Locomotor Adaptations in Plio-Pleistocene Large Carnivores from the Italian Peninsula: Palaeoecological Implications

Current Zoology 57 (3): 269−283, 2011

Locomotor adaptations in Plio-Pleistocene large carnivores from the Italian Peninsula: Palaeoecological implications

Carlo MELORO*

Hull York Medical School, The University of Hull, Loxley Building, Cottingham Road Hull HU6 7RX, UK

Abstract Mammalian carnivores are rarely considered for environmental reconstructions because they are extremely adaptable and their geographic range is usually large. However, the functional morphology of carnivore long bones can be indicative of locomotor behaviour as well as adaptation to specific kind of habitats. Here, different long bone ratios belonging to a subsample of extant large carnivores are used to infer palaeoecology of a comparative sample of Plio-Pleistocene fossils belonging to Italian paleo-communities. A multivariate long bone shape space reveals similarities between extant and fossil carnivores and multiple logistic regression models suggest that specific indices (the brachial and the Mt/F) can be applied to predict adaptations to grassland and tropical biomes. These functional indices exhibit also a phylogenetic signal to different degree. The brachial index is a significant predictor of adaptations to tropical biomes when phylogeny is taken into account, while Mt/F is not correlated any- more to habitat adaptations. However, the proportion of grassland-adapted carnivores in Italian paleo-communities exhibits a negative relationship with mean oxygen isotopic values, which are indicative of past climatic oscillations. As climate became more unstable during the Ice Ages, large carnivore guilds from the Italian peninsula were invaded by tropical/closed-adapted species. These species take advantage of the temperate forest cover that was more spread after 1.0 Ma than in the initial phase of the Quaternary (2.0 Ma) when the climate was more arid [Current Zoology 57 (3): 269–283, 2011]. Keywords Carnivora, Long bones, Quaternary, Climate change

Although members of Carnivora are literally defined relationship between the morphological variability of by virtue of what they eat, they exhibit a great variabi- long bones shape and the environment occupied by lity also in locomotor adaptations (Ewer, 1973; Taylor, both extant and extinct species. 1989; Van Valkenburgh, 1987). All the elements of the Interestingly, several long bone length ratios of car- appendicular skeleton of carnivores are usually modi- nivores appear to be related with the running perfor- fied in order to maximise locomotor performance which mance (e.g., maximum speed). In particular, metatar- can be remarkable if compared with that exhibited by sal/femur length ratio (= Mt/F) has been intensively other mammals (e.g., the cheetah Acinonyx jubatus can investigated because it can be applied also to the scatter achieve the highest speed ever recorded among extant record of extinct carnivores and their prey (Janis and terrestrial mammals). Such intuitive relationship between Wilhem, 1993). Garland and Janis (1993) revealed that bones and performance is reflected in the external mor- the relationship of such ratio with speed (in large mam- phology of the pelvis, of long bones and other appen- mals) is not an artefact of phylogenetic relatedness even dicular elements like the scapula. It is also useful to if it results not in a straightforward line. Christiansen infer locomotor behaviour in extinct species (Taylor, (1999) confirms that in both carnivores and ungulates 1989; Van Valkenburgh, 1987; Munthe, 1989; Ginsburg, Mt/F co-varies with speed but other long bone ratios can 1999; Argant, 2004; Antón et al., 2005). Accordingly, also be considered as even better descriptor of running Taylor (1989) categorised carnivores for their ability in performance (e.g., Olecranon/Radius length). Carrano running or climbing or digging. These categories are (1999) also demonstrated that long bone indices are reflected also in long bone shape and dimension. The useful for determining locomotor habit in both mam- latter observation allows quantitative analyses, which, mals and dinosaurs. Cursorial species tend to exhibit for carnivores, were firstly performed by Van Valken- longer metatarsal, more slender limb elements and burgh (1985, 1987) in her survey of extant and extinct shorter femora. large carnivore guilds. Her study reveals a complex Samuels and Van Valkenburgh (2008) recently con-

Received July 10, 2010; accepted Jan 14, 2011. ∗ Corresponding author. E-mail: [email protected] © 2011 Current Zoology 270 Current Zoology Vol. 57 No. 3 firmed that long bone ratios are useful behavioural pre- 1 Materials and Methods dictors also in rodents that exhibit a remarkable range of locomotion types. Clearly, the analysis of long bone 1.1 Long bones indices proportions is a powerful analytical tool to explore lo- Twenty two extant species of large carnivores be- comotor adaptations in a wide range of vertebrates. longing to families Canidae, Felidae, Hyaenidae and However, they can be also interpreted to elucidate com- Ursidae were considered in order to reconstruct loco- plex behaviours especially in mammalian carnivores. motor behaviour in extinct Italian Plio-Pleistocene large For instance, Lewis (1997) extended the analysis of carnivores. long bone indices to clarify ecological adaptations of Polly (2010) recently demonstrated that eco-morphology African Plio-Pleistocene carnivores. She evidenced a of carnivore communities can be used for environmental difference in eco-morphology between extant African reconstruction. Mean calcaneal gear ratio of extant car- carnivores and Plio-Pleistocene species. The extinct nivore communities correlates with several environ- sabertooth cats (Homotherium, Megantereon and Dinofelis) mental parameters such as precipitation or temperature were prey grapplers. The African Plio-Pleistocene Canis at large geographical scale (all North America). This spp. were more adapted to omnivory and to cursoriality. approach is very promising but it assumes that all Similarly, Meachen-Samuels and Van Valkenburgh members of a carnivore community are represented in a (2009) analysed different ratios from forelimb elements fossil site. This condition is quite rare because of ta- to predict prey size range in extant felids. The study phonomic bias favouring preservation of large taxa clearly shows that forelimb long bone proportions are (Damuth, 1982). Consequently, I restrict my analyses to better predictors of prey size rather than locomotion. large taxa (defined as species bigger than 7 kg) whose Felids adapted to kill large prey exhibit more robust fossil record has been extensively investigated for the humerus and radius at the midshaft and distal epiphyses. Italian peninsula (Raia et al., 2005, 2006a, b, 2007; Nonetheless, felids represent a special case in Carnivora Meloro et al., 2007, 2008). because they are homogenously adapted to hunt verte- For each extant species humerus, radius, femur and brates. This condition rarely applies to different families tibia length (L in cm) measurements were extracted of Carnivora where species tend to exhibit a wider range from the literature (Bertram and Biewener, 1990) and of feeding adaptations (from insectivory to herbivory, combined into the brachial index (BI = radius/humerus Gittleman, 1985; Meiri et al. 2005). length cf. Lewis, 1997) and the tibia/femur index (T/F= Here, I present a survey of long bone indices in a tibia/femur length). subset of both extant and fossil species in order to elu- Data in Janis and Wilhem (1993) and Christiansen cidate eco-morphology of Italian Plio-Pleistocene large (2002) were reviewed to extract for each species meta- carnivore guilds. I focus on the implication of recon- tarsal/femur ratio (Mt/F = metatarsal/femur length). structing locomotor behaviour for palaecological Taken overall, three variables were considered for each reconstructions and, in particular, environmental ones. extant large carnivore: BI, T/F, Mt/F (see Appendix 1). As Lewis (1997) pointed out, some indices like the Even if the number of indices is quite limited, they are brachial (radius/humerus length), can be useful to de- broadly representative of long bone functional mor- scribe the adaptation of a large carnivore for particular phology (Carrano, 1999) and they were selected to cre- habitats (closed, mixed, open savannah). This rela- ate a consistent comparative dataset for the fossil data, tionship will be empirically tested in order to define which are inevitably fragmentary (see next section). for each fossil carnivore particular preference for cer- 1.2 Fossil sample tain biomes. Phylogenetic relatedness among species For extinct Plio-Pleistocene taxa, the latter variables will also be taken into account to validate the statisti- were computed combining measurements directly taken cal models based on the interspecific dataset of extant from museum collections or after an extensive review of Carnivora (Garland et al., 2005). Closely related spe- the Italian and European literature on each fossil large cies tend to resemble each other in several morpho- carnivore (Appendix 2). As the fossils are usually in- logical as well as behavioural traits because of their complete and scattered in space, several long bone in- shared evolutionary history (Blomberg et al., 2003). dices were obtained combining long bone lengths from For this reason, comparative methods are necessary to different localities. When possible, multiple long bone apply when interspecific data are explored for pa- measurements from the same locality were combined by laeoecological interpretations. computing their mean as representative of the species MELORO C: Locomotion in Quaternary carnivores 271 mean value. In most cases, it was possible to use linear case will be considered with caution in the interpretation measurements of specimens from Italian localities. of the results bearing in mind that for some species sev- Among caniforms only Italian specimens were used to eral long bone estimates can be inaccurate. estimate indices in: Canis etruscus, Canis arnensis, Ly- 1.3 Multivariate analyses caon falconeri, Ursus spelaeus and Ursus arctos while Species are analysed in a multivariate eco-morpho- for feliforms only in Lynx issiodorensis and in the logical space (cf. Van Valkenburgh, 1985; Lewis, 1997) Pleistocene European lynx Lynx lynx. in order to obtain better information on their general Among extinct species, I included also European long bone proportion. A Principal Component Analysis Galerian canids C. mosbachensis and C. aff. arnensis (PCA, Blackith and Reyment, 1971) was applied to from L’Escale for which linear measurements were identify the possible interrelationship between the in- available (Appendix 2). It was not possible to obtain dices in a sample of 22 extant and 20 extinct large car- linear measurements of Pliocrocuta perrieri and the nivores. The Principal Component vectors were ex- Pleistocene leopard Panthera pardus for which Del tracted using a correlation matrix that allows to maxi- Campana (1947) reported the same variability as in ex- mize the variance of the indices analysed. The tant leopards. It is worth mentioning that for most three-dimensional morphospace was interpreted in Plio-Pleistocene species only fragments of some long order to infer locomotor behaviour of extinct bones exist. In this case, the maximum length was esti- Plio-Pleistocene carnivores. An UPGMA clustering mated from medio-lateral or antero-posterior diameter was also applied to the Euclidean dissimilarity matrix by using regression equations (at family level) with the extracted by considering the three indices simultane- highest “r” value reported in Bertram and Biewener ously (cf. Lewis, 1997). (1990). Although this approach is not ideal, it allows 1.4 Habitat and climate variables reconstructing long bones length for several species: In order to validate the possible relationship between C. arnensis, L. falconeri, A. pardinensis, L. spelaeus, M. locomotor behaviour and habitat, each extant large car- cultridens and P. gombaszoegensis (see Appendix). nivore (n = 22) was assigned to a habitat value accor- When no diameter was available, long bone length was ding to Ortolani and Caro (1996). Ortolani and Caro estimated by applying regression equations relative to (1996) identified the presence of mammalian carnivores another element (of the same limb anterior or posterior) into broad biome categories: temperate forest, tropical for which allometric scaling can be evidenced at family forest, grassland, arctic, riparian and desert. These level (data Bertram and Biewener 1990, Appendix 3). categories are not mutually exclusive so that species This was the only way to estimate the radius length in such as the grey wolf are considered adapted to arctic, A. pardinensis (from humerus length), the femur length temperate forest, grassland and desert as well. Conse- in L. falconeri, Lynx spelaeus, Chasmaporthetes quently, for each habitat category (e.g., arctic) a score of lunensis and the tibia in C. arnensis. Taken overall on 0.0 was assigned as indicative of “not adapted” while 20 extinct species analysed, 35% have at least one 1.0 as “adapted”. length’s element reconstructed from another element of Separate logistic regression models (with forward the same limb. Wald option) were performed by considering all studied For Chasmaporthetes lunensis long bone linear indices (BI, T/F, Mt/F) as potential predictors of the measurements were available only for the tibia (from Mt. dependent variable (0.0 = not adapted, 1.0 = adapted), in Perrier) and the III metacarpal (from Layna) (in Kurtén order to predict adaptations to each of the six habitat and Werdelin, 1988). The length of other elements was categories. The Wald option allows considering a step- estimated by considering long bone measurements of wise approach where covariates are singularly added to both Chasmaporthetes borissiaki and Chasmaporthetes the model step by step if the Wald statistic corresponds ossifragus (in Berta, 1981). Although Kurtén and Wer- to a probability value smaller than 1.0. If the Wald delin (1988) reported some differences in length propor- probability is bigger than 0.10, variables are not in- tions of Chasmaporhetes spp. it is, here, assumed that cluded in the model (Hair et al., 1998) and the stepwise such differences are minimal when compared to inter- procedure stops. specific variability. To assess the accuracy of the logistic regression This approach allows obtaining estimates of the three model, the Hosmer-Lemeshow goodness-of-fit statistic long bone indices which are useful to describe the lo- was considered. This goodness-of-fit statistic is a robust comotor behaviour of extinct taxa (Appendix 3). Each approach for small sample size (Hair et al., 1998) and it 272 Current Zoology Vol. 57 No. 3 is based on grouping cases into deciles of risk and lected by the logistic regression models and habitat comparing the observed probability with the expected categorisation. This model includes phylogeny as an probability within each deciles. A chi square distribution error term in the former equation: Y = bX + ε where (in is used for such comparison. If the associated probabil- this case) Y is the habitat categorisation, X is the long ity value is not significant, then the model is stable and bone index and the error term ε is quantified by the shows a good fit. phylogenetic covariance matrix computed directly from These models allow obtaining prediction of the pos- the phylogenetic tree. The phylogenetic signal was as- sible habitat adaptations in fossil Plio-Pleistocene car- sessed using the module Physig for Matlab (Blomberg nivores by their long bone proportions. The latter pa- et al., 2003) while PGLS analyses were performed with laeoecological reconstruction of large carnivores will be NTSYS vr 2.2 (Rohlf, 2006b). discussed on the light of the climate changes raised by 2 Results Italian Plio-Pleistocene Paleo-Communities (PCOM) spanning in time from 3.2 until 0.8 Ma (PCOMs, Raia et 2.1 Locomotor behaviour al.，2005, 2006a, 2006b, 2007; Meloro et al., 2007, Three Principal Component axes define a morpho- 2008). For each PCOM a proportion of large carnivores space of 22 extant large carnivores and 20 fossil forms. adapted to particular habitats (e.g., 50% of tropic The first two axes explain together more than 95% of adapted and 50% of Grassland adapted) was assessed. the total variance (Table 1). They are informative of the Oxygen 18 isotopic values drawn from Kroon et al. intra family variation of locomotory features (Fig. 1), (1998), site 967 which exhibited a continuous record with which they are strongly associated: the first axis is from 3.2 Ma, were then considered as a proxy for cli- positively associated with the hind limb ratios (both T/F mate in each PCOM interval (cf. Raia et al., 2005; and Mt/F that are highly interrelated r = 0.836) while Meloro et al., 2008). A total of 441 records were con- the second is strongly influenced by the brachial index sidered spanning 3.2 – 0.8 Ma. For each time bin of that shows no significant correlation with the other in- each PCOM mean of Oxygen 18 values were computed. dices (Table 1). The third axis is loaded by the Mt/F The latter value is indicative of major climate shifts while the other indices contribute only to a very low from warm to cold conditions that could be correlated degree to the variance explained. with the relative proportion of carnivores adapted to The three dimensional plot clearly shows a separation particular habitats. in between families, that occupy very distinct areas of 1.5 Model validation the morphospace (Fig. 1). Indeed, families are signifi- The association between functional long bone indices cantly different when PC scores are used as dependent and habitat adaptation could be the result of shared evo- variables (MANOVA Wilk’s Lambda = 0.091, F9.0, 87.77 lutionary history in Carnivora. Blomberg et al. (2003) = 16.405, P < 0.0001). evidenced a strong phylogenetic signal in Mt/F for the Interestingly, several fossil taxa show functional as- dataset of Carnivora and Ungulata. For this reason, I sociation with extant species independently of phylog- first present the test of Blomberg et al. (2003) for the eny. For instance, Homotherium crenatidens clusters selected functional indices using the phylogeny of Bin- with extant hyenas while Pachycrocuta brevirostris is inda-Emonds et al. (1999). The K statistic (which is a much closer to the polar bear than to the other hyenas. measure of the strength of phylogenetic signal) is com- Exceptions exist also in extant taxa with the cheetah puted for all the long bone indices and a randomisation clearly clustering with several canids (Fig. 1). In this test is presented to assess statistical significance (Blomberg et al., 2003). The Bininda-Emonds et al. Table 1 Eigenvalues, % of explained variance and loadings for each Principal Component (1999) supertree was manipulated with Mesquite vs. 1.1 (Maddison and Maddison, 2009) in order to select only PC1 PC2 PC3 the 22 extant taxa used for this analyses. This supertree Eigenvalue 00.026 00.009 000.001 was favoured among others, because the branch lengths % Variance 71.870 24.470 003.660 are expressed as time of divergence after a robust cali- Cum. % 71.870 96.340 100.000 bration with the fossil record. Loadings Phylogenetic Generalised Least Square (PGLS, Rohlf, BI 00.475 00.878 0−0.060 2001, 2006a) is applied to test for the significant asso- T/F 00.963 −0.237 0−0.130 ciation between the long bone functional indices se- Mt/F 00.934 −0.010 000.356 MELORO C: Locomotion in Quaternary carnivores 273

Fig. 1 Scatter plot of three Principal Component axes extracted from three long bone indices The position of some fossil taxa is highlighted by labelling. plot, taxa occupy the morphospace so that canids clearly tive score). show higher values of Mt/F and T/F as an adaptation to If PCA scatter plot is informative of both phyloge- cursorial and strictly terrestrial locomotion, while at the netic and functional influences on long bone indices opposite side bears and hyenas exhibit lower T/F values. variability, UPGMA confirms this observation and al- PC2 shows taxa with small brachial index (short ra- lows interpreting species clustering in a more consistent dius relative to the humerus) occupying low scores (e.g., way (Fig. 2). Both extant and fossil bears form a the extant tiger, the leopard), while at the opposite side well-defined cluster that is close to some large felids species with a more balanced forelimb (radius and hu- (the lion, the tiger, the jaguar but also Megantereon and merus with the same proportion) such as the long legged the puma) and fossil giant hyenas. These taxa are more maned wolf Chrysocyon brachyurus and the striped generalists prey grapplers (all exhibit a low Mt/F values) hyena Hyaena hyaena. contrary to the cheetah, the wolf, the golden jackal The PC3 explains only a small portion of variance Canis aureus but also Lynx lynx, fossil dogs and Lycaon (3%) and it adds a limited amount of information rela- pictus which are cursorial (upper side of the tree, Fig. 2). tive to the magnitude of Mt/F and BI with species at its Interestingly, the extinct long legged Chasmaporthetes opposite extremities tending to exhibit lower (or higher) clusters with extant hyenas that are closed to the fossil Mt/F relative to their BI (e.g., the fossil Panthera gom- lion and lynx Lynx spelaeus but also Homotherium. baszoegensis negative score or Lycaon falconeri, posi- These forms show a shorter tibia relative to the femur. 274 Current Zoology Vol. 57 No. 3

Fig. 2 UPGMA tree obtained from Euclidean distance matrix extracted on the basis of three locomotor indices computed for each species. Families are labelled near each taxon: C = Canidae, F = Felidae, H = Hyaenidae, U = Ursidae. The cophenetic correlation is 0.735.

In the other cluster both extant and fossil dogs are 2.2 Habitat adaptations associated with medium sized cats like the serval, the Several logistic models were considered to predict lynx but also the leopard and several fossil species. habitat preference in extinct species from long bone Some clusters are consistent with taxonomic similarities indices. The presence/absence of large carnivores in such as the group of Canis latrans and Canis aff. arnen- temperate habitats cannot be predicted because no long sis but also the association of all Plio-Pleistocene fossil bone indices enter in a logistic regression model. Canis spp. On the other hand, BI is the only variable in a model Clearly, this cophenetic tree shows a mix of both computed to predict adaptation for tropical environ- taxonomic and functional signals that are not always ments (Hosmer and Lemeshow at first step: χ2 = 0.885, consistent because the cophenetic correlation value is df = 8; P = 0.999) (Table 2). In this case, 95% of species not indicative of a high fit (r = 0.735). not present in tropics are well predicted while 75% of MELORO C: Locomotion in Quaternary carnivores 275

Table 2 Parameters for different logistic regression models with Open/Closed, Tropic or Grassland variable as dependent variables

Predicted Steps Parameters B SE Wald df Sig. BI −28.509 14.747 3.737 1 0.053 Tropic 1 Constant 023.535 12.456 3.57 1 0.059 Mt/F 034.723 16.218 4.584 1 0.032 Grassland 1 Constant −10.254 05.35 3.674 1 0.055 tropic-adapted carnivores are classified correctly. For enigmatic also because several long bone indices are grassland environments only the Mt/F is the relevant estimates for this species (see Appendix 3). Few variable included in the model (Table 2) which is able to Plio-Pleistocene species are classified as possibly classify adaptation (score 1.0) at 100%, but adapted to tropical environment: the extinct dirk toothed non-adaptation at a lower rate (75%). Adaptations to the cat Megantereon cultridens, the Etouare lynx Lynx is- other biomes as described in Ortolani and Caro (1996) siodorensis and the cave bear Ursus spelaeus. Interes- cannot be predicted by any of the other logistic regres- tingly, the other fossil bears (Ursus minimus and U. sion models, which are always non-significant. Hence, etruscus) are predicted to be non-adapted to both long bone indices of large carnivores cannot be used to Grassland and Tropical biome as it is the case also for predict their adaptation to extreme environments like the extant bears. the arctic or desert but even the riparian ones. As for Nyctereutes megamastoides and Pliocrocuta The significant logistic models allow making some perrieri, no long bones were available. Their locomotor predictions on extinct Plio-Pleistocene forms. Some are adaptation could have been similar to that of their extant reliable while others need to be considered with caution counterpart Nyctereutes procyonoides and Parahyena (Table 3). For instance, most canids are predicted as brunnea (Kurtén, 1968), respectively. Consequently, the grassland specialist even if the result for L. falconeri is habitat preference was scored accordingly with the data available for the latter extant species in Ortolani and Table 3 Predicted habitat preference for Plio-Pleistocene large Caro (1996). carnivores 2.3 Test for phylogenetic inheritance Species Tropical Grassland The phylogenetic signal is always present and statis- Canis etruscus 0 1 tically significant in the functional indices considered Canis arnesis 0 1 (Table 4). The T/F exhibits the highest K while the bra- Lycaon falconeri 1 1 chial index the lowest. PGLS model supports the strong Canis mosbachensis 0 1 association between BI and adaptation to tropical biome C. aff.arnensis (L’Escale) 0 1 in the selected extant Carnivora (b = −1.945, r2 = 0.209, Canis lupus Pleist. 0 1 Fs = 5.283, df = 1, 20, P = 0.032) thus validating the Homotherium crenatidens 0 1 logistic regression model. On the other hand, Mt/F is Megantereon cultridens 1 1 not associated to grassland adaptation when phylogeny Acinonyx pardinensis 0 1 is taken into account (b = 1.746, r2 = 0.161, Fs = 3.829, Lynx issiodorensis 1 1 df = 1, 20, P = 0.064). Lynx spelaeus 0 1 2.4 Carnivore guilds and climate Panthera leo Pleist 0 1 The proportion of grassland or tropical carnivores in Panthera gombaszoegensis 0 1 each PCOM is significantly associated to Quaternary 18 Pachycrocuta brevirostris 0 1 climatic proxies (mean δ O). Interestingly and counter Crocuta crocuta Pleist 0 1 Table 4 Values of the K statistic, Mean Squared Error (MSE) for Chasmaporthetes lunensis 0 1 a star phylogeny, MSE for the candidate tree and associated Ursus minimus 0 0 probability after 999 randomizations

Ursus etruscus 0 0 K MSE Star MSE candidate P Ursus spelaeus 1 0 Mt/F 0.529 0.006 0.004 0.000 Ursus arctos Pleist 0 0 BI 0.215 0.012 0.017 0.037 0 = “open” and “non-adapted” for columns tropical and grassland; 1 = T/F 0.757 0.014 0.006 0.000 “closed” and “adapted” for columns tropical and grassland 276 Current Zoology Vol. 57 No. 3 intuitively a positive correlation occurs between mean (Lewis, 1997) which is supported after phylogeny is δ18 O and the proportion of tropic adapted taxa while the taken into account. On the other hand, the T/F has the same relationship is negative when grassland taxa are highest phylogenetic signal suggesting that it is strongly considered (Tab. 5; Fig. 3). That is: cold PCOMs (Gale- inherited in large Carnivora. For this reason, it is not a rian, Aurelian) have higher proportion of “tropical good predictor of habitat adaptations. The lack of asso- adapted” carnivores and a lower proportion of “grass- ciation between Mt/F and grassland adaptation after land adapted” carnivores. No correlation occurs with the PGLS probably reflects the better association of this variance of δ18 O computed for each PCOM. index with cursorial performance in Carnivora (Garland 3 Discussion and Janis, 1993). When all indices are considered simultaneously, a Like feeding apparatus, locomotor appendices of mix of phylogenetic and functional similarities occur large carnivores describe the dichotomy between func- between certain specific taxa (Fig. 2). For this reason, tional morphology and phylogenetic constraint. The the results obtained for Plio-Pleistocene carnivores survey performed on long bone indices variability of needs to be interpreted case wise. both extant and Plio-Pleistocene carnivores clearly Among canids, most of the analysed extinct species shows that morphological convergence is a rare (but possess only a few complete long bone specimens. In possible) event in the recent evolutionary history of the spite of the approximation in computing several indices, group (Van Valkenburgh, 1985, 1999, 2007). Such con- the long bone shape variability of extinct species is vergence occurs in long bone proportion between the concordant with that exhibited by extant dogs. It is not grey wolf and the cheetah or the spotted hyena and the useful to define if one species was faster than another by fossil sabertooth Homotherium. Clearly, the long bone virtue of its long bone proportion but what emerged is indices describe phylogenetic resemblance so that dis- Table 5 Pearson correlation coefficients (below) and associated P tinct families occupy separate portion of the functional values (above the diagonal) for proportion of carnivores and iso- morphospace (Fig. 1). This pattern is consistent with topic values computed for each PCOM (n = 9) previous observations as in Jones (2003). She analysed Pearson r / p Tropical Grassland mean δ18 O Var δ18 O several long bone ratios in both placental and marsupial Tropical − 0.000 0.004 0.171 carnivores evidencing similarities more at family level than by virtue of locomotor adaptation. Not surprisingly, Grassland −1.000 − 0.004 0.171 all the analysed functional indices exhibit a phyloge- mean δ18 O 0.848 −0.848 − 0.035 netic signal even if the brachial index has a very low K. Var δ18 O 0.500 −0.500 0.703 − This, in turn reflects its stronger functional signal Bold values are significant at P < 0.005

Fig. 3 Scatter plot of mean δ18 O vs proportion of grassland carnivores The exactly opposite trend applies when proportion of tropical adapted taxa are plotted on the Y axis. MELORO C: Locomotion in Quaternary carnivores 277 the relatively lack of peculiar morphologies among portions of extinct and modern large cats. The Pleisto- Plio-Pleistocene dogs. Canis etruscus and Canis arnen- cene lion from Equi (measurements taken in Del Cam- sis cluster within a large group including the extant pana, 1947) is remarkably similar to the sabertooth golden jackal (Canis aureus) but also medium sized tooth Homotherium crenatidens. This fact probably re- lynx, as well as the maned wolf supporting the idea that flects similarity in body mass but also peculiar func- their morphological variability was similar –but not tional convergence imposed by the emphasized cursori- entirely the same- to extant canids (Del Campana, 1913; ality in H. crenatidens as compared to the other modern Torre, 1967). Interestingly, the Galerian dog from felids (see Turner and Antón, 1997; Antón et al., 2005). Untermassfield clusters with Canis etruscus and the On the other hand, the relatively long legs of the Pleis- Pleistocene wolf. This cluster possibly reflects phyloge- tocene lion could be an adaptation to hunt in the severe netic similarities. Indeed, Rook and Torre (1996a, b) sug- snow conditions of the coldest glacial periods. There are gested the possibility of a single lineage with Canis also implications for the hunting behaviour of both spe- etruscus being the ancestor of Canis mosbachensis and cies: H. crenatidens was adapted to hunting very large the latter the ancestor of the grey wolf. On the other prey (Lewis, 1997; Turner and Antón, 1997; Antón et al., hand, the extant grey wolf does not cluster with the 2005), and probably also the Pleistocene lion shows Pleistocene form suggesting a possible morphological such adaptation. It is likely that the European Pleisto- change through time. Lycaon falconeri does not con- cene lion hunted species much heavier than modern Af- verge in long bone proportion with its extant counterpart rican ungulates. The results obtained for the dirk tooth Lycaon pictus leaving open the question about the run- Megantereon confirm its adaptation as a stalker hunter ning adaptation of this extinct dog. The lack of the first of closed forests (Turner and Antón, 1997) like the metacarpal digit supports the adaptation of Lycaon fal- modern puma. Panthera gombaszoegensis is, here, con- coneri to a cursorial lifestyle (Rook, 1994), but more sidered more cursorial than expected for a jaguar-like long bone data are necessary to validate such observa- large cat (cf. Figs.1, 2) but this result needs to be con- tion. In fact, it could also be possible that the latter sidered with caution because complete long bones are morphological character is not a case of functional con- scanty in the fossil record. Similarly, Acinonyx vergence but it is due to its phylogenetic relationship pardinensis was certainly a cursorial adapted carnivore with the extant Lycaon pictus (Martinez-Navarro and but it clusters with the racoon dog Nyctereutes procyon- Rook, 2003). oides and the leopard. They share a small brachial index Clearly all Plio-Pleistocene dogs were more cursorial (short radius relative to the humerus) that for A. than extant and fossil large cats or bears. Within this pardinensis was inferred on the basis of fragmentary cursorial group, it is possible to include (as expected) materials. Consequently, it is difficult to deduce if A. also the extant hyenas together with Chasmaporthetes. pardinensis was faster than the extant cheetah, but this The indices obtained for the latter species are estimates analysis does not support such hypothesis: it is plausible but they are based on C. borrissiaki and C. ossifragus that A. pardinensis was a sprinter (Turner and Antón, which should be less long legged than European C. 1997) but, because of its large body mass, its agility and lunensis (Kurtén and Werdelin, 1988). speed was probably not as extreme as in A. jubatus. On the other hand, the giant Pachycrocuta brevi- Lynxes are not so enigmatic in clustering with me- rostris but also the cave hyena (C. crocuta of the Pleis- dium-sized canids, but it is possible to note the rela- tocene) appears to be similar to the extant big cats in tively shorter forelimb proportions of Lynx issiodorensis long bone proportion. This fact is partially confirmed by from Olivola if compared to extant lynxes and the serval palaecological description of Turner and Antón (1996). (Figs. 1, 2). This fact confirms the observations of Accordingly, the tibia of P. brevirostris is shorter than Kurtén (1980) and Werdelin (1981) and such a feature that of extant and other Plio-Pleistocene hyenas sug- could be the consequence of two different factors. gesting a greater mechanical advantage in the hindlimb. Firstly, Lynx issiodorensis was less specialised than ex- Relative shortening is exhibited also by the radius. The tant lynxes in hunting in different conditions (from the similarity between P. brevirostris, the cave hyena and American deserts to the snowy mountains of Carpathi- big cats could possibly be the result of their similarity in ans) and as a typical Villafranchian taxon, was probably mean body mass (127 kg and 102 kg respectively, data specialised in woody habitats (to escape competition in Meloro et al., 2007). with cursorial canids). Secondly, Lynx issiodorensis was Different observations emerge in the long bone pro- probably more adapted to hunting medium-sized ungu- 278 Current Zoology Vol. 57 No. 3 lates rather than lagomorphs. Medium-sized ungulates carnivores appear to be plausible at least if we consider generally represent an important source of food for ex- valid the adaptation to grassland biome. The preference tant lynxes (Tumlison, 1987; Nowell and Jackson, 1996) of Lycaon falconeri, Megantereon cultridens, Lynx is- and the fact that lynxes are lagomorph specialists is siodorensis and Ursus spelaeus to potential tropical en- considered a “misconception” (Nowell and Jackson, vironments needs to be interpreted as a preference to 1996). For instance, Eurasian lynx Lynx lynx kills large more forested areas. This survey demonstrates only that prey fairly regularly, and although the Spanish lynx long bone indices can be used to predict habitat prefer- Lynx pardinus kills mostly water fowl and lagomorphs ence at least for the non-mutually exclusive variables now, this was not the case historically before habitat “grassland” and “tropical”. The other habitat categories destruction (Gil-Sánchez et al., 2006). Because of this, it can be probably predicted by morphological features is likely that the Plio-Pleistocene lynxes were well when more taxa are included without any body mass adapted in hunting small-medium sized ungulates, with threshold. Interestingly, the proportion of grassland- Lynx issiodorensis being more specialised in large prey. adapted or tropical-adapted large carnivores in each In extinct bears, it is possible to note again a pattern Plio-Pleistocene ecosystem is related to major climatic of long bone variability driven by phylogenetic rela- oscillations supporting the idea that large carnivores are tionships (Mazza and Rustioni, 1994). The lineage of important components of extant and extinct biomes. As brown bear is defined by U. minimus - U. etruscus the climate became colder in the Mediterranean area, group which are also similar in long bone proportions. the proportion of grassland-adapted carnivores dimin- On the other hand, U. spelaeus clusters with the polar ished. If we consider biome changes, this is plausible bear, suggesting that its limb proportions reflected a because several vegetation analyses support the arid particular adaptation to very cold climates. conditions of Villafranchian (ca 2.0 Ma) ecosystems This species by species approach revealed useful in characterised by abundance of Artemisia shrub lands defining locomotor behaviour of extinct large carnivores (Suc et al., 1995; Fauquette et al., 1999; Ghinassi et al., but, most interestingly, long bone indices are also in- 2004). Instead, Galerian (1.0 Ma) and Aurelian (0.3 Ma) dicative to define general patterns of association be- are characterised by the alternation of cold / temperate tween carnivores and habitat adaptations. Lewis (1997) phases which result in the spread of montane-subalpine already mentioned and utilised the brachial index as an forests (Malatesta, 1985). indicator of adaptability to open, mixed or closed habi- Torre et al. (1992) already mentioned a structural tats. But it is important to note that the logistic regres- change in Plio-Pleistocene large mammal fauna of Ital- sion models presented here are statistically more robust. ian peninsula directly related to climate oscillations. Such a relationship is complex because carnivores are Results presented here validate previous observations on not habitat specialists like ungulates or rodents. Large the spread of large carnivores in Italy with cursorial carnivores have large home ranges and their habitat se- grassland-adapted canids being favoured during the Vil- lection depends on prey availability but also on the den- lafranchian (which characterises also the Wolf event) sities of other competitors (Gittleman, 1985; Carbone while pantherine cats and modern taxa of temperate and Gittleman, 2002; Woodroffe and Ginsberg, 2005). habitat (e.g., wolf, brown bear) spread in the colder Consequently, several species morphologically well- Pleistocene phases. adapted to particular habitats may select sub-optimal habitats by virtue of external environmental factors. For Acknowledgments I am grateful to P. Raia, F. Carotenuto, C. Barbera and A. Kotaskis for their encouragments during the instance, the African wild dog is a large carnivore preliminary phases of this study. E. Cioppi and L. Rook kindly adapted to cursoriality and it is supposed to be a grass- assisted me during my visit to the Museo di Geologia e Pale- land species (as in Ortolani and Caro, 1996). But recent ontologia dell’Università di Firenze, Firenze, Italy. A special studies demonstrated that Lycaon pictus selects decidu- thank goes to A. Colamarco for her continuos support to my ous woodlands more than expected (Creel and Creel, research. Three anonymous reviewers increased the quality of 2002). This is just one case that is considered to have the manuscript with important comments and suggestions. ecological explanations (avoiding competition with li- Many thanks also to the editor Zhi-Yun JIA for his encourag- ons and hyenas) but most other large carnivores are ing invitation. usually constrained in sub-optimal habitat because of References habitat fragmentation due to human activity. On the other hand, the results obtained for extinct Antón M, Galobart A, Turner A, 2005. Co-existence of scimitar- MELORO C: Locomotion in Quaternary carnivores 279

toothed cats, lions and hominins in the European Pleistocene: Germany: Verlag 109–148. Implications of the post-cranial anatomy of Homotherium Ghinassi M, Magi M, Sagri M, Singer BS, 2004. Arid climate 2.5 latidens (Owen) for comparative palaeoecology. Quaternary Ma in the Plio-Pleistocene Valdarno Basin (Northern Apenni- Science Reviews 24: 1287–1301. nes, Italy). Palaeogeography, Palaeoclimatology, Palaeoecology Argant A, 2004. Les Carnivores du gisement Pliocéne final de 207: 37–57. Saint-Vallier (Dróme, France). The Carnivores of the late Gil-Sanchez JM, Ballesteros-Duperon E, Bueno-Segura JF, 2006. Pliocene site of Saint-Vallier (Dróme, France). Geobios 37: Feeding ecology of the Iberian lynx Lynx pardinus in eastern S133–S182. Sierra Morena (Southern Spain). Acta Theriologica 51: 85–90. Berta A, 1981. The Plio-Pleistocene hyaena Chasmaporthetes Gittleman JL, 1985. Carnivore body size: Ecological and taxo- ossifragus from Florida. Journal of Vertebrate Palaeontology 1: nomic correlates. Oecologia 67: 540–554. 341–356. Hair JF, Anderson RE, Tatham RL, Black WC, 1998. Multivariate Bertram JEA, Biewener AA, 1990. Differential scaling of the long Data Analysis. 5th edn. Upper Saddle River: Prentice Hall. bones in the terrestrial Carnivora and other mammals. Journal Janis CM, Wilhem PB, 1993. Were there mammalian pursuit of Morphology 204: 157–169. predators in the Tertiary? Dances with wolf avatars. Journal of Bininda-Emonds ORP, Gittleman JL, Purvis A, 1999. Building Mammalian Evolution 1: 103–125. large trees by combining phylogenetic information: A complete Jones M, 2003. Convergence in ecomorphology and guild struc- phylogeny of the extant Carnivora (Mammalia). Biol. Rev. ture among marsupial and placental carnivores. In: Jones M, Camb. Philos. Soc. 74, 143–175. Dickman C, Arcker M ed. Predators with Pouches: The Bio- Blackith RE, Reyment RA, 1971. Multivariate Morphometrics. logy of Carnivorous Marsupials. Collingwood: CSIRO Pub- London & New York: Academic Press. lishing, Australia, 285–296. Blomberg SP, Garland, T, Ives AR, 2003. Testing for phylogenetic Kroon D, Alexander I, Little M, Lourens LJ, Matthewson A et al, signal in comparative data: Behavioral traits are more labile. 1998. Oxygen isotope and sapropel stratigraphy in the eastern Evolution 57: 717–745. mediterranean during the last 3.2 million years. Proc. Ocean Carbone C, Gittleman JL, 2002. A common rule for the scaling of Drill. Prog. 160: 181–189. carnivore density. Science 295: 2273–2276. Kurtén B, 1968. Pleistocene mammals of Europe. Chicago: Aldine Carrano MT, 1999. What, if anything, is a cursor? Categories Publishing Company. versus continua for determining locomotor habit in mammals Kurtén B, 1980. The lynx from Etouaires, Lynx issiodorensis and dinosaurs. Journal of Zoology, London 247: 29–42. (Croizet and Jobert), late Pliocene. Annales Zoologici Fennici Christiansen P, 1999. Scaling of the limb long bones to bodymass 15: 314–322. in terrestrial mammals. Journal of Morphology 239: 167–190. Kurtén B, Werdelin L, 1988. A review of the genus Chas- Christiansen P, 2002. Locomotion in terrestrial mammals: The maporthetes Hay, 1921 (Carnivora, Hyaenidae). Journal of influence of body mass, limb length and bone proportions on Vertebrate Paleontology 8: 46–66. speed. Zoological Journal of the Linnean Society 136: Lewis ME, 1997. Carnivoran paleoguilds of Africa: Implications 685–714. for hominid food procurement strategies. Journal of Human Creel S, Creel NM, 2002. The African Wild Dog (Behaviour, Evolution 32: 257–288. Ecology and Conservation). Princeton & Oxford: Princeton Maddison WP, Maddison DR, 2009 Mesquite: A modular system University Press. for evolutionary analysis. Version 2.6 http://mesquiteproject.org Damuth J, 1982. Analysis of the preservation of community Malatesta A, 1985. Geologia e Paleobiologia dell’Era Glaciale. structure in assemblages of fossil mammals. Paleobiology, 8: Roma: La Nuova Italia Scientifica. 434–446. Martínez-Navarro B, Rook L, 2003. Gradual evolution in the Del Campana D, 1913. I cani Pliocenici di Toscana. Palaeonto- African hunting dog lineage Systematic implications. C. R. graphia Italica XIX: 189–254. Palevol 2: 695–702. Del Campana D, 1947. Carnivori quaternari della Tecchia e della Mazza P, Rustioni M, 1994. On the phylogeny of the Eurasian caverna di Equi nelle Alpi Apuane (mustelidi, canidi, felidi). bears. Palaeontographica Abt. A 230: 1–38. Palaeontographia Italica XLIV: 1–42. Meachen-Samuels J, Van Valkenburgh B, 2009. Forelimb indica- Ewer RF, 1973. The Carnivores. New York: Cornell University tors of prey-size preference in the Felidae. Journal of Mor- Press. phology 270: 729–744. Fauquette S, Suc J-P, Guiot J, Diniz F, Feddi N et al, 1999. Cli- Meiri S, Dayan T, Simberloff D, 2005. Variability and correlations mate and biomes in the West Mediterranean area during Plio- in carnivore crania and dentition. Functional Ecology 19: cene. Palaeogeography, Palaeoclimatology, Palaeoecology 152: 337–343. 15–36. Meloro C, Raia P, Barbera C, 2007. Effect of predation on prey Garland T Jr, Bennett AF, Rezende EL, 2005. Phylogenetic ap- abundance and survival in Plio-Pleistocene mammalian com- proaches in comparative physiology. Journal of Experimental munities. Evolutionary Ecology Research 9: 505–525. Biology 208: 3015–3035. Meloro C, Raia P, Carotenuto F, Barbera C, 2008. Diversity and Garland Jr T, Janis CM, 1993. Does metatarsal/femur ratio predict turnover of Plio-Pleistocene large mammal fauna from the the maximal running speed in cursorial mammals? Journal of Italian Peninsula. Palaeogeography, Palaeoclimatology, Pa- Zoology 229: 133–151. laeoecology 268: 58–64. Ginsburg L, 1999. Order Carnivora. In: Rössner GE, Heissig K ed. Munthe K, 1989. The Skeleton of the Borophagine (Carnivora, The Miocene Land Mammals of Europe. Dr. F. Pfeil. München, Canidae): Morphology and function. University of California 280 Current Zoology Vol. 57 No. 3

Publications in Geological Sciences Vol. 133, 115 pp. locomotor adaptations in living and extinct rodents. Journal of Nowell K, Jackson P, 1996. Wild Cats: Status, Survey and Con- Morphology 269: 1387–1411. servation Action Plan. Gland, Switzerland: IUCN. Suc JP, Bertini A, Combourieu-Nebout N, Diniz F, Leroy S et al, Ortolani A, Caro TM, 1996. The adaptive significance of color 1995. Structure of West Mediterranean vegetation and climate patterns in carnivores: Phylogenetic tests of classic hypotheses. since 5.3 ma. Acta Zool. Cracov. 38: 3–16. In: Gittleman JL ed. Carnivore Behavior, Ecology, and Evolu- Taylor ME, 1989. Locomotor adaptations. In: Gittleman JL ed. tion. Vol 2. Ithaca, IL: Cornell University Press, 132–186. Carnivore Behavior, Ecology, and Evolution. Vol. 1. Ithaca, Polly PD, 2010. Tiptoeing through the trophics: Geographic varia- NY: Cornell University Press, 382–409. tion in carnivoran locomotor ecomorphology in relation to en- Torre D, 1967. I cani villafranchiani della Toscana. Palaeonto- vironment. In: Goswami A, Friscia A ed. Carnivoran Evolu- graphia Italica 63: 113–138. tion: New Views on Phylogeny, Form, and Function. Cam- Torre D, Ficcarelli G, Masini F, Rook L, Sala B, 1992. Mammal bridge UK: Cambridge University Press, 374–410. dispersal events in the early Pleistocene of Western Europe. Raia P, Piras P, Kotsakis T, 2005. Turnover pulse or Red Queen? Courier Forsch.-Inst. Senckenberg 153: 51–58. Evidence from the large mammal communities during the Tumlison R, 1987. Felis lynx. Mammalian Species. 269: 1–8. Plio-Pleistocene of Italy. Paleogeography, Paleoclimatology, Turner A, Antón M, 1996. The giant hyaena Pachycrocuta brevirostris Paleoecology 221: 293–312. (Mammalia, Canrivora, Hyaenidae). Geobios 29: 455–468. Raia P, Meloro C, Loy A, Barbera C, 2006a. Species occupancy Turner A, Antón M, 1997. The Big Cats and Their Fossil Relatives. and its course in the past: Macroecological patterns in extinct New York: Columbia University Press. communities. Evolutionary Ecology Research 8: 181–194. Van Valkenburgh B, 1985. Locomotor diversity between past and Raia P, Piras P, Kotsakis T, 2006b. Detection of Plio-Quaternary present guilds of large predatory mammals. Paleobiology 11: large mammal communities of Italy: Integration to biochro- 406–428. nology. Quaternary Science Review 25: 846–854. Van Valkenburgh B, 1987. Skeletal indicators of locomotor be- Raia P, Meloro C, Barbera C, 2007. Inconstancy in predator/prey haviour in living and extinct carnivores. Journal of Vertebrate ratio in Quaternary large mammal communities with an ap- Paleontology 7: 162–182. praisal of mechanisms. Quaternary Research 67: 255–263. Van Valkenburgh B, 1988. Trophic diversity in past and present Rohlf FJ, 2001. Comparative methods for the analysis of continuous guilds of large predatory mammals. Paleobiology 14: 155–173. variables: Geometric interpretations. Evolution 55: 2143–2160. Van Valkenburgh B, 1999. Major patterns in the history of car- Rohlf FJ, 2006a. A comment on phylogenetic correction. Evolu- nivorous mammals. Annual Review of Earth and Planetary tion 60: 1509–1515. Science 27: 463–493. Rohlf FJ, 2006b. NTSYSpc v. 2.21c. New York: Exeter Software. Van Valkenburgh B, 2007. De´ja`vu: The evolution of feeding Rook L, 1994. The Plio-Pleistocene Old World Canis (Xenocyon) morphologies in the Carnivora. Integrative and Comparative ex gr. Falconeri. Boll. Soc. Paleontologica Italiana 33: 71–82. Biology 1–17. Rook L, Torre D, 1996a. The wolf-event in western Europe and Werdelin L, 1981. The evolution of lynxes. Annales Zoologici the beginning of the Late Villafranchian. Neuse Jahrbuch Fennici 18: 37-71. Geologie unt Palantologie Mh H 8: 495–501. Woodroffe R, Ginsberg JR, 2005. King of the beasts? Evidence Rook L, Torre D, 1996b. The latest Villafranchian, early Galerian for guild redundancy among large mammalian carnivores. In: small dogs of the Mediterranean area. Acta Zool. Cracoviense Ray JC, Redford KH, Steneck RS, Berger J ed. Large Carni- 39: 427–34. vores and the Conservation of Biodiversity. Washington: Island Samuels JX, Van Valkenburgh B, 2008. Skeletal indicators of Press, 230–246. MELORO C: Locomotion in Quaternary carnivores 281

Appendix 1 Long bone indices computed for extant large carnivores

Species Family Tropic Grasslan Mt/Femur BI T/F

Canis aureus Canidae 0.0 1.0 0.42 1.01 0.97

Canis latrans Canidae 0.0 1.0 0.43 1.03 1.01

Canis lupus Canidae 0.0 1.0 0.44 1.00 0.98

Cerdocyon thous Canidae 0.0 1.0 0.43 0.92 1.00

Lycaon pictus Canidae 0.0 1.0 0.46 1.06 1.00

Nyctereutes procyonoides Canidae 0.0 0.0 0.37 0.87 0.99

Chrysocyon brachyurus Canidae 0.0 1.0 0.47 1.09 1.10

Acinonyx jubatus Felidae 0.0 1.0 0.44 1.00 1.00

Puma concolor Felidae 1.0 1.0 0.31 0.84 0.89

Leptailurus serval Felidae 0.0 1.0 0.41 0.97 0.99

Lynx lynx Felidae 0.0 1.0 0.46 0.98 0.98

Lynx rufus Felidae 0.0 1.0 0.42 0.94 0.94

Panthera leo Felidae 0.0 1.0 0.35 0.90 0.86

Panthera onca Felidae 1.0 1.0 0.32 0.79 0.82

Panthera pardus Felidae 1.0 1.0 0.40 0.82 0.92

Panthera tigris Felidae 1.0 1.0 0.35 0.81 0.83

Crocuta crocuta Hyaenidae 0.0 1.0 0.36 1.04 0.77

Parahyaena brunnea Hyaenidae 0.0 1.0 0.39 1.12 0.79

Hyaena hyaena Hyaenidae 0.0 1.0 0.40 1.14 0.87

Ursus americanus Ursidae 0.0 0.0 0.20 0.86 0.75

Ursus arctos Ursidae 0.0 0.0 0.22 0.92 0.65

Ursus maritimus Ursidae 0.0 0.0 0.25 0.76 0.70

Mt/F = metatarsus length/femur length; BI = Brachial Index that is Radius length/Humerus length; T/F = Tibia length/ Femur length. Data source Bertram and Biewener (1990); Janis and Wilhem 1993, Christiansen (2002). Adaptations to tropical or grassland environments as in Ortolani and Caro (1996): 0.0 = Non adapted; 1.0 = adapted.

References predators in the Tertiary? Dances with wolf avatars. Journal of Mammalian Evolution 1: 103–125. Bertram JEA, Biewener AA, 1990. Differential scaling of the long Ortolani A, Caro TM, 1996. The adaptive significance of color bones in the terrestrial Carnivora and other mammals. Journal patterns in carnivores: phylogenetic tests of classic hypotheses. of Morphology 204: 157–169. In: Gittleman JL ed. Carnivore Behavior, Ecology, and Evolution. Janis CM, Wilhem PB, 1993. Were there mammalian pursuit Vol 2. Ithaca, IL: Cornell University Press, 132–186.

282 Current Zoology Vol. 57 No. 3

Appendix 2 Long bone indices computed for Plio-Pleistocene large carnivores Species Family Mt/F BI T/F Localitites References Canis etruscus Canidae 0.49 0.93 1.11 Olivola Meloro, 2003 Canis arnesis Canidae 0.42 0.96 1.00 Tasso but F estimated Meloro, 2003 Lycaon falconeri Canidae 0.49 0.82 0.99 Pirro but H, F, R estimated Meloro, 2003 Canis mosbachensis Canidae 0.46 0.98 1.08 UntermaBfeld Sotnikova, 2003 Canis aff. arnensis Canidae 0.42 1.02 1.02 L'Escale Sotnikova, 2003 Canus lupus_Pleist Canidae 0.44 0.97 1.13 G. Jaruens but F estimated Ballesio, 1979 Homotherium crenatidens Felidae 0.34 1.03 0.86 Pirro and Senéze Sardella, 1994 Megantereon cultridens Felidae 0.32 0.80 0.89 Pirro, Argentario, St.Vallier, Valdarno Sardella, 1994 Acinonyx pardinensis Felidae 0.41 0.87 0.96 Olivola, Casa Frata posterior bones; St Ficcarelli, 1984; Argant, Vallier anterior 2004 Lynx issiodorensis Felidae 0.46 0.77 1.10 Tasso Meloro, 2003 Lynx spelaeus Felidae 0.40 1.02 0.90 Observatoire Testu, 2006 Panthera leo_Pleist Felidae 0.33 0.99 0.79 Equi Del Campana, 1947 Panthera gombaszoegensis Felidae 0.32 0.92 1.04 Valdarno and H from UntermaBfeld Del Campana, 1916; Hemmer, 2001 Pachycrocuta brevirostris Hyaenidae 0.37 0.88 0.74 China but Mt is from Valdarno Turner and Antón, 1996; Meloro, 2003 Crocuta crocuta_Pleist Hyaenidae 0.30 0.90 0.75 G. Jaruens Ballesio, 1979 Chasmaporthetes lunensis Hyaenidae 0.42 1.09 0.88 as in C. ossifragus Berta, 1981 Ursus minimus Ursidae 0.20 0.93 0.73 Gaville Berzi, 1966 Ursus etruscus Ursidae 0.19 0.90 0.67 St. Vallier but Mt from Valdarno Argant, 2004; Meloro, 2003 Ursus spelaeus Ursidae 0.20 0.76 0.67 Equi but Mt from Potocka zijalka Francassi, 1920; Withalm, 2004 Ursus arctos_Pleist Ursidae 0.20 0.89 0.72 Malaspino Koby, 1945 Mt/F = metatarsus length/femur length; BI = Brachial Index that is Radius length/Humerus length; T/F = Tibia length/ Femur length; Humerus = H; Femur = F; Radius = R; Mt = III Metatarsus.

References Untermaßfield bei Meiningen (Thüringen), Teil 3. Monographien des Römisch-Germanischen Zentralmuseums Argant A, 2004. Les Carnivores du gisement Pliocéne final de Mainz 40: 699–771. Saint-Vallier (Dróme, France). The Carnivores of the late Plio- Koby F, 1945. Un squelette d'ours brun du pléistocène italien. cene site of Saint-Vallier (Dróme, France). Geobios 37: Verhandlungen der Naturforschenden Gesellschaft in Basel, 68 S133–S182. and 71. Ballesio R, 1979. Le gisement Plèistocéne Supérieur de la grotte Meloro C, 2003. Canis etruscus, Canis arnensis, Canis (Xenocyon) de Jaurens a Nespouls, Corrèze, France: I Canidae et falconeri: ecologia ed evoluzione. First degree thesis in Verte- Hyaenidae. Nouv. Arch. Mus. Hist. nat. Lyon 17: 25–55. brate Palaeontology. Università degli Studi di Napoli “Fede- Berta A, 1981. The Plio-Pleistocene hyaena Chasmaporthetes rico II”. ossifragus from Florida. Journal of Vertebrate Palaeontology 1: Sardella R, 1994. Sistematica e distribuzione stratigrafica die 341–356. macairodonti dal Miocene superiore al Pleistocene. PhD Thesis. Berzi A, 1966. L’orso di Gaville nel Valdarno Superiore. Universities of Modena, Bologna, Firenze, Roma “La Sapienza”. Palaeontographia Italica LX: 1–31. Sotnikova MV, 2003. Remains of Canidae from the lower Pleis- Del Campana D, 1916. Nuove ricerche sui felini del Pleistocene tocene of Untermassfeld. In Kalkhe RD et al. ed. Das italiano. Palaeontographia Italica XXI, XXII: 1–33. Pleistozän von Untermaßfield bei Meiningen (Thüringen), Teil Del Campana D, 1947. Carnivori quaternari della Tecchia e della 3. Monographien des Römisch-Germanischen Zentralmuseums caverna di Equi nelle Alpi Apuane (mustelidi, canidi, felidi). Mainz 40: 607–632. Palaeontographia Italica XLIV: 1–42. Testu A, 2006. Etude paléontologique et biostratigaphique des Ficcarelli G, 1984. The Villafranchian cheetahs from Tuscany and Felidae et Hyaenidae pléistocènes de l’Europe méditerranéenne. remarks on the dispersal and evolution of the genus Acinonyx. PhD thesis Universite de Perpignan. Palaeontographia Italica LXXIII: 94–103. Turner A, Antón M, 1996. The giant hyaena Pachycrocuta brevirostris Fracassi G, 1920. Ricerche sulle ossa di orso della grotta di Equi (Mammalia, Canrivora, Hyaenidae). Geobios 29: 455–468. in Lunigiana. Palaeontographia Italica XXVI: 157–180. Withalm G, 2004. Analysis of the Cave Bear Metapodial Bones Hemmer H, 2001. Die feliden aus dem epivillafranchium von from Potocka zijalka (Slovenia). Mitt. Komm. Quartärforsch. Untermaßfield. In: Kalkhe RD et al. ed. Das Pleistozän von Österr. Akad. Wiss. 13: 149–160. MELORO C: Locomotion in Quaternary carnivores 283

Appendix 3 Data source for the estimated long bone in fossil taxa Species Bone Equation source Radius Estimate from equation in Bertram and Biewener, 1990_independent is D_LL

Canis arnensis Femur Estimate from equation in Bertram and Biewener, 1990_independent is D_AP Estimate from equation: y = 0.941x + 9.5736; Tibia R2 = 0.980 where x is femur length Humerus Estimate from equation in Bertram and Biewener, 1990_independent is D_AP Radius Estimate from equation in Bertram and Biewener, 1990_independent is D_LL Lycaon falconeri Estimate from equation: y=1.0416x-7.1575; Femur R2 = 0.980 where x is tibia length Canis lupus Femur Estimate from Mt/F in Janis and Wilhem,1993. The Mt value is in Ballesio, 1979 Estimate based on Bertram and Biewener, 1990 only Feliformia, Equation y = 1.228x - 23.163; Femur R2 = 0.960 where x=tibia L Estimate based on Bertram and Biewener, 1990 only Feliformia, Equation: log y = 0.7107 logx + Chasmaporthetes lunensis Radius 1.4086; R2 = 0.884 where x= Radius D_LL Estimate based on Bertram and Biewener, 1990 only Feliformia,Equation: y = 1.0151x + 10.69; Humerus R2 = 0.9137 where x=radius length Estimate based on Betram and Biewener, 1990 only Feliformia, Equation: y = 0.8428x + 6.2563; Acinonyx pardinensis Humerus R2 = 0.958 where x = Humerus length Estimate based on Bertram and Biewener, 1990 only Feliformia, Equation: y = 1.228x - 23.163; Femur 2 Lynx spelaeus R = 0.9602 where x is Tibia length Humerus Estimate from equation in Bertram and Biewener, 1990 where D_AP is independent Radius Estimate from equation in Bertram and Biewener, 1990 where D_LL is independent Panthera gombaszoegensis Femur Estimate from equation in Bertram and Biewener, 1990 where D_AP is independent Megantereon cultridens Humerus Estimate from equation in Bertram and Biewener, 1990 where D_LL is independent When equation is not specified it means that it was reported in the reference and not modified. D_LL = mediolateral length at the midshaft. D_AP = anteroposterior length at the midshaft

References de Jaurens a Nespouls, Corrèze, France: I Canidae et Hyaeni- dae. Nouv. Arch. Mus. Hist. nat. Lyon 17: 25–55. Bertram JEA, Biewener AA, 1990. Differential scaling of the long Janis CM, Wilhem PB, 1993. Were there mammalian pursuit bones in the terrestrial Carnivora and other mammals. Journal predators in the Tertiary? Dances with wolf avatars. Journal of of Morphology 204: 157–169. Mammalian Evolution 1: 103–125. Ballesio R, 1979. Le gisement Plèistocéne Supérieur de la grotte