Miocene deposits of the Santa Cruz Formation, Patagonia, comprise a diverse and excellently preserved vertebrate fauna, allowing detailed paleobiological and paleoecological studies based on three ecological parameters: body mass, diet, and substrate preference. In contrast to the small and arboreal extant sloths, Bradypus and Choloepus, Santacrucian sloths were much more diverse and larger, and comprised 11 genera previously characterized as arboreal or climbing forms. Here, we focus on body mass estimation based on measurements of postcranial elements. We present a morphometric database comprising 64 linear, base-ten logged variables applied to Santacrucian sloths and a wide sample of extant mammals, as well as the body mass of the extant taxa as reported in the literature. To detect any potential phylogenetical bias, we performed a variance decomposition test on our sample of extant mammals. Based on four orthogram statistics, logged body mass was found not to be dependent on phylogenetic tree topology. Predictive equations for the body mass of extant mammals were generated through multiple regression analysis, using weighting procedures to avoid taxonomic biases and stepwise analysis to discard redundant variables. Using this procedure, we derived separate equations for the scapula, humerus, radius, ulna, pelvis, femur, tibia plus fibula, astragalus, and calcaneum. These equations were then applied to estimate the body mass of our sample of Santacrucian sloths. We obtained an average body mass of about 70 kg for the megalonychid Eucholoeops. Among stem megatherioids, Hapalops ranged between 30 and 80 kg, Analcimorphus was estimated at 67 kg, and Schismotherium at 44 kg. Larger genera included the megatheriid Prepotherium (∼123 kg), and the mylodontids Analcitherium (∼88 kg) and Nematherium (∼89 kg). The medium to large body size of Santacrucian sloths imposed constraints on their climbing abilities. Megalonychids and stem megatherioids were likely unable to access the finest branches, while megatheriids and mylodonts were more terrestrial forms.
The mostly fluvial deposits of the Santa Cruz Formation (late Early Miocene, ∼18–16 Ma) along the Atlantic coast of Santa Cruz Province (Patagonia, Argentina) preserve a rich and diverse vertebrate fauna, comprising xenarthrans such as sloths (Folivora), anteaters (Vermilingua), and armadillos and glyptodonts (Cingulata), as well as marsupials, astrapotheres, notoungulates, litopterns, rodents, primates, and birds (Tauber 1997a,b; Vizcaíno et al. 2012c). The remains are excellently preserved, thus offering the possibility of detailed paleobiological analyses.
Eleven genera of relatively large sloths have been identified from the Santa Cruz Formation. These include Eucholoeops and Megalonychotherium (forming part of the Megalonychidae); Planops and Prepotherium (Megatheriidae); Hapalops, Hyperleptus, Analcimorphus, Schismotherium, and Pelecyodon (stem megatherioids sensu Gaudin 2004); and Nematherium and Analcitherium (Mylodontidae). These fossil taxa contrast in their diversity and size with the extant sloths, which are represented by only two genera (Bradypus and Choloepus), both fully arboreal and weighing less than 10 kg (Nowak 1999). The Santacrucian genera are considered to represent the first major radiation of three of the five major folivoran families (Megalonychidae, Megatheriidae and Mylodontidae; see Fig. 1). By contrast, no representatives of Nothrotheriidae are known from the Santa Cruz Formation, while Bradypodidae lacks any fossil record whatsoever (see Gaudin 2004; McDonald and De Iuliis 2008).
Previous studies have analyzed the position of mammalian species within their respective communities in terms of three ecological parameters: body mass, diet, and substrate preference (including locomotion) (Andrews et al. 1979; Van Couvering 1980; Reed 1998; Kay and Madden 1997; Vizcaíno et al. 2006, 2008, 2010; Kay et al. 2012). Out of the latter, the diet and substrate preference of Santacrucian sloths have been the focus of several studies. Bargo et al. (2009) analyzed the masticatory movements and potential food sources of Eucholoeops, and made comparisons with other megatherioids, such as Hapalops, and mylodontids, such as Nematherium. White (1993, 1997) used several functional indices to propose that Hapalops and Eucholoeops were mainly arboreal or semi-arboreal, while Prepotherium was likely more terrestrial. Finally, a recent morphometric study focusing on the forelimbs of the Santacrucian sloths concluded that they resembled extant anteaters and pangolins, rather than extant sloths, in terms of their functional capabilities (Toledo et al. 2012). This finding potentially implies the ability to dig, and probably arboreal and semiarboreal habits for these taxa.
Body mass is among the most important variables in mammalian biology because it correlates with several significant biological parameters, such as basal metabolic rate, cardiac rhythm, sleeping times, or energetic requirements (e.g., Smith and Savage 1955; Eisenberg 1981; Hildebrand 1988; Brown and West 2000 and references therein). Previous body mass estimates for the Santacrucian sloths have mainly relied on regressions based on the bicondylar width of the femur or the diameter of the femoral head (White 1993; Croft 2000, 2001).
Obtaining reliable mass estimates of extinct mammals from allometric functions has been controversial (Millien and Bovy 2010). Mass estimates based on dental measurements were reported by Fortelius (1990) and by Millien and Bovy (2010), who concluded that, when available, other measures are preferable. Several authors have provided body mass estimates for Pleistocene ground sloths based on linear regressions (e.g., Fariña et al. 1998; Christiansen and Fariña 2003), but cautioned that taking a wider set of variables into account would likely lead to more robust estimates.
De Esteban-Trivigno et al. (2008) developed a robust set of predictive equations based on fore- and hind limb measurements from a wide sample of extant mammals, and applied them to Pleistocene xenarthrans, including ground sloths. These authors used weighting procedures to avoid taxonomic bias and indices of perceptual error to evaluate the accuracy of their estimates. Postcranial elements have been claimed as the best elements on which to base mass estimates, because they support the weight of the body (Gingerich 1990; White 1993; De Esteban-Trivigno et al. 2008 and references therein). This is especially true for fossil sloths, as their limbs supported relatively larger muscular masses than found in any living mammal, except those with fossorial adaptations (Bargo et al. 2012). Unfortunately, the equations of De Esteban-Trivigno et al. (2008) are not applicable to the Santacrucian remains, because there are few specimens retaining all of the fore- and hind limb elements, despite the generally excellent state of preservation. Here, we present a set of predictive body mass equations for fossil sloths, based on limb bone measurements, which are suitable for incomplete specimens.
Institutional abbreviations.—AMNH, American Museum of Natural History, New York, USA; FMNH, Field Museum of Natural History, Chicago, USA; MLP, Museo de La Plata, La Plata, Argentina; MPM-PV, Museo Regional Provincial Padre M. Jesús Molina, Río Gallegos, Argentina; YPM-VPPU, Yale Peabody Museum, Vertebrate Paleontology, Princeton University Collection, New Haven, USA.
Other abbreviations.—QMLE, Quasi Maximum Likelihood Estimator; PE, Percentual Error of the estimate; RE, Ratio Estimator; SE, Smearing Estimator.
Material and methods
The fossil material studied here was collected by expeditions during the 19th and early 20th centuries (MLP, YPM-VPPU, AMNH, and FMNH), as well as MLP-Duke University joint expeditions (MPM-PV) over the last nine years (2003–2011). The specimens of extant mammals are housed mostly at the AMNH and the FMNH.
We constructed a morphometric database of extant mammals including xenarthrans, marsupials, pholidotans, tubulidentates, carnivorans, primates and rodents (SOM 1 and 2, Supplementary Online Material available at http://app.pan.pl/SOM/app59-Toledo_etal_SOM.pdf), based on 64 linear variables (Appendix 1 and Figs. 2, 3) commonly used to quantify the proportions and dimensions of the appendicular skeleton (e.g., Alexander et al. 1979; Scott 1990; Sargis 2002a, b; Elissamburu and Vizcaíno 2004; Candela and Picasso 2008; De Esteban-Trivigno et al. 2008). When available from the literature (Nowak 1999), gender-specific body mass was included. A second database of 44 Santacrucian sloths was built using the same measurements (Table 1; SOM 2).
Phylogenetic bias analysis
In order to identify a putative phylogenetic bias in the distribution of body mass within our sample of extant mammals, we performed a variance decomposition test (Ollier et al. 2006). This test decomposes the variance of a given trait with respect to the topological structure of a phylogenetic tree (in our case, the phylogeny of Asher and Helgen 2010; Fig. 4), and determines whether it is concentrated at one or more particular nodes of the latter. This method does not take into account the distance between individual nodes, but focuses only on the topology of the tree. The total absence of phylogenetic dependency was considered the null-hypothesis. Following Ollier et al. (2006), we compared the observed variance decomposition to a theoretical distribution based on 20 000 Monte-Carlo permutations using four statistics: R2Max, with high values indicating significant changes in body mass variance at a single node; Dmax, corresponding to the Kolmogorov-Smirnov statistic, which tests whether the vector describing the studied variable matches a random sample from a uniform distribution; SkR2k, indicating the degree of asymmetry (skewness) in body mass variance; and SCE, which measures the local averaged variation in the orthogram values. All calculation were performed in R (R Development Core Team 2011) using the ‘ade4’ package (Dray and Dufour 2007).
Santacrucian fossil sloth specimens included in this study.
Predictive equations for body mass estimation
We developed separate multiple regression models for the scapula, humerus, ulna, radius, pelvis, femur, tibia-fibula, astragalus, and calcaneum, using the 64 size variables as predictors and body mass as the dependant variable. Owing to the volumetric nature of body mass, and to ensure a normal distribution, all of the data were base-ten log-transformed prior to the analysis. Following De Esteban-Trivigno et al. (2008), a weighting factor was introduced to account for the over-representation of some clades (xenarthrans and carnivorans). Redundant variables were discarded from all models using a stepwise method (backwards-forwards; Hocking 1976), based on the Akaike Information Criterion (Sakamoto et al. 1986; SOM 1). The final models were then used to estimate the body mass of Santacrucian sloths.
When body mass values inferred using least squares regression are transformed back to arithmetic units, log-adjustment can result in biased estimates (Smith 1993). We corrected for this potential bias (and thus restored dimensionality) by applying the “Ratio Estimator” (RE) of Snowdon (1991), the “Quasi Maximum Likelihood Estimator” (QMLE) of Fergusson (1986), and the “Smearing Estimator” (SE) of Duan (1983).
The equation reliability was evaluated using the Percent Prediction Error %PE = 100 x [observed-predicted]/predicted (Smith 1984). This index is more reliable than the correlation coefficient, which is affected by the range of values of the dependent variable and thus a poor indicator of the predictive power of the independent variable(s) (e.g., De Esteban-Trivigno et al. 2008; Cassini et al. 2012).
All calculations were performed using the open-access statistical package R (R Development Core Team 2011). Details of the scripts and modules used here are provided in SOM 1.
Phylogenetic bias analysis.—None of the four orthogram statistics for variance independence resulted in significant values (see Table 2 and Fig. 5), thus indicating that the (log-transformed) body mass of extant mammals is not dependent on tree topology.
Predictive equations for body mass.—Those equations with lower R2 values (radius and calcaneum) showed higher %PE values, and vice versa (humerus and femur). The former also showed higher QMLE values. All the SE and QMLE values were very close to 1, which suggests little, but not necessarily no bias. We therefore conservatively used the RE coefficient to re-transform the logged body size estimates (Table 3; see also Christiansen and Harris  and De Esteban-Trivigno et al. ). While all equations showed high R2 (>0.80) and highly significant P values, the humeral and femoral equations stood out as those with the lowest %PE (∼9 and ∼8%, respectively). Therefore, we assumed that the most reliable equations were those corresponding to the humerus and femur, while those based on the radius, calcaneum and ulna were less accurate (Table 3). The final equations used to estimate the body mass of the Santacrucian sloths are shown in Table 4. No fibular variable was conserved in the final equations.
Obtained values for the four statistics for variance independence, based on 20 000 Monte-Carlo permutations.
Statistics for each of the predictive equations. R2, coefficient of determination; adjusted-R2, coefficient of determination adjusted by the number of explanatory terms in the model; F, value of Fisher Test for null dependency and associated degrees of freedom (df); p-F, probability of the F-test; RE, “Ratio Estimator”; QMLE, “Quasi Maximum Likelihood Estimate”; SE, “Smearing Estimate”; %PE, Percent Prediction Error of the estimate.
Final predictive equations for each of the studied elements of the appendicular skeleton. See Appendix 1 for explanations of abbreviations.
Estimated body mass of Santacrucian sloths.—The mean body mass and ponderated mean value (weighted by the %PE value; see Christiansen and Harris 2005) for each specimen are listed in Table 5. Mean values for each genus are shown in Table 6. Estimated body mass values ranged between 40 and 80 kg (Tables 5, 6), except for the megatheriids and mylodontids. The genus with the most reliable estimates was the stem megatherioid Hapalops, owing to the great number of available specimens and preserved elements, followed by the megalonychid Eucholoeops. Conversely, the estimates for the remaining genera were less reliable, owing to the relatively small number of specimens and/or elements on which they were based.
Eucholoeops showed an average body mass of about 60 kg, with E. fronto weighing about 65 kg, and E. ingens estimated at 53 kg (Table 5). For Hapalops, the mean was about 45 kg, but with high intra- and interspecific variability (Table 5). Thus, for example, the smallest and largest specimens of H. ruetimeyeri and H. elongatus differed by more than 10 kg. Hapalops rectangularis, on the other hand, showed less variation, ranging between 31.5 and 38 kg. Other species of the genus were heavier, reaching weights of 49.4 kg (H. platycephalus) and 56 kg (H. indifferens). Finally, H. longiceps was the biggest species of the genus (83.8 kg). The remaining basal megatherioids included Analcimorphus (67 kg) and Schismotherium (44 kg).
Larger genera were represented by the megatheriid Prepotherium (123.2 kg), the mylodontids Analcitherium (88.2 kg) and Nematherium (89.3 kg), and some taxonomically unassigned specimens showing even larger estimates. Based on the tibial, astragalar and calcaneal equations, the mylodontid MPM-PV 3406 was about 170 kg, which was only surpassed by the megatheriid MPM-PV 3408 at around 200 kg (based on the astragalar equation).
Mean and weighted estimates of body mass (in kg) for specimens analyzed in this work. Weighted means were calculated taking into account %PE. Abbreviations: Pond. Mean, mean ponderated by the %PE index for each equation; Pond. #, number of equations per specimen. SD, standard deviation.
Estimation of body mass.—The absence of any obvious phylogenetic bias in our analysis of extant mammals likely is a result of the wide range of sampled taxa, and corroborates the conclusions of Mendoza et al. (2006). However, this finding does not exclude the possibility that phylogenetic biases may affect patterns within particular clades, such as folivorans. Most of our predictive equations include the diaphyseal width of each bone (humerus, ulna, femur, and tibia), as well as the width and depth of the distal epiphysis, thus reflecting the overall robustness of the element.
Our equations represent an improvement on those of White (1993, 1997) and Croft (2000, 2001) in including more than one variable, more than a single postcranial element, and being derived from a more diverse sample of extant mammals. In addition, although our equations are not as robust as the ones provided by de Esteban-Trivigno et al. (2008), they have the advantage of being applicable to incomplete specimens—a very common situation in paleontology, The best estimators are the humeral and femoral equations (lowest %PE values), while were radial and calcaneal equations show a greater degree of distortion. These findings are in agreement with those of Gingerich (1990) and Biknevicius (1993).
Our body mass estimates for Eucholoeops differ markedly from the 40 kg obtained by Croft (2000) based on regressions against the femoral bicondylar width, and are closer to the results of Bargo et al. (2009), who based them on the length of the humerus using allometric equations provided by Scott (1990). The apparent variability in the body size of Hapalops is also consistent with the results of Bargo et al. (2009), which ranged between approximately 30 and 80 kg, but differs strikingly from the estimates of other authors. Our estimate of around 84 kg for H. longiceps is very similar to the 73.8–81.4 kg calculated by White (1993) based on the femoral head diameter, but lower than the 156 kg obtained by Croft (2001). The same holds true for H. indifferens, with our estimate of 56 kg being much higher than the 20.1 kg estimated by Croft (2000), but similar to the 58.3 kg estimated by White (1993).
Mean and weighted estimates of body mass (in kg) for the genera included in this study. Abbreviations: Pond. Mean, weighted mean; Pond. Estimate, body mass estimate weighted by the number of equations instead of number of specimens (only for Hapalops and Eucholoeops), SD, standard deviation.
At 44 kg, Schismotherium is one of lightest megatherioids. However, owing to its relatively poor state of preservation, we were unable to estimate the body mass of Pelecyodon (AMNH 9240), which seems to be a juvenile individual. By contrast, the mylodontids (Analcitherium and Nematherium, around 90 kg) and the megatheriids (Prepotherium, about 120 kg) are among the heaviest Santacrucian sloths. This estimate is lower than the 174–229 kg proposed for Prepotherium by White (1993). Cassini et al. (2012) classified Santacrucian ungulates into three base-ten logarithmic body size ranges: small-sized forms (1–10 kg), medium-sized-forms (10–100 kg), and large-sized forms (100–1000 kg). According to this scale, most of the Santacrucian sloths are therefore medium- sized, with the exception of the large-sized megatheriids and the mylodontids.
Body size and substrate preference.—Previous authors characterized the Santacrucian sloths as arboreal or semi-arboreal (White 1993, 1997). Access to arboreal substrates offers several advantages, such as food resources not available at ground level, shelter, and more efficient locomotion when the ground is rough or flooded (Hildebrand 1988). However, locomotion on arboreal substrates involves its own difficulties, which are related to two main issues: movement on a three-dimensional and discontinuous substrate, and the risk of falling during locomotion and resting. Most extant arboreal mammals are either small or medium-sized (Table 7; Eisenberg 1981). Indeed, there are no capable climbers as large as Santacrucian sloths in our sample of extant mammals, except for Gorilla (150 kg) and some specimens of Acynonix and Panthera tigris (50 and 200 kg, respectively). A relatively large body size may decrease the risk of predation and allow the animal to budget its energetic resources long-term (Hildebrand 1988). The latter advantage is especially important for xenarthrans, owing to their low basal metabolic rate (McNab 1985, 2000; Vizcaíno et al. 2006; Bargo and Vizcaíno 2008; Vizcaíno 2009). On the other hand, large body size also imposes strong constraints on, for example, reproduction, substrate preference and locomotor strategies (Hildebrand 1988).
Among extant xenarthrans, anteaters, the sister group to sloths, seem to be better analogues than extant sloths in terms of understanding the relationship between the size and substrate preference of the Miocene taxa. Body size among extant xenarthrans is correlated with substrate preference: the minute Cyclopes (less than 0.5 kg) is strictly arboreal, Tamandua (around 5 kg) is a semi-arboreal form, and the giant anteater Myrmecophaga (about 30 kg) is terrestrial and only climbs occasionally (Taylor 1978; White 1993; Rodrigues et al. 2008). It therefore seems that body size imposes a strong constraint on arboreal capabilities (Taylor 1985), implying that some of the larger Miocene sloths, such as Prepotherium and the mylodonts (exceeding 80 kg), were almost certainly too heavy to be fully arboreal.
An important factor involved in arboreality is the resistance to a fall. Heavier animals reach higher kinetic energies during a fall than smaller ones for equal conditions of distance, time and gravity acceleration (Hill 1950), making them more vulnerable to injury. Thus, most of the more agile and acrobatic arboreal animals are small (Cartmill 1985). The only fully arboreal primate as heavy as the Santacrucian sloths is the orangutan (Pongo pygmaeus, ∼90 kg; Nowak 1999), whose locomotor style consists of suspending its body from branches using most or all of its long limbs at the same time, thus distributing its weight over several supports. The animal is considered to be very close to the maximum operational body size for feeding in an arboreal substrate (Cartmill 1985). Among carnivorans, many ursids are heavier than the orangutan and are excellent climbers (e.g., the spectacled bear Tremarctos ornatus, and the American black bear Ursus americanus, both around 120 kg, Nowak 1999). However, these bears are among the smallest members of the family, and they are not as specialized as primates in their climbing abilities.
Body mass and substrate preference of the extant mammals included in this analysis.
Constraints on arboreal locomotion imposed by large body size are also related to the maximum resistance of the support. A large animal is restricted to move only on those supports that can resist its weight. Thus, Tamandua is unable to move along the finest, terminal branches on which Cyclopes moves (Taylor 1985). The same likely holds true for most of the Santacrucian sloths, especially Prepotherium and the mylodontids, which were characterized by White (1997) as “more terrestrial”. Circumventing the problem of fine branches, many primates and extant sloths move by distributing their weight among two or more slender supports. However, the forelimbs of Santacrucian sloths are short and robust compared to those of the extant species, which are more similar to primates such as Hylobates. Instead, the forelimbs of the Miocene taxa resemble those of digging mammals such as anteaters, pangolins and aardvarks (Toledo et al. 2012; Bargo et al. 2012), although lacking the extreme ulnar morphology seen in armadillos (White 1993; Vizcaíno et al. 1999, 2001, 2006, 2008, 2011, 2012b; Vizcaíno and Milne 2002).
In summary, Santacrucian sloths likely were arboreal and semiarboreal forms restricted in their climbing capabilities by their relatively large body size. While megalonychids (Eucholoeops) and stem megatherioids (Hapalops, Analcimorphus, Schismotherium and Pelecyodon) were probably arboreal climbers, albeit restricted to the thickest branches, megatheriids (Prepotherium) and mylodontids (Analcithe-Analcitherium and Nematherium) were more terrestrial forms that climbed only occasionally, for example to avoid predation. Further morphofunctional studies are being performed to gain further insights into these issues. Our conclusions are supported by the diet of the Santacrucian mylodontids, which likely consisted of tubers and fruits (Bargo et al. 2012). In contrast to extant sloths, which generally weigh less than 10 kg, Santacrucian sloths ranged from more than 10 kg to around 120 kg. The heavier Santacrucian genera belonged to two families, Megatheriidae and Mylodontidae, which also included a range of other large genera, some with gigantic body sizes, during the Plio-Pleistocene (Vizcaíno et al. 2012a). In this lights, we interpret the small body size of Choloepus as a derived condition reached independently from Bradypus. Based on the fossil evidence, the last common ancestor of the extant Choloepus and Bradypus was likely of large body size.
The variation in body mass shown by the extant mammals analyzed in this study does not appear to be biased by phylogeny.
The predictive equations developed here based on several elements of the appendicular skeleton show relatively low predictive error values, especially for the scapular, humeral, femoral, and tibia-fibular equations.
Most Santacrucian sloths ranged between 40 and 80 kg (megalonychids and stem megatherioids), while others (megatheriids and mylodontids) ranged between 80 and 150 kg.
Medium and large body sizes imposed constraints on climbing capabilities of Santacrucian sloths, thus likely preventing megalonychids and stem megatherioids from accessing the finest branches, while restricting megatheriids and mylodonts largely to the ground.
The authors wish to thank Marcelo Reguero and Lucas Pomi (both MLP), Walter Joyce and Daniel Brinkman (both YPM), John Flynn, Nancy B. Simmons, and Christopher Norris (all AMNH), and William F. Simpson and Kenneth D. Angielczyk (both FMNH) for kindly providing access to the collections under their care, and to the Dirección de Patrimonio Cultural of Santa Cruz Province and the Museo Regional Provincial P.M.J. Molina of Río Gallegos, for support during field work. We are grateful to Jonathan M.G. Perry (Center for Functional Anatomy and Evolution, Johns Hopkins University, School of Medicine, Baltimore, USA) for their helpful comments, and the reviewers Jennifer White (East Stroudsburg University, USA) and Richard Fariña (Universidad de la República, Montevideo, Uruguay) for their valuable comments and suggestions that greatly improved the manuscript. This is a contribution to the projects PIP-CONICET 1054, PICT 0143, and UNLP N647, National Geographic Society to SFV and Richard F. Kay, and National Science Foundation to R.F. Kay.