taxonID	type	description	language	source
CB5187EBFFB6FF894D04FD53BF52A004.taxon	description	Machaeroidines are known from lower to middle Eocene deposits of North America (Fig. 1). They differ dentally from carnivorans in having the second lower molar (rather than the first) modified into a carnassial (see Gunnell, 1998). Only two machaeroidine specimens have mandibles sufficiently complete for detailed analysis: the small Machaeroides eothen (USNM 17059; see Gazin, 1946) and the larger Apataelurus kayi (CM 11920; see Scott, 1938). Even though the mandibular flange of Apataelurus kayi is only partially preserved, the mandibular symphysis is complete and allows for an accurate determination of mandibular depth at the canine. Additionally, partial mandibles for Machaeroides eothen (USNM 361372) and the slightly more derived and younger Machaeroides simpsoni (CM 36397; see Dawson et al., 1986) were studied. Lacking the postcarnassial segment of the ramus, mandibular force profiles could not be evaluated for these specimens because accurate determination of the moment arm length for each interdental gap is impossible. However, the bending strengths at each interdental gap, Zx and Zy, and relative mandibular force (Zx / Zy) could be calculated and compared to those of the complete specimen of M. eothen. By doing so, intrageneric variability can be evaluated, providing insight into possible phylogenetic trends. Mandibular force profiles Although Machaeroides eothen and Apataelurus kayi have slightly different mandibular force profiles, both exhibit a negative slope in dorsoventral (Zx / L) and labiolingual (Zy / L) forces between M 1 M 2 and P 3 P 4 and a very strong symphyseal region, much stronger than the corpus at the carnassial (Fig. 5 A). The peak in dorsoventral force at P 4 M 1 in M. eothen is an artefact produced by the anomalous expansion of the dorsal margin of the mandibular corpus into the interdental gap between the right P 4 and M 1 in USNM 17059, artificially increasing mandibular depth at that interdental gap. Comparison of the dorsoventral and labiolingual bending strengths (Zx and Zy) for the two specimens of M. eothen and M. simpsoni confirms the anomalous mandibular properties at P 4 M 1 in USNM 17059 (Fig. 5 B): while the dorsoventral bending strength of the mandibular corpus varies little along the tooth row in the incomplete specimens of M. eothen and M. simpsoni, the Zx values at P 4 M 1 are much higher in USNM 17059. This anomaly is not as apparent labiolingually, where the bending strength values increase slightly anteriorly in both specimens of M. eothen; in contrast, the bending strength values decrease slightly anteriorly in M. simpsoni (Fig. 5 B). The relative mandibular force (Zx / Zy) profiles of all machaeroidine specimens studied (Fig. 5 A – B) can be compared. Omitting the exaggerated P 4 M 1 value in USNM 17059, the Zx / Zy profiles of M. eothen and A. kayi show a slow, linear decrease anteriorly between the carnassial and the canine. In contrast, a gradual increase in Zx / Zy values is observed in M. simpsoni. All machaeroidines attain minimum Zx / Zy values at the canine: Zx / Zy canine values for M. eothen vary between 1.33 and 1.60 (mean of 1.46 when mandibular dimensions are averaged prior to determination of section moduli), Zx / Zy canine = 2.15 for M. simpsoni, and Zx / Zy canine = 1.48 for Apataelurus kayi. Interpretation The dorsoventral and labiolingual force profiles of machaeroidines give insight into their bite force (Table 1). At the carnassial (M 1 M 2), the bite force of Apataelurus kayi was approximately 3.50 times greater than that of the smaller Machaeroides eothen. By comparing these results with those obtained for extant taxa, A. kayi was likely able to generate bite forces comparable to those of Panthera pardus, a felid of similar mandibular length. Although felids of mandibular length similar to M. eothen were not studied by Therrien (2005), the bite force of the small machaeroidine appears to have been half that of Eusmilus cerebralis, a nimravid of similar mandibular length, one third that of Puma concolor, a felid with a mandible 50 % longer, and close to that of Canis latrans, a canid with a mandible 63 % longer. The dorsoventral and labiolingual force profiles (Zx / L and Zy / L, respectively) also indicate that the mandible of machaeroidines is stronger at the canine than at the cheek teeth, suggesting that they delivered powerful canine bites to kill their prey, like felids, rather than shallow bites like canids and hyaenids A Dorsoventral force Labiolingual force Relative force (Log Zx / L) (Log Zy / L) (Zx / Zy) Machaeroides eothen (L = 8.95 cm) - 0.47 2.05 1.48 Apataelurus kayi (L = 14.59 cm) (Fig. 4). The close correspondence of the section modulus values of the incomplete M. eothen and M. simpsoni specimens to those of the complete M. eothen specimen reveals that the killing behaviour of these individuals must have been similar. The relative mandibular force (Zx / Zy) profiles reveal that the mandible of machaeroidines, with the exception of M. simpsoni, becomes rounder from M 1 M 2 toward the canine (Fig. 5). These decreasing Zx / Zy values indicate that torsional stresses are more important toward the canine, most likely in response to prey capture. Machaeroides simpsoni is unique among machaeroidines in having a mandibular corpus that becomes deeper anteriorly between M 1 M 2 and P 3 P 4 (Fig. 5). Although none of the extant carnivorans studied by Therrien (2005) displayed this feature, the fact that some nimravids and machairodonts exhibit a similar trend in mandibular shape (see below) indicates that this feature evolved independently in different sabretooth lineages. The Zx / Zy profiles reveal that the mandible of the two Machaeroides species is better dorsoventrally buttressed than that of A. kayi. In other words, labiolingual stresses are relatively more important in the larger machaeroidine species than in the smaller species. In fact, whereas the Zx / Zy profile near the cheek teeth of A. kayi is within the range of extant felids, those of Machaeroides are slightly higher (Figs 4, 5). This difference could potentially reflect the fact that small machaeroidines captured small prey, which would have induced significantly less torsional stresses than larger prey, thus allowing the mandible to be better dorsoventrally buttressed than that of Apataelurus. However, beam models for extant carnivorans of small body size, similar to Machaeroides, and feeding on small prey need to be constructed to test these hypotheses. The Zx / Zy canine values of M. eothen and A. kayi are similar to those of extant conical-toothed felids with special killing techniques (Fig. 4). From these mandibular properties, it can be surmised that machaeroidines delivered canine bites on prey that offered relatively less resistance than those of extant, large, conical-toothed felids. Thus, prey must have been at least partially restrained before the sabre-like canines were used to kill. Although postcranial remains for A. kayi are not known, the humerus of M. eothen (USNM 17059) is robust and possesses a large deltopectoral crest reminiscent of that observed in dirk-toothed sabretooth ecomorphs (Gazin, 1946; pers. observ.). Although Anyonge (1996) did not include machaeroidines in his study, he demonstrated that dirk-toothed ecomorphs had extremely powerful forelimbs. This interpretation is supported by the brachial index (maximum length of radius / maximum length of humerus) of M. eothen (BI = 0.76, based on measurements in Gazin, 1946), which is lower than that of Smilodon but slightly higher than those of Barbourofelis lovoerum and B. fricki (see below). Given the robust postcranium of machaeroidines, sabretoothed creodonts could have been able to at least partially immobilize prey with their forelimbs prior to delivering a canine bite. Finally, the fact that M. eothen and A. kayi possess lower Zx / Zy canine values than later sabretooths (possibly with the exception of Nimravus brachyops, see below) suggests that machaeroidines may have been less efficient at restraining their prey than later sabretooths (see Discussion). Machaeroides simpsoni contrasts with other machaeroidines in having a high Zx / Zy canine value (2.16; Fig. 5 B). Such a value, although within the range of machairodonts (Figs 9, 10), is still low compared to other flanged sabretooths (Figs 6, 7, 8, 11). Nevertheless, it does indicate that stresses were better constrained dorsoventrally in the symphysis when M. simpsoni delivered its canine bite than in other machaeroidines. Thus, M. simpsoni may have been able to immobilize its prey more efficiently than other machaeroidines prior to delivering canine bites.	en	Therrien, François (2005): Feeding behaviour and bite force of sabretoothed predators. Zoological Journal of the Linnean Society 145 (3): 393-426, DOI: 10.1111/j.1096-3642.2005.00194.x, URL: https://academic.oup.com/zoolinnean/article-lookup/doi/10.1111/j.1096-3642.2005.00194.x
CB5187EBFFB4FF8D4FC1FC00B83CA1FC.taxon	description	Mandibular force profiles Among nimravids, two distinct patterns, corresponding to the scimitar-toothed and dirk-toothed ecomorphs, emerge from the dorsoventral and labiolingual mandibular force profiles (Figs 6, 7). Although members of both ecomorphs exhibit a decrease in dorsoventral and labiolingual force values between post-M 1 and P 3 P 4, they differ mainly by their mandibular crosssectional properties at the canine. Dirk-toothed nimravids are characterized by a symphyseal region slightly stronger dorsoventrally than the corpus at the carnassial, the symphysis being much stronger than at the premolars (P 3 P 4; Fig. 6). Labiolingually, the symphyseal region is still stronger than at the premolars but is weaker than at the carnassial, except in Eusmilus cerebralis. In contrast, the symphyseal region of scimitar-toothed nimravids is much weaker, both dorsoventrally and labiolingually, than the corpus at the carnassial (Fig. 7). The symphysis is only slightly stronger than at the premolars (P 3 P 4), a condition reminiscent of that observed in extant canids (Fig. 4) (Therrien, 2005). A smaller, presumably less mature Dinictis felina individual (YPM 13587; see Bryant, 1996 b) is characterized by a symphyseal region weaker than the corpus at P 3 P 4. Finally, N. brachyops differs from other scimitar-toothed sabretooths in having higher postcarnassial Zx / L and Zy / L - values and a labiolingually stronger symphyseal region. The ecomorph dichotomy among nimravids is also recognizable in the relative mandibular force profiles (Zx / Zy, Figs 6, 7). Whereas the shape of the mandibular ramus tends to remain the same among scimitar-toothed nimravids (Zx / Zy = 2.00), it deepens anteriorly in dirk-toothed nimravids, as indicated by the increasing Zx / Zy values. Although the symphyseal region in both ecomorphs is strongly buttressed against dorsoventral loads, having minimal Zx / Zy canine values of 2.37 (excluding N. brachyops), dirk-toothed nimravids have significantly higher Zx / Zy canine values than scimitar-toothed nimravids (P <0.006, P <0.04 without Nimravus). The mandibular force profile of N. brachyops is generally similar to that of other scimitar-toothed nimravids, but it differs markedly in having a symphyseal region much better adapted towards labiolingual and torsional loads (Zx / Zy canine = 1.89), similar to the condition observed in the extant Neofelis nebulosa (Fig. 4) (Therrien, 2005). Interpretation Again assuming similarity of cortical bone thickness and of safety factors, nimravids are shown to have been able to generate bite forces equal to or even greater than felids of similar mandibular length (Table 1). For example, at the carnassial (P 4 M 1 interdental gap), Pogonodon platycopis had a maximum bite force 10 % superior to Panthera pardus and the bite of Hoplophoneus primaevus was 66 % more powerful than that of Acinonyx jubatus. The difference in mandibular force profiles between ecomorphs indicates that dirk-toothed and scimitartoothed nimravids killed their prey differently. The dorsoventrally and labiolingually stronger mandibular symphysis of dirk-toothed nimravids demonstrate that greater loads were exerted on the anterior extremity of the mandible than in scimitar-toothed nimravids, suggesting that the former may have emphasized the canine bite as a means to subdue their prey. With their elongate sabres and inferred ambush predatory style, dirk-toothed nimravids may have delivered powerful canine bites, driving the sabres deep into their prey to kill it quickly. Consequently, the symphysis had to be reinforced dorsoventrally and labiolingually to sustain the localized stresses generated during the bite. In contrast, the close similarity of the mandibular force profiles of scimitar-toothed nimravids to those of canids suggests they may not have relied on a single, powerful canine bite to kill their prey. Instead, these cursorial sabretoothed predators probably pursued their prey and delivered slashing canine bites to weaken it, their short, coarsely serrated sabres proving ideal for that purpose (e. g. Martin, 1980). The weaker symphyseal region of the smaller D. felina individual suggests that it may have hunted slightly smaller prey than its larger and older conspecifics. The relatively uniform Zx / Zy profile near the cheek teeth in scimitar-toothed nimravids (Fig. 7) indicates that the shape of the mandibular ramus remains constant throughout its length. The constancy of the Zx / Zy ratio reflects a feeding behaviour in which the ramus was not adapted for high-load applications, such as bone-cracking, but for the unique purpose of slicing flesh as seen in conical-toothed felids (see Therrien, 2005). The anteriorly increasing Zx / Zy values along the mandibular ramus in dirk-toothed nimravids (Hoplophoneini; Fig. 11) indicate that the ramus becomes relatively deeper towards the symphysis and, thus, that dorsoventral stresses become increasingly more important. All nimravids, except for N. brachyops, have strongly dorsoventrally buttressed mandibular symphyses compared to conical-toothed felids, with minimum Zx / Zy canine values of 2.37 for Eusmilus cerebralis and maximum values of 4.15 in Hoplophoneus dakotensis. This adaptation suggests that torsional stresses and the unpredictability of load orientation, such as those produced by struggling prey, must have been greatly reduced when sabretoothed nimravids delivered canine bites. In other words, prey must have been restrained and offered relatively little resistance when the sabretooths employed their sabres or else the symphysis would have been better adapted to withstand labiolingual and torsional stresses. Dorsoventral force Labiolingual force Relative force (Log Zx / L) (Log Zy / L) (Zx / Zy) Interestingly, the Zx / Zy canine values of hoplophoneines are significantly higher than those of scimitartoothed nimravids, suggesting that the former may have been better at immobilizing their prey. Given that dirk-toothed sabretooths had limbs as robust as those of bears while scimitar-toothed nimravids had less robust limbs more typical of cursorial predators (Anyonge, 1996), mandibular stresses may not have been as well constrained within the sagittal plane during bite in scimitar-toothed nimravids, thus requiring Dorsoventral force Labiolingual force Relative force (Log Zx / L) (Log Zy / L) (Zx / Zy) 1.04 1.46 Dinictis felina (L = 11.21 cm) Dinictis felina (subadult; L = 9.54 cm) Pogonodon platycopis (L = 14.65 cm)	en	Therrien, François (2005): Feeding behaviour and bite force of sabretoothed predators. Zoological Journal of the Linnean Society 145 (3): 393-426, DOI: 10.1111/j.1096-3642.2005.00194.x, URL: https://academic.oup.com/zoolinnean/article-lookup/doi/10.1111/j.1096-3642.2005.00194.x
CB5187EBFFB0FF934D02FD5DBB9FA7B5.taxon	description	Mandibular force profiles No dichotomy of mandibular force profiles occurs in barbourofelids, reflecting the sole occurrence of dirktoothed ecomorphs in this lineage (Fig. 8). The dorsoventral and labiolingual force profiles show that force values are greater at P 3 P 4 than at the carnassial, an observation apparently contradicting beam theory, which states that bite force should decrease with increasing distance from the articulation (see Material and methods above). These perplexing results indicate that a significant change in mandibular morphology occurs near the cheek teeth. Indeed, the mandibular corpus of barbourofelids is labially inclined at the cheek teeth to insure proper occlusion with the upper carnassial. This inclination places the cheek teeth ‘ on a sort of lateral “ island ” away from the ramus’ (Schultz et al., 1970: 11), P 3 being situated farther laterally than M 1. Consequently, the labiolingual diameter of the corpus increases at each interdental gap (see Schultz et al., 1970: fig. 5), producing the observed anterior increase in bending force. Although strong in absolute terms, the mandibular symphysis of barbourofelids appears to be subequal or only slightly stronger than at the cheek teeth (Fig. 8), a condition reminiscent of the canid profile (Fig. 4) (see Therrien, 2005). However, this similarity is solely an artefact of the labial rotation of the ramus described above, which inflates the force values at the cheek teeth. When compared to the dorsoventral and labiolingual force values behind the carnassial (where the ramus is not rotated), the symphyseal region is seen to be stronger than the mandibular corpus, a condition also typical of dirk-toothed nimravids. The relative mandibular force profile (Zx / Zy) reflects the fact that the barbourofelid mandible is effectively wider near the cheek teeth relative to the postcarnassial region, the Zx / Zy values being lower at the P 3 P 4 and P 4 M 1 interdental gaps (Fig. 8). At the canine, the mandible is strongly dorsoventrally buttressed, attaining Zx / Zy canine values of 2.77 or higher. Interpretation Barbourofelids have been described as having extremely robust mandibular rami (Schultz et al., 1970), a fact clearly supported by the biomechanical models presented here. When comparison of Zx / L - values is made posterior to M 1, barbourofelids are shown to have possessed more powerful bites than nimravids and extant felids of similar mandibular length (Table 1). For example, B. morrisi is shown to have had a bite force 44 % superior to the similar-sized Hoplophoneus dakotensis, 66 % superior to the similarsized Panthera pardus, and equal to P. tigris, a larger felid (28 % - longer mandible). Barbourofelis fricki, the largest barbourofelid, is estimated to have been able to generate a bite 63 % more powerful than P. leo. As was the case in dirk-toothed nimravids, the dorsoventrally and labiolingually strong mandibular symphysis of barbourofelids demonstrates that important loads were exerted on the anterior extremity of the mandible, indicating that a powerful canine bite was used to subdue prey. The Zx / Zy profiles of barbourofelids are atypical for dirk-toothed predators. Although the mandibular ramus increases in depth anteriorly (Schultz et al., 1970), as in hoplophoneines (see above), this fact is obscured in the biomechanical models by the lateral rotation of the ramus near the cheek teeth, effectively decreasing the Zx / Zy values at P 3 P 4 and P 4 M 1 (Fig. 8). Nevertheless, those Zx / Zy values are still within the range of modern felids (Fig. 4), indicating an adaptation for slicing flesh (Therrien, 2005), and the anteriorly increasing depth of the ramus suggests that loads are better constrained within the sagittal plane towards the symphysis. As was the case for dirk-toothed nimravids, the Zx / Zy canine values of barbourofelids are significantly higher than those of scimitar-toothed nimravids, sug- Dorsoventral force Labiolingual force Relative force (Log Zx / L) (Log Zy / L) (Zx / Zy) Barbourofelis whitfordi (L = 13.01 cm) gesting that the dirk-toothed barbourofelids could more efficiently immobilize their prey. Anyonge (1996) included barbourofelids in his study of extinct predator locomotor behaviour and obtained enlightening results. By comparing the brachial index (maximum length of the radius / maximum length of the humerus) and crural index (maximum length of the tibia / maximum length of the femur) of extinct predators to those of modern carnivores, Anyonge (1996: fig. 3) discovered that while the crural index of the leopard-sized B. morrisi was similar to that of the lion-sized Smilodon fatalis, ambush predators and ambulators (i. e. bears), its brachial index was lower than that of the machairodont and ambush predators, being closer to the index of ambulators. Because lower intralimb indices (shorter distal element relative to proximal element) indicate a greater power output at the limb extremity, due to a shorter out-lever arm (Hildebrand et al., 1985; Hildebrand, 1995), Anyonge’s (1996) results suggest that the forelimbs of the small barbourofelid had the same biomechanical advantage or leverage as those of the larger S. fatalis while its hindlimbs were biomechanically capable of generating a relatively greater force. Similarly, Baskin (in press) calculated brachial and crural indices for B. loveorum (BI = 0.72, CI = 0.75) and B. fricki (BI = 0.72, CI = 0.65) and discovered that they are even smaller than those of S. fatalis (BI = 0.82, CI = 0.84). These results suggest that the power output of the limbs of barbourofelids may have been greater than in S. fatalis! Therefore, the dirk-toothed barbourofelids could have easily subdued prey with their powerful forelimbs, thus limiting torsional stresses induced by struggling prey, prior to delivering a powerful canine bite to rapidly kill it.	en	Therrien, François (2005): Feeding behaviour and bite force of sabretoothed predators. Zoological Journal of the Linnean Society 145 (3): 393-426, DOI: 10.1111/j.1096-3642.2005.00194.x, URL: https://academic.oup.com/zoolinnean/article-lookup/doi/10.1111/j.1096-3642.2005.00194.x
CB5187EBFFADFF944F95FAA2BAF6A670.taxon	description	Represented by a single, long-lived species, Thylacosmilus atrox, this clade of dirk-toothed marsupials lived in South America from the late Miocene until the late Pliocene (Fig. 1) (Pascual & Bondesio, 1982; Goin & Pascual, 1987). Extensive work has been done on the craniodental and postcranial adaptations of thylacosmilines and their comparison with those of eutherian sabretooths (Riggs, 1934; Marshall, 1976; Turnbull, 1976, 1978; Churcher, 1985; Goin & Pascual, 1987). The different dental formula (I 1, C 1, P 2, M 4) and absence of a carnassial tooth homologous to that of carnivorans in Thylacosmilus (the fourth molar is modified into a shearing blade) requires additional measurement sites. Even though the thylacosmiline interdental gaps may not be functionally equivalent to those of eutherians, they give an overall appreciation of the force variation along the mandible (from post- M 4 to P 2 P 3, and behind the canine). Both an adult (P 14531, holotype) and a juvenile individual (P 14344, paratype) were studied in order to elucidate differences in mandibular force profiles, which could give insight into changes in feeding behaviour occurring during thylacosmiline ontogeny. A brief note on the proper orientation of the Thylacosmilus mandible must be made. Originally, Riggs (1934: plate 3) believed the sabres of Thylacosmilus to be strongly divergent. Consequently, he reconstructed the mandible with laterally flaring mandibular flanges, even though this resulted in a shorter mandibular symphysis than indicated by the rugosity on the medial aspect of the flanges. Later studies revealed that the sabres were not divergent and that the mandibular symphysis extended nearly to the lower extremity of the mandibular flange (Turnbull, 1976; Churcher, 1985; Goin & Pascual, 1987), which resulted in a slight reorientation of the mandibular corpus. Beam models designed for both mandibular orientations have delivered nearly identical results, except for the values at the canine due to the shorter mandibular symphysis of Riggs’ (1934) reconstruction. Owing to the similarity of results, only the model for the revised mandibular orientation will be discussed. Mandibular force profiles Other than differing in absolute values, the profiles of the two thylacosmilines are nearly identical (Fig. 11). Dorsoventrally (Zx / L), mandibular force values decrease slowly along the cheek teeth, attaining a minimum at P 2 P 3. Labiolingually (Zy / L), the mandibular force values remain relatively constant over the molars and decrease rapidly over the premolars, attaining a minimum at P 2 P 3. The mandible at the canine is extremely strong, both dorsoventrally and labiolingually, being much stronger than at the carnassial-analogue M 4. Juvenile and adult Thylacosmilus differ the most from other carnivorous mammals and, surprisingly, from each other in their relative mandibular force profile (Zx / Zy; Fig. 11). In the adult, the relative mandibular force profile remains relatively constant (Zx / Zy = 1.50) over the entire tooth row, the mandible being slightly narrower behind the last molar (Zx / Zy = 1.80). These values are lower than those observed in carnivorous eutherians. In the juvenile, the mandible maintains a relatively constant shape under the anteriormost molars (Zx / Zy ≈ 1.72), but is narrower under the last molar and the premolars (Zx / Zy> 2.13). Finally, the symphyseal region is extremely deep in thylacosmilines, with Zx / Zy canine values rivalling those of Barbourofelis fricki and Hoplophoneus dakotensis. The juvenile Thylacosmilus has higher Zx / Zy canine values than the adult (4.27 and 3.15, respectively), although it is impossible to evaluate the statistical significance of this difference. Interpretation In addition to assuming similarity of cortical bone thickness and of safety factors for bite force estimation purposes, tooth homology must also be considered when comparing eutherians and metatherians. Because the M 3 M 4 interdental gap in Thylacosmilus is situated at the same relative position as the P 4 M 1 interdental gap in sabretoothed eutherians (∼ 42 – 44 % of mandibular length), although slightly farther back than in conical-toothed carnivorous eutherians and marsupials (∼ 50 %; Emerson & Radinsky, 1980; Goin & Pascual, 1987; Werdelin, 1987; Bryant & Russell, 1995), comparison with the eutherian precarnassial interdental gap (P 4 M 1) is considered valid. The biomechanical approach indicates that Thylacosmilus could exert a bite force 34 % greater than Panthera tigris, a felid of similar mandibular length (Table 1). Both the dorsoventral and labiolingual force profiles show that the symphyseal region of Thylacosmilus was stronger in bending than the mandibular corpus near the cheek teeth, indicating that this predator delivered powerful canine bites. The very shallow negative slope formed by the Zx / L - values along the mandibular ramus suggests that force was uniformly applied along the entire tooth row (Fig. 11). In order to explain the unusual labiolingual force profile, one must look at the shape of the mandibular ramus. As was the case in barbourofelids and smilodontines, the mandibular ramus is labially inclined near the cheek teeth, being at its farthest lateral position at M 2 (also see Riggs, 1934); this feature is even more obvious in the juvenile specimen studied. The labial rotation of the mandibular ramus slightly increases the labiolingual diameter near the molars, explaining the high labiolingual force over the posterior molars. The relatively constant Zx / Zy values between M 4 and P 2 in the adult thylacosmiline (Fig. 11) reflect the uniform shape of the mandibular ramus (see Riggs, 1934; Churcher, 1985; Goin & Pascual, 1987). The constancy of the Zx / Zy ratio indicates that the ramus is not adapted to sustain localized high loads, such as in bone cracking, but for the unique purpose of slicing flesh (Therrien, 2005). This interpretation is corroborated by the dentition of thylacosmilines, which is unique among carnivorous marsupials in being specialized for pure slicing (Riggs, 1934; Churcher, 1985; Goin & Pascual, 1987). Even the highly carnivorous borhyaenines, the sister taxon to thylacosmilines, possessed robust premolars that have been interpreted as an adaptation for bone cracking (Marshall, 1978). The lower Zx / Zy values of the mandibular ramus in the adult thylacosmiline relative to other carnivorous Dorsoventral force Labiolingual force Relative force (Log Zx / L) (Log Zy / L) (Zx / Zy) Thylacosmilus atrox, adult (L = 20.54 cm) mammals is presumably an artefact of the lateral rotation of the mandibular corpus. The fact that the lowest Zx / Zy values in the juvenile thylacosmiline correspond to the farthest lateral extension of the tooth row strongly supports this interpretation. The high Zx / Zy canine values of thylacosmilines indicate that the symphyseal region was extremely well buttressed dorsoventrally and that loads were well constrained within the sagittal plane during the canine bite (Fig. 11). Similar buttressing is achieved only by the dirk-toothed predators H. dakotensis and B. fricki. Consequently, Thylacosmilus must have used its sabres against a restrained or immobilized prey, for struggling prey would have induced unpredictable stresses for which the mandible was not designed to sustain. The reduced incisor battery of Thylacosmilus argues against any significant grasping function for the anterior extremity of the mandible (Goin & Pascual, 1987) but its appendicular skeleton best resembles that of eutherian dirk-toothed sabretooths. The humerus is extremely robust with a well-developed deltopectoral crest, indicative of a powerful forelimb (Riggs, 1934; Argot, 2004 a, b). The humeral and ulnar articulations of the radius, combined with the prominent lateral epicondylar crest of the humerus, indicate that the forelimb was capable of extensive pronation / supination movements for grasping (Riggs, 1934; Argot, 2004 b). The presence of a semiopposable pollex in Thylacosmilus further supports the interpretation that the forelimbs were used for grasping (Argot, 2004 a, b). Although the radius is incomplete, Riggs (1934: 25) stated that it ‘ appears to have been almost as large as the tibia’; one must presume that Riggs referred to the tibia of the paratype (juvenile) because the holotype (adult) lacks this element. If this is the case, then the brachial index for the adult Thylacosmilus would have been near 0.75 (derived from measurements in Riggs, 1934), lower than Smilodon fatalis but similar to Barbourofelis loveorum (see above; Baskin, in press). The hindlimb was also robust. The femur has a straight shaft and a prominent, rounded head, a condition unlike that of any other marsupial but reminiscent of bears and eutherian sabretooths (Riggs, 1934; Argot, 2004 a, b). Although a complete tibia is not preserved in the adult, the crural index of the juvenile Thylacosmilus, derived from dimensions reported by Riggs (1934), is 0.71, which is also slightly lower than S. fatalis but within the range of B. loveorum and extant bears (see above; Anyonge, 1996; Baskin, in press). These characteristics do support the hypothesis that Thylacosmilus, like dirk-toothed sabretooths, was a powerful ambush predator capable of pulling down large prey and immobilizing it with its forelimbs (see Argot, 2004 a, b). Despite the clearly robust postcranial morphology of Thylacosmilus, Goin & Pascual (1987) argued that this predator could not have subdued large prey (e. g. macraucheniids, megatheriids, mylodontids or toxodonts) with its forelimbs because: (1) these prey would have been too large to be tackled by the ‘ cougar-size’ Thylacosmilus; (2) the absence of gregarious hunting behaviour among living polyprotodonts argues against collaborative hunting in Thylacosmilus to subdue large prey; and (3) thylacosmilines did not possess retractile claws, which would have allowed them to grasp prey. Instead, the authors suggested that Thylacosmilus rammed its prey in order to knock it down, its robust head being capable of absorbing the shock of the impact, a highly unusual, if not improbable, technique. Two of the three arguments employed by Goin & Pascual (1987) to discredit a dirktooth-like behaviour in thylacosmilines are invalid. First, the holotype of Thylacosmilus (P 14531) has cranial and mandibular dimensions comparable to those of a tiger (the skull being only slightly shorter due to the vertically orientated occipital region), thus suggesting an animal much larger than a cougar (Puma concolor). Additionally, the robust appendicular skeleton of the adult Thylacosmilus, almost bear-like in proportions [compare intermembral indices presented above with Anyonge’s (1996) results], suggests that it could have tackled large prey (if not fully-grown adults of the aforementioned prey species, then at least juveniles and / or subadults). Secondly, other than among canids and hyaenids, gregariousness is uncommon among eutherian carnivores; even among modern felids feeding on large prey, collaborative hunting is the exception rather than the rule (Schaller, 1972; Scheel & Packer, 1991; Caro, 1994). Gregarious behaviour among eutherian sabretooths has neither been clearly established (see McCall, Naples & Martin, 2003), nor is it a requisite for their inferred killing technique on large prey. Recently, Van Valkenburgh & Sacco (2002) investigated sexual dimorphism in Rancho La Brea carnivores and observed that the degree of sexual dimorphism exhibited by S. fatalis suggested that this animal lived either in a typical solitary felid social organization or in a monogamous breeding system like that of wolves (with one dominant pair and offspring from the current and / or previous years), but clearly did not form prides like African lions. Thus, the dramatic ramming behaviour inferred by Goin & Pascual (1987) rests solely on the apparent lack of retractile claws in Thylacosmilus, as inferred on the basis of a single proximal half of an ungual bearing likeness to those of Borhyaena tuberata (see Riggs, 1934) and the absence of such features among marsupials in general. Re-examination of the type material of Thylacosmilus by Argot (2004 b) revealed that none of the osteological correlates of claw retraction (see Gonyea & Ashworth, 1975) are present. It is conceivable, albeit just a supposition, that the ever-growing sabres of Thylacosmilus could have evolved in response to a reduced efficiency to immobilize prey. In strong contrast with eutherian sabretooths, Thylacosmilus possessed a reduced incisor battery and lacked retractile claws, which raise the possibility that this predator would not have been as good at subduing prey as its eutherian analogues. Consequently, the risk of accidental sabre breakage would have been increased, particularly during the initial phase of the attack when torsional stresses are at their greatest. Ever-growing sabres may have been the solution adopted by thylacosmilines to compensate for their less efficient grappling adaptations due to their borhyaenoid phylogenetic heritage (e. g. Argot, 2003, 2004 a, b). Finally, the higher Zx / Zy canine values of juvenile Thylacosmilus relative to the adult is reminiscent of the situation observed in modern felids characterized by extended periods of parental care (Therrien, 2005). Although caution is advisable given the limited sample size, this interpretation is not unreasonable. Indeed, the study of dental eruption sequence in Homotherium serum (Rawn-Schatzinger, 1983), S. fatalis (Tejada-Flores & Shaw, 1984), and Barbourofelis species (Bryant, 1988, 1990) has revealed that juvenile eutherian sabretooths underwent extended parental care and became independent around the same age as modern felids. During this extended period of parental care, juvenile sabretooths would have learned and honed their skills to kill prey with sabre-like canines under adult supervision while they fed on prey killed by a parent. In juvenile S. fatalis, the deciduous sabre (dC 1) is preserved until the permanent canine attains a similar length (c. 65 – 70 mm). At that time, dC 1 is shed and the permanent sabre pursues its eruption. Therefore, juveniles do not undergo a stage where they lack functional sabres (Tejada-Flores & Shaw, 1984). Comparison of microwear on deciduous and permanent upper canines revealed that juveniles were not as precise or careful with their sabres as adults (Akersten, 1985), possibly reflecting a learning period. Consequently, juveniles could have practised their killing technique with their deciduous canines. By the time the permanent sabres were in place, the individual had refined its killing techniques and would have been apt to use the sabres efficiently and safely. In contrast, the dC 1 in Barbourofelis, and most nimravids, is nearly as large as the permanent sabre (from 80 % to 98 % its length) and its eruption is delayed until the permanent cheek teeth have erupted, an equivalent stage of dental eruption to the permanent canines of extant lions (Bryant, 1988, 1990). Thus, juvenile barbourofelids underwent a long ontogenetic interval without functional upper canines before the large dC 1 started acting as ‘ a second pair of permanent sabres at the expense of the deciduous pair and hence a fully functional replacement in the event of breakage to the first pair early in life’ (Bryant, 1988: 305). With its ever-growing upper canines, Thylacosmilus probably did not suffer too seriously from the detrimental consequences of sabre breakage. Juveniles could have learned gradually how to use their sabres and in the event that they broke a sabre accidentally, owing to inexperience and / or carelessness, the tooth would have eventually regrown to its original size. The sabre could then have been resharpened through thegosis against the lower canines (Goin & Pascual, 1987).	en	Therrien, François (2005): Feeding behaviour and bite force of sabretoothed predators. Zoological Journal of the Linnean Society 145 (3): 393-426, DOI: 10.1111/j.1096-3642.2005.00194.x, URL: https://academic.oup.com/zoolinnean/article-lookup/doi/10.1111/j.1096-3642.2005.00194.x
