both femora are present and are generally well preserved. The right femur ( Fig. 21A–D View Figure 21 ) has had cos- metic restoration around the femoral head and along the anterolateral face around the middle of the diaphysis. The medial condyle is missing, but has been restored, albeit not to a credible morphology. The left femur has had more extensive restoration around the femoral neck, and the femoral head and distal condyles are weathered, and cosmetic restoration has been carried out at several places along the diaphysis and distal condyles. Femoral length in Megantereon ( Table 4) is within the lower size range of lions (N = 17: 262.5–402.5 mm) and tigers (N = 19: 272.3– 429.6 mm), and is larger than in jaguars (N = 8: 229.8–265.5 mm), leopards (N = 16: 200.9–255.3 mm) and pumas (N = 6: 226.3–274.3 mm). It is well below the values in Smilodon fatalis (N = 7: 370.5–423.5), although distinctly smaller adult specimens have been found than are present in our sample (see *Distal end of tuber restored.
Diaphysial diameters are taken at the site of circumference, which is least circumference in the femur, but is taken at midshaft in the tibia.
Abbreviations: AP, anteroposterior diameter; LM, lateromedial diameter; r, restored. A plus sign indicates weathering, and that actual measurements should have been greater.
Merriam & Stock, 1932), S. populator (N = 2; 390.2 and 395.0 mm; see Christiansen & Harris, 2005) and Homotherium serum (N = 5; 323.3–349.7 mm; Rawn- Schatzinger, 1992: table 23). The single specimen of S. gracilis available for this study (UF82652) is smaller than Megantereon (269.1 mm).
The femora are solid in build compared with those of extant large felids, and the least circumference of the femoral diaphysis to articular length of the femur in Megantereon (0.289; Fig. 22 View Figure 22 ) is significantly higher than in Panthera leo (0.271 ± 0.004; P <0.001), P. onca (0.263 ± 0.006; P = 0.002), P. pardus (0.246 ±
No known copyright restrictions apply. See Agosti, D., Egloff, W., 2009. Taxonomic information exchange and copyright: the Plazi approach. BMC Research Notes 2009, 2:53 for further explanation.