The Calcaneum On the Heels of Marsupial Locomotion

Transcription

1 J Mammal Evol (2009) 16:1 23 DOI /s ORIGINAL PAPER The Calcaneum On the Heels of Marsupial Locomotion Mina Bassarova & Christine M. Janis & Michael Archer Published online: 9 September 2008 # Springer Science + Business Media, LLC 2008 Abstract The potential for making functional interpretations from a single postcranial element for marsupials was investigated through morphometric analysis of the calcanea of 61 extant species from Australia and New Guinea. Extant species were grouped into locomotor categories and a canonical variates analysis was carried out on measurements of their calcanea. A relationship between measurements of the calcanea and the locomotor behavior of species was found, allowing for prediction of locomotor behavior from calcaneum morphometrics. This was applied to fossil marsupial taxa, from early late Miocene/?Pliocene deposits at Riversleigh, in an attempt to determine their locomotor behavior. Hopping (saltatorial) taxa are distinguished from quadruped terrestrial taxa and taxa capable of climbing (arboreal and scansorial) by their relatively longer tuber calcis and wider calcaneal head, by their dorso-ventrally thicker calcaneal head, and by their calcaneocuboid facet being less steeply angled antero-posteriorly. Taxa capable of climbing are distinguished from quadruped terrestrial taxa by their shorter tuber calcis relative to the calcaneal head and by their smaller calcaneo-astragalar facet. The locomotor categories distinguished in this study (arboreal/ scansorial, quadruped terrestrial, and saltatorial) highlight differences between species in their use of available substrates and thus are informative with regards to the structural components of their habitat. The results of this analysis can be used, in combination with other data, to make inferences about the habitats of paleocommunities at Riversleigh through the Miocene. The calcaneum is a dense and very robust element and, therefore, has a good chance of being preserved. This method provides a quick and easy way of inferring locomotion and has a wide potential for application to many fossil deposits because it requires only a single element. Keywords Morphometric functional analysis. Fossils. Riversleigh. Canonical variates analysis M. Bassarova (*) : M. Archer Vertebrate Palaeontology Laboratory, School of Biological, Earth and Environmental Science, University of New South Wales, Kensington, Sydney, NSW 2052, Australia mina.bassarova@unswalumni.com C. M. Janis Department of Ecology and Evolutionary Biology, Brown University, Providence, RI 02912, USA Introduction The locomotion of animals reflects the substrates on which they move. The basic media or substrates available to mammals for locomotion are water (for aquatic mammals such as the platypus), air (for chiropterans), the ground (for bipedal or quadrupedal

2 2 J Mammal Evol (2009) 16:1 23 terrestrial mammals), or vegetation (for arboreal and semi-arboreal, or scansorial, mammals). Species may be specialized for locomotion on only one substrate, or they may utilize more than one to varying degrees. Locomotor behavior of individual animals reflects habitat use, and at the community level, locomotor behavior reflects the general habitat/vegetation structure inhabited by that community (Damuth 1992). The basic mammalian morphotype is to be small and generalized, with a scansorial mode of locomotion (i.e., adept at moving both on the ground and within vegetation, much like a present-day squirrel). Smaller mammals (under 5 kg) often lack specific adaptations for generalized climbing, while larger forms may exhibit a greater degree of morphological specialization (see Cartmill 1985). Likewise, few small mammals show specialized adaptations for running ( cursorial adaptations); rabbits show some skeletal modifications but to nowhere near the extent seen in large ungulates. However, small hopping mammals are fairly distinctive, at least in their relative limb proportions and elongated hindlimbs. At larger body sizes animals show more distinctive morphological specializations for particular locomotor behaviors, because of allometric constraints on skeletal design (Bertram and Biewener 1990). Locomotor behavior, especially in larger and more specialized animals, is reflected in their skeletal morphology, and thus locomotion can be inferred from an animal s postcranial remains. The traditional approach to inferring the behavior of extinct animals is to compare their anatomy with those of other animals (preferably close relatives) of similar form, with the assumption that function follows form (Walker 1974; Hickman 1988). This type of approach used to be largely qualitative, but with the rise of personal computers over the past few decades such studies are usually quantitative (see Janis et al. 2002). A prime assumption here is that anatomical form should be biomechanically linked to the presumed function, so that anatomical similarities of extinct animals to extant ones are not just the spurious reflections of phylogenetic legacy. Biomechanical principles also help us to infer the behavior of extinct animals that lack close living relatives, such as dinosaurs. However, whether qualitative or quantitative, this type of comparison of extinct forms with extant ones is inevitably inference of a particular behavior rather than direct evidence. Behavior certainly may precede morphological adaptations in evolution (for example, it is doubtful that if polar bears were known only from fossils that their swimming behavior would be deduced), and animals may be capable of doing things that nobody would ever predict from their specialized morphology (e.g., goats climbing trees). Furthermore, the feeding, postural and manipulative behavior of an animal may influence the extent to which overlap into different types of locomotor behavior occurs (Argot 2001). Therefore, in considering applications to the fossil record, broader locomotor categories (e.g., simply arboreal rather than the precise mode of climbing) may be useful than more narrowly-defined ones in making comparisons of different types of animals over a range of paleocommunities. Thus, the extent of detail necessary, in terms of describing locomotor behavior, varies with the particular aims of a study and is subject to the limitations of the available evidence. The form of skeletal elements reflects, to varying degrees, the loading placed on the skeleton by locomotion (Szalay 1994; Bishop 1997) and is constrained by biomechanical principles. Locomotor behavior thus can be inferred from skeletal features. Among the studies that demonstrate this are: Oxnard s (1968) canonical analyses of forelimb and shoulder measurements, which distinguish arboreal and nonarboreal forms of locomotion in mammals; Walker s (1974) investigation of correlation between skeletal anatomy and locomotion in primates; Jenkins and McLearn s (1984) exploration of the relation between foot structure and locomotor use in climbing animals; Van Valkenburgh s (1987) multivariate analyses of locomotor behavior of carnivore species; the biomechanical and multivariate analyses of macaque calcanea by Fa-Hong et al. (1993); Szalay s (1994) research into morphological and functional relationships in marsupial pedal and tarsal elements; Lemelin s (1999) and Argot s (2001) functional analyses of didelphid feet and forelimbs, respectively; Szalay and Sargis (2001) study of marsupial postcranial form function relationships; the statistical comparisons of Janis et al. (2002) of camelid and ruminant artiodactyl limb anatomy as it relates to locomotion; Argot s (2002) functional adaptive analysis of marsupial hindlimbs; and Youlatos (2003) examination of primate calcaneal features. The aims of this study were to find a relationship between the known locomotor behavior of extant

3 J Mammal Evol (2009) 16: Australo-Papuan marsupial species and measurements of their calcanea and to use these relationships to determine the locomotor behavior of fossil marsupial taxa from measurements of their calcanea. The fossil taxa are from Camel Sputum, Encore, Mike s Menagerie, and Quantum Leap Sites from the Riversleigh World Heritage Area fossil deposits, northwestern Queensland, Australia (Archer et al. 1989, 1997; Creaser 1997). The calcaneum was chosen for this study for a number of reasons. Firstly, it is very robust and easily identifiable (in terms of identification of tarsal elements) and therefore provides the largest sample size for each fossil site compared to other elements of the foot. Secondly, the fact that it is a foot element is important because hindlimbs, through necessary adaptations to particular substrates, are more restricted in function than forelimbs, which are additionally used for habitat exploration, food manipulation, and social activities (Szalay 1994). Thirdly, the fact that the calcaneum is so robust not only provides for a better sample size (because it is more likely to be preserved), but also indicates that it has important load-bearing functions. The load-bearing mechanics of the hindfoot are associated with locomotor behavior (Szalay 1994), and the calcaneum is suited to high tensile, bending and compressive forces, providing a firm support for the body weight (Hall and Shereff 1993). The calcaneum has a very important role as the lever arm for the gastrocnemius and soleus muscles, which allow strong and rapid foot retraction. It also is involved, along with adjoining foot elements, in rotation, flexion/extension, and adduction/abduction (sideways inward and outward) movements of the foot. The contour and orientation of the articulating surfaces of the calcaneum are particularly informative in terms of indicating the extent of these movements (Sarrafian 1993). For example, cursorial and saltatorial locomotions require stability during high-speed movement over ground, which results from suppressed rotation of the foot through restricted lateral movement at the calcaneocuboid articulation and close contact between the astragalus and calcaneum (Bishop 1997). Arboreal locomotion, on the other hand, involves specific adaptations for greater freedom of movement that allow climbing on the uneven and potentially unstable surfaces of vegetation. Thus, measurements of these articulating surfaces are potentially useful in determining locomotor behavior. Ideally, the entire foot complex/unit should be used for such a study because articulating bones are closely adapted functionally for particular bioroles (Szalay 1994). However, due to the nature of most fossil deposits, associated or articulated elements rarely occur, making it impossible to determine which elements belong together as part of one individual s functional unit. Materials and methods Specimens Specimens of modern marsupials examined (Appendices A C ) are from the collections of the Australian Museum ( A, M, P or S prefix), the South Australian Museum ( SAM prefix), and the University of New South Wales ( UNSWZ and AR prefixes). A list of the specimens belonging to living species measured and the raw measurement data (each measurement taken three times and averaged) are provided in Appendices A C. Appendix D lists fossil specimens and their raw measurements. Fossil specimens from the Queensland Museum collection ( QMF prefix) were collected and prepared by staff, students, and volunteers from the University of New South Wales. Fossil calcanea used in this study could not be assigned to species already described, because most Riversleigh fossil species have been described on the basis of their dentition, which in most cases is not associated with postcranials. For this reason, fossil calcanea were assigned to groups based on similarity in size and morphology; each group is an unnamed morphotype. As long as the number of morphotypes can be obtained, identification of the actual species involved is not necessary (see discussion of ecomorphs by Damuth 1992). Terminology and abbreviations for calcaneum morphology follow Szalay (1982, 1994). Measurements and variables Table 1 describes the raw calcaneal measurements and abbreviations (as they appear in the Appendices). Figure 1 illustrates the raw measurements, which were taken with Brown and Sharpe SO 9001 digital calipers (TESA Cal 1P65) to the nearest 0.01 mm.

4 4 J Mammal Evol (2009) 16:1 23 Table 1 Descriptions of raw measurements for calcanea Abbreviations (as in Szalay 1994): CLAJP continuous lower ankle joint pattern, CaCu calcaneocuboid, CaFi calcaneofibular, cflf calcaneofibular ligament facet Raw measurement description [abbreviation] Tuber length (from posterior end of calcaneum to posterior end of CLAJP in dorsal view) [ONE] Tuber width (maximum width in dorsal view) [TWO] Calcaneal head length (from posterior end of CLAJP to anterior end of CaCu in dorsal view) [THREE] Calcaneal head width (maximum medio-lateral width in dorsal view) [FOUR] Calcaneocuboid maximum width (medio-lateral direction in distal view) [FIVE] Region of the calcaneofibular maximum length (in antero-posterior direction) [SIX] Region of the calcaneofibular maximum width (in dorso-plantar direction) [SEVEN] Lateral calcanal head length (from posterior end of CaFi/cflf to anterior end of CaCu in lateral view) [EIGHT] Lateral calcaneal head width (maximum width of calcaneal head in dorso-plantar direction) [NINE] Entire calcaneum length (in lateral view) [TEN] Medial tuber length (from posterior end of CLAJP to posterior end of calcaneum in medial view) [ELEVEN] Medial calcaneal head length (from posterior end of CLAJP to anterior end of calcaneum in medial view) [TWELVE] Calcaneocuboid maximum height (dorso-plantar direction in distal view) [THIRTEEN] Region of the calcaneofibular protrusion (distance of outward protrusion from tuber in dorsal view) [FOURTEEN] The choice of measurements covers all surfaces of the calcaneum, incorporating the dimensions of articulating surfaces. The fibula does not articulate with the calcaneum in all species (Szalay 1994). Where an area for articulation is not present, there is a calcaneofibular ligament facet. Measurements of this facet or of a calcaneofibular articulation are referred to here as region of the calcaneofibular. Ratios were created from the fourteen raw measurements (Table 2; the average of the ratios was used in cases where multiple specimens of a species were used) and were subsequently used as the variables in Fig. 1 Illustration of raw measurements for calcanea. A is dorsal view, B is lateral view, C is anterior view (see Table 1 for measurement description and abbreviations).

5 J Mammal Evol (2009) 16: Table 2 Variables used in canonical variates analysis Ratios of raw measurements Variable name Description of variable Variables are derived from raw measurements TWO/ONE TALONID Talonid shape THREE/FOUR HEAD Shape of dorsal surface of calcaneal head THREE/ONE HEADTAL Calcaneal head length relative to talonid length THIRTEEN/FIVE CACU Calcaneocuboid shape SIX/SEVEN CAFI Shape of calcaneofibular region EIGHT/NINE LATHEAD Dorso-plantar thickness of calcaneal head relative to lateral length TWELVE/EIGHT MEDLAT Medio-lateral orientation of calcaneocuboid FOURTEEN/TEN CAFI/ TOTAL Extent of outward protrusion of region of the calcaneofibular the statistical analysis. The reasons for this were as follows: firstly, to clarify shape aspects of the bones (described in Table 2); secondly, comparisons between groups are easier when values are not absolute; and thirdly, to reduce the number of variables relative to species, as the analysis requires. Allometric relationships may confound morphometric analyses. Ratio variables were regressed against body mass for the extant marsupials studied here to determine whether there was a significant correlation (which indicates allometric relationship). No correlation was found, thus allometry is not expected to confound the results. Locomotor categories The present study distinguishes three categories of locomotion for Australian and New Guinean marsupials: (1) arboreal/scansorial, includes species that spend the majority of their time on vegetation (feeding and sleeping) and species that spend a considerable proportion of time on the ground but are capable climbers; (2) terrestrial, includes quadruped species restricted mostly to the ground or underground; and (3) saltatorial or hopping, includes bipedal hopping species. These categories serve to highlight differences between species in their use of available substrates and thus to illustrate their distribution among the structural components of the habitat. Arboreal and scansorial taxa are included together in a single category here in an effort to avoid a priori misclassification based on limited behavioural information, a possibility identified by Van Valkenburgh (1987). Some animals may be intermediate between locomotor categories, but for the purposes of analysis have to be allocated to one group only. Species were assigned to these locomotor categories based on field observations of their behavior recorded in literature (see Table 3). Small mammals, due to their small body size, are often confronted with uneven substrates and (for them) steeply inclined surfaces of tree roots, fallen logs, and rocks (Jenkins 1974). For this reason, even those that spend most of their time on the ground are actually able to climb. Small mammals (under approximately 500 g body mass, for the purposes of this study) therefore have a versatile locomotor pattern and in this study are classified as arboreal/scansorial, because they have the functional potential to locomote on the ground or in vegetation. The issue of size and scaling is also important, with large animals being more specialized (see earlier comments on this). Data A list of the modern species used in the analysis, the locomotor categories they were assigned to, and the corresponding values for each variable used for the analysis are provided in Table 3. Fossil morphotypes entered into the analysis and corresponding values for each variable are listed in Table 4. Measurements were taken of as many specimens as were available for each modern marsupial species. The means of these measurements for each species were used in calculating the ratio variables used in the analysis. A test was devised to make sure that withinspecies variation was less than between-species variation. For the modern northern brown bandicoot (Isoodon macrourus), individuals of one sex were entered into the analysis as known and individuals

6 6 J Mammal Evol (2009) 16:1 23 Table 3 Data for modern species used in canonical variates analysis Order Species Locomotion Body mass (g) TALONID HEAD HEADTAL CACU CAFI LATHEAD MEDLAT CAFI/TOTAL DIP Phascolarctos cinereus a Arboreal 5, Petaurus norfolcensis a Arboreal b Pseudocheirus peregrinus a Arboreal Pseudochirops archeri a Arboreal 2, Spilocuscus maculatus a Arboreal c 3, Trichosurus vulpecula a Arboreal 2, Onychogalea fraenata a Hopper 5, Aepyprymnus rufescens a Hopper 3, Burramys parvus Terrestrial b Hemibelideus lemuroides Arboreal Petauroides volans Arboreal 1, Petaurus breviceps Arboreal b Phalanger gymnotis Scansorial c 2, Phalanger orientalis Arboreal c 2, Pseudochirulus cf. canescens Arboreal b,c Lasiorhinus latifrons a Terrestrial 25, Vombatus ursinus Terrestrial 26, Bettongia gaimardi cuniculus Hopper 1, Bettongia penicillata Hopper 1, Dendrolagus dorianus Arboreal c 10, Dendrolagus goodfellowi Arboreal c 8, Dendrolagus inustus Arboreal c 10, Dendrolagus lumholtzi Arboreal 6, Dendrolagus matschiei Arboreal c 9, Dendrolgus scottae Arboreal c 9, Dorcopsis afrata Hopper Dorcopsis muelleri Hopper c 5, Hypsiprymnodon moschatus Scansorial Lagorchestes hirsutus Hopper 1, Macropus agilis Hopper 15, Macropus eugenii Hopper 6, Macropus fuliginosus Hopper 40, Macropus giganteus Hopper 49, Macropus parma Hopper 5, Macropus parryi Hopper 13, Macropus rufus Hopper 46, Onychogalea unguifera Hopper 6,

7 J Mammal Evol (2009) 16: Table 3(continued) Order Species Locomotion Body mass (g) TALONID HEAD HEADTAL CACU CAFI LATHEAD MEDLAT CAFI/TOTAL Petrogale brachyotis Hopper 4, Petrogale mareeba Hopper 8, Petrogale penicillata Hopper 7, Petrogale persephone Hopper 6, Potorous tridactylus Hopper 1, Wallabia bicolor Hopper 15, DAS Dasyuroides byrnei a Terrestrial b Dasyurus maculatus a Scansorial 5, Antechinus agilis Scansorial b Antechinus stuartii Scansorial b Dasyurus viverrinus Scansorial 1, Parantechinus apicalis Terrestrial b Phascogale tapoatafa Arboreal b Sarcophilus harrisii Scansorial 8, Sminthopsis murina Terrestrial b Thylacinus cynocephalus Terrestrial 25, PER Isoodon macrourus a Terrestrial 1, Echymipera kalubu Terrestrial c Echymipera rufescens Terrestrial 1,000 d Isoodon obesulus Terrestrial Macrotis lagotis Terrestrial 1, Perameles nasuta Terrestrial Peroryctes raffrayana Terrestrial NOTO Notoryctes typhlops Terrestrial b DAS Dasyuromorphia, PER Peramelemorphia, DIP Diprotodontia, NOTO Notoryctemorphia a Average of more than one specimens b Species placed in arboreal/scansorial category in analysis due to having <500 g body mass c Flannery (1995) as reference; remainder use Strahan (1998) d Weight was inferred based on body length

8 8 J Mammal Evol (2009) 16:1 23 Table 4 Data for fossil morphotypes used in canonical variates analysis Site Order Morphotype TALONID HEAD HEADTAL CACU CAFI LATHEAD MEDLAT CAFI/TOTAL CS DIP CSB CSI CSS CSS CST CSV CSC CSJ CSU CSQ CSR DAS CSP PER CSa CSd CSe CSf CSg CSh CSm CSn CSo CSma CSx CSy CSz MM DIP MM MM MM PER MM MM MM MM MM Ec DIP R.flanneryi EcG EcH EcJ

9 J Mammal Evol (2009) 16: Table 4(continued) Site Order Morphotype TALONID HEAD HEADTAL CACU CAFI LATHEAD MEDLAT CAFI/TOTAL EcK EcL EcM EcN EcO Trichosurus DAS EcE EcF PER EcA EcB EcC EcD QL DIP Nambaroo sp PER QLA QLB QLC QLD QLE QLF QLG DAS Dasyuromorphia, PER Peramelemorphia, DIP Diprotodontia, NOTO Notoryctemorphia, CS camel sputum, MM Mike s menagerie, Ec encore, QL quantum leap

10 10 J Mammal Evol (2009) 16:1 23 of the other sex as unknowns (see Appendix C for specimen list and raw data). The results verify that unknowns are placed in the appropriate category by the analysis. The northern brown bandicoot was chosen for its relatively larger sample size and because the average body mass of males is almost twice that of females (Gordon 1998), thus exhibiting some degree of sexual dimorphism. Phylogeny, form and function Form is phylogenetically influenced within taxa and functional adaptations develop under these influences (Szalay 1982; Hickman1988); therefore, function and phylogeny often coincide. This study is concerned only with the relationship between form and function. The results of the analysis undertaken will reveal whether or not function (or locomotor behavior) can be inferred from morphometrics, irrespective of phylogeny. If species group together within the same locomotor category, despite being only distantly related, or conversely, if closely related taxa are distinguished on the basis of locomotor behavior, then the interplay of function and morphology may be considered sufficient for the purposes of this study. Statistical analysis and assumptions A canonical variates analysis (multiple discriminant analysis) was carried out to determine whether the calcaneal measurements distinguish species on the basis of their locomotor behavior. Canonical variates analysis is a multi-group discriminant analysis that highlights differences among groups (locomotor categories in this case) by spacing out the means of the groups to the maximum extent through construction of linear functions (Oxnard 1968). The separation of the groups is maximized relative to the variation within each group and the analysis calculates the likelihood of an unknown (in this case a fossil) belonging to each of the groups (Reyment et al. 1984). The analysis was performed using SYSTAT (for Windows) version 7.0 (1997) and graphs were produced with PAST (Palaeontological Statistics) version 1.27 (Hammer et al. 2004). Assumptions of canonical variates analysis include normality of data, equivalence of variances/covariances across groups, and minimal multicollinearity of the variables. There is no direct test for multivariate normality, but if all variables exhibit univariate normality then departures from multivariate normality are usually inconsequential (Hair et al. 1998). Each variable was examined for normality statistically for the modern sample (not shown here). All variables, except TALONID and CAFI, are normally distributed at the 0.01 probability level. A couple of outliers are responsible for the TALONID and CAFI variables not being normally distributed (Dendrolagus dorianus and D. matschiei for TALONID and Hypsiprymnodon moschatus and Sminthopsis murina for CAFI). This is most likely due to measurement error. The analysis was carried out excluding these taxa and results (not shown here) were effectively the same as those presented here; therefore, this deviation from univariate normality is considered inconsequential. A statistical test for equality of covariance matrices is Box s M test, but it is very sensitive to departures from normality. A violation of the assumption of equivalence of covariance matrices across groups has minimal impact if the groups are of approximately equal size if the largest group size divided by the smallest group size is less than 1.5 (Hair et al. 1998). The arboreal/scansorial category in this analysis consists of three times as many species as the terrestrial category and almost 1.5 times as many species as the saltatorial category. The analysis was re-run with the arboreal/scansorial and saltatorial categories reduced to group sizes equal to that for the terrestrial category and results were very similar to those presented here for the larger groups of arboreal/ scansorial taxa and hopping taxa. Canonical variates analysis is a robust technique in any case, that can tolerate deviations from these assumptions to some extent (Turbón et al. 1997). One way to assess the impact of multicollinearity of variables is to look at tolerance values for the variables, where tolerance is one minus the proportion of the variable s variance explained by the other predictor variables (Hair et al. 1998). Tolerance values approaching one indicate little collinearity, but variables with tolerance values approaching zero are accounted for by other variables in the analysis to a great extent (Hair et al. 1998). Tolerance levels for variables in this analysis are close to one for the CAFI, MEDLAT and CAFI/TOTAL variables, above 0.2 for CACU, TALONID, HEADTAL and LATHEAD variables, and very close to 0.2 for the HEAD variable.

11 J Mammal Evol (2009) 16: This indicates that at least 20% of the variation explained by each variable (in the case of CACU, TALONID, HEADTAL, LATHEAD and HEAD variables) and over 80% (in the case of the remaining variables) is unique to that variable. Results Summary statistics Table 5 shows summary statistics for the locomotor categories. The tuber of the calcaneum is narrowest in saltatorial taxa and widest in arboreal/scansorial taxa, as indicated by average TALONID values for these categories. The dorsal surface of the calcaneal head is approximately square in species of the saltatorial category and it is longer than it is wide in arboreal/ scansorial and terrestrial taxa, as indicated by average HEAD values. HEADTAL values under one are indicative of a longer talonid relative to the calcaneal head. This is seen for saltatorial species. Terrestrial taxa generally exhibit a calcaneal head approximately equal in length to the tuber, and arboreal/scansorial taxa have shorter tubers relative to calcaneal heads. The CACU variable describes calcanoecuboid shape, which approaches a square form in terrestrial species but is wider than it is deep in arboreal/scansorial and saltatorial taxa. The LATHEAD variable describes the dorso-plantar thickness of the calcaneal head, which is seen to be considerably greater in hopping species compared to arboreal/scansorial and terrestrial taxa. The medio-lateral orientation of the calcaneocuboid, or its position relative to the antero-posterior axis of the calcaneum, is described by the MEDLAT variable. Means for this variable indicate that the calcaneocuboid is situated more perpendicular to the anteroposterior axis in saltatorial taxa compared to those with other forms of locomotion. Canonical variates analysis Figure 2 shows the pattern of separation of locomotor categories on the canonical variates plot for modern marsupial species. Canonical variate 1 distinguishes the hopping category from the terrestrial and arboreal/ scansorial categories. This separation is due mainly to the combination of TALONID, LATHEAD, and MEDLAT variables (Table 6), which are indicative of a thinner tuber, dorso-plantarly thicker calcaneal head, and the near perpendicular orientation of the calcaneocuboid relative to the antero-posterior axis of the calcaneum in saltatorial taxa. Canonical variate 2 distinguishes the terrestrial category from the arboreal/ scansorial category and, to a large extent, from the saltatorial category. Variables contributing most to this separation include TALONID, HEAD, HEADTAL, and CACU (Table 6). The tuber of terrestrial taxa is intermediate in width between that of arboreal/ scansorial and saltatorial taxa. The length of the tuber in terrestrial taxa is approximately equal to the length of the calcaneal head, in contrast to arboreal/ scansorial taxa and saltatorial taxa, which have shorter tubers and longer tubers respectively. Terrestrial species also have calcaneocuboids that are approximately square in shape. Figure 3 summarizes the main morphological differences in calcanea among the three locomotor categories. Table 5 Summary statistics for modern species Group Statistic TALONID HEAD HEADTAL CACU CAFI LATHEAD MEDLAT CAFI/TOTAL Arboreal/scansorial (N=30) X OR SD Terrestrial (N=10) X OR SD Hopping (N=21) X OR SD X group means, OR observed ranges, SD standard deviations, N number of species within each group

12 12 J Mammal Evol (2009) 16:1 23 Fig. 2 Canonical variates plot for modern marsupial sample. Morphological features that score positively or negatively high along the axes (and thus distinguish locomotor categories) are indicated along respective axes. Eigenvalues (Table 6) indicate the variation explained by the canonical axes of the plot. The first canonical variate accounts for 77.4% of the variation in calcaneal morphometrics and the second for 22.6%. A global test of differences between multivariate means showed a significant difference between locomotor categories (Pillai s trace=1.398, df=16, 104, P<0.0000). Table 7 shows the classification results. Predicted locomotion for fossil specimens and modern unknowns This study aimed to develop a model with which to allocate subjects (fossil calcanea) to groups (locomotor categories). Table 8 lists fossil morphotypes and their predicted group allocation with associated probabilities, based on the distance of each specimen from the locomotor group centroid on the canonical variates plot. Modern species entered into the analysis as unknowns were grouped correctly (statistical output not shown). This is discussed further below. Discussion The results of this study indicate that calcaneal morphometrics are informative in terms of locomotor function for marsupials. Arboreal/scansorial, terrestrial, and saltatorial locomotor behaviors were distinguished on the basis of calcaneal articular surface shapes and orientations and relative sizes of structural components of the calcaneum. The clearest distinctions are between the saltatorial category and the arboreal/scansorial category. Fossorial behavior was not distinguished in the current study, despite the fact that several species with fossorial and semi-fossorial adaptations (wombats, the bilby, and the marsupial mole) were included. The reason for this is that digging adaptations are usually more pronounced in forelimbs as opposed to hindlimbs, but this study only takes the calcaneum into account. Despite the fact that these animals are known to use their hindlegs to varying extents when digging, the analysis is not sensitive enough to distinguish them as a separate group. The clear separation of the saltatorial category from other locomotor categories may be due partly to

13 J Mammal Evol (2009) 16: Fig. 2 (continued) phylogenetic influences. All species in this category are macropodoids. However, the fact that the tree kangaroos and the musky rat kangaroo, macropodoids of genera Dendrolagus and Hypsiprymnodon, cluster within the arboreal/scansorial category indicates that differences in functional morphology are evident despite phylogenetic influences. The mix of taxa comprising the arboreal/scansorial and terrestrial categories also indicates that not only phylogeny but also function is reflected in morphometric data.

14 14 J Mammal Evol (2009) 16:1 23 Table 6 Canonical discriminant functions Variable Canonical variate 1 2 CONSTANT TALONID 7.99 (1.17) 7.30 (1.07) HEAD 0.69 (0.17) 3.04 (0.73) HEADTAL 2.86 ( 0.68) 6.62 ( 1.59) CACU 2.97 (0.47) 4.48 (0.70) CAFI 0.45 ( 0.16) 0.33 ( 0.12) LATHEAD 4.53 (1.22) 1.69 (0.46) MEDLAT 4.87 ( 0.56) 0.69 (0.08) CAFI/TOTAL 0.81 ( 0.02) ( 0.41) Eigenvalue Variance (%) stability necessary for terrestrial locomotion (Szalay and Decker 1974). The current analysis shows hopping species to have longer, thinner talonids and hopping and terrestrial taxa to have wider, shorter calcaneal heads compared to arboreal and scansorial Values in brackets are standardised coefficients for canonical variates As indicated by the statistical analyses, and as shown in Figure 3, saltatorial taxa have mediolaterally thinner and longer talonids (lower TALONID values), wider calcaneal heads (lower HEAD values) with a dorsally oriented calcaneo-astragalar facet allowing for close contact between the astragalus and calcaneum, and dorso-plantarly thicker calcaneal heads (lower LATHEAD value) relative to other locomotor categories. For species capable of climbing (arboreal and scansorial), calcaneal heads are longer than talonids (higher HEADTAL values) and the area of contact between the calcaneum and the astragalus (the continuous lower ankle joint pattern as in Szalay 1994) is situated in a more medial direction compared to ground-dwelling species (terrestrial and saltatorial). These features in terrestrial species generally appear to be intermediate between saltatorial and arboreal/scansorial species. Additionally, the orientation of the surface of the calcaneocuboid relative to the antero-posterior axis of the calcaneum (indicated by MEDLAT values) in saltatorial species is close to perpendicular, whereas in terrestrial and arboreal/scansorial species it is situated at a more acute angle, or mediodistally. The surface of the calcaneocuboid of terrestrial species differs from that of species with climbing ability in that it is approximately square in shape (higher CACU value). These results are consistent with previous studies. A long tuber has been associated with cursorial habits (see Youlatos 2003 and references within) and a square-shaped calcaneo-astragalar facet with subtalar Fig. 3 Generalized calcaneal features for arboreal/scansorial, terrestrial and saltatorial locomotion. A shows dorsal view; B shows medial view. CaCu calcaneocuboid, CLAJP continuous lower ankle joint pattern.

15 J Mammal Evol (2009) 16: Table 7 Classification results from canonical variates analysis Allocated locomotion category Arboreal/scansorial Terrestrial Hopping Actual locomotion Arboreal/scansorial (30) 100/83 0/17 0/0 Terrestrial (10) 0/20 90/70 10/10 Hopping (21) 0/10 0/0 100/90 Values indicate the percentage of allocations of individuals to each group; values in italics indicate jackknife classification results; values in parentheses are the numbers of species for each group taxa. This is not unexpected given that fast hopping is not unlike cursorial locomotion in the requirement for stability. Stability is provided by a wide calcaneal head region and a secure calcaneo-astragalar articulation (Szalay 1994; Bishop 1997). Climbing, on the other hand, requires inversion and eversion of the foot that is a result of a long, narrow calcaneo-astragalar facet (Youlatos 2003). Additionally, bulbous enlargements of bones near joints are indicative of large forces on the joints (Szalay 1994). The greater dorsoplantar thickness of the calcaneal head in hoppers relative to climbing and quadruped terrestrial taxa is thus interpreted to be related to the great force exerted on the region of the calcaneo-astragalar articulation during hopping. The size and orientation of the calcaneocuboid articulation are important in determining pedal and tarsal movements in mammals. Flexion extension, rotation, and adduction/abduction of the forefoot are a result of force transmission from the calcaneum to the cuboid at this articulation (Szalay 1994). Stability in the feet of terrestrial species is indicated by the dorsoplantarly thicker calcaneocuboid articulation (relative to climbing species) that restricts such movements to some extent. Orientation of the calcaneocuboid also restricts such movements to a greater extent when it is perpendicular to the antero-posterior axis, as seen in hoppers. The more steeply angled this articulation is relative to the antero-posterior axis, the more adduction/ abduction, and perhaps rotation, are possible, as seen in species with climbing ability. The calcaneofibular region did not appear to be informative with respect to mode of locomotion in this analysis. This is probably due to the fact that in the data set used here, measurements were provided for the calcaneofibular ligament facet where a calcaneofibular articulation is absent (and these are in fact two different measurements). The calcaneofibular articulation, where present, is informative in terms of extent of mobility between the crus (tibia and fibula) and the pes (Bishop 1997). This region is more pronounced in terrestrial species relative to climbing species; reduction of the articulating region between the fibula and calcaneum, as seen in Dendrolagus, suggests greater mobility at the articulation (Bishop 1997). Modern species entered into the analysis as unknowns were grouped correctly. They included specimens of female northern brown bandicoots (Isoodon macrourus), to test for possible effects of sexual dimorphism, and juveniles of the common ringtail possum (Pseudocheirus peregrinus), the common spotted cuscus (Spilocuscus maculatus), the common brushtail possum (Trichosurus vulpecula), and the bridled nailtail wallaby (Onychogalea fraenata). Juveniles most likely grouped correctly because the articular patterns in the bones of mammals are established prior to adult size being reached the bone is ossified from an epiphysis at the posterior end of the tuber (Szalay 1994). The issue of immature specimens is really most important for long bone length, which in this case might affect the length of the talonid. One modern species used as a known in the analysis was reallocated to a different locomotor category to that specified for it. The thylacine (Thylacinus cynocephalus), a quadruped terrestrial species, is grouped with saltatorial species in the analysis (indicated by the shorter Mahalanobis distance of this taxon from the saltatorial group centroid full statistical output not shown here). A possible explanation for this is that the thylacine was a cursorial carnivore (Strahan 1998; Paddle 2000) and as such had a functional requirement for stability during high speed movement, as is the case for macropodids. Close contact between the astragalus and calcaneum (sup-

16 16 J Mammal Evol (2009) 16:1 23 Table 8 Percent probabilities for locomotor group allocations for fossil morphotypes Morphotype Taxon Arboreal/scansorial Terrestrial Hopping CSa Peramelemorphian CSB Macropodoid CSC Macopodoid CSd Peramelemorphian Cse a Peramelemorphian CSf a Peramelemorphian CSg a Peramelemorphian CSh a Peramelemorphian CSI Macropodoid CSJ Vombatiform CSM Peramelemorphian CSN a Peramelemorphian CSO a Peramelemorphian CSP Dasyuromorphian CSQ Possum CSR a Possum CSS1 Macropodoid CSS2 Macropodoid CST Macropodoid CSU Vombatiform CSV Macropodoid CSMA Peramelemorphian CSX a Peramelemorphian CSY a Peramelemorphian CSZ a Peramelemorphian R. flanneryi Macropodoid Trichosurus Possum EA Peramelemorphian EB Peramelemorphian EC a Peramelemorphian ED a Peramelemorphian EE a Dasyuromorphian EF a Dasyuromorphian EG Macropodoid EH Macropodoid EJ Macropodoid EK Macropodoid EL Macropodoid EM Macropodoid EN Macropodoid EO Macropodoid MM1 Macropodoid MM2 Macropodoid MM3 Macropodoid MM4 a Peramelemorphian MM5 a Peramelemorphian MM6 Peramelemorphian MM7 Peramelemorphian MM8 a Peramelemorphian Nambaroo sp. 3 Macropodoid QLA a Peramelemorphian QLB Peramelemorphian

17 J Mammal Evol (2009) 16: Table 8(continued) Morphotype Taxon Arboreal/scansorial Terrestrial Hopping QLC Peramelemorphian QLD a Peramelemorphian QLE a Peramelemorphian QLF Peramelemorphian QLG a Peramelemorphian a Calcaneum length under 1 mm, therefore body mass presumed to be under 500 g pressing rotation of the foot) is apparent (seen in the square shape of the calcaneal head) in macropodids and thylacines. It is interesting to note the position of the rock wallabies, wombats, bilby, and tree kangaroos on the canonical variates plot (Fig. 2). These cases indicate that the analysis is sensitive to foot stability and range of motion capabilities as reflected in the morphology of the calcaneum. The four rock wallabies used in the analysis (Petrogale brachyotis, P. mareeba, P. penicillata, and P. persephone) cluster within the hopping category, but are at the edge closest to the arboreal/ scansorial group. Rock wallabies differ from other hopping macropodoids in their ability to move over steeply inclined and uneven rock and cliff surfaces. This requires more flexibility and a greater range of motion in the feet compared to fast hopping on relatively flat surfaces. Within the terrestrial group the two wombats (Vombatus ursinus and Lasiorhinus latifrons) and the bilby (Macrotis lagotis, which is number 3 in Fig. 2) fall out closer to the arboreal/ scansorial group than any of the other terrestrial taxa. This may be explained by their digging behavior. The bilby and wombats use their hindlimbs to kick away soil as they burrow (Wells 1989; Strahan 1998). This action requires a certain amount of flexibility in the ankle that may not occur in non-burrowing terrestrial taxa. The three tree kangaroos within the arboreal/ scansorial group that are closest to the saltatorial cluster (Dendrolagus lumholtzi, D. inustus and D. goodfellowi) are known to hop, whereas the other tree kangaroos used in the analysis (D. matschiei, D. dorianus, and D. scottae), positioned well away from the saltatorial group, are noted as being largely arboreal and capable of walking rather than hopping (Flannery et al. 1996: p. 132, and references within). The method described here of inferring locomotion requires a single element and thus could easily be applied to isolated marsupial calcanea from any fossil site. The fossil calcanea entered into this analysis indicate all macropodoids from Mike s Menagerie, Camel Sputum, and Encore sites were likely to have been saltatorial and most peramelemorphians were probably terrestrial (Table 8), except where species have the ability to climb due to their small size. It is interesting that one peramelemorphian entered into the analysis (morphotype CSa) with a body mass likely to have exceeded 500 g, based on the size of the element, clusters within the arboreal/scansorial category; no modern peramelemorphians are known to climb. All fossil possums and vombatiforms enteredintothisanalysishavebeenclassifiedarboreal/ scansorial. The allocation of fossil taxa from Riversleigh deposits to categories of locomotor behavior will form a component of the data set to be used in future paleoecological studies of these deposits, along with dietary behavior data. These categories of ecological diversity will be compared to those of modern mammalian communities from known habitats. Inferences about the habitats of fossil communities at Riversleigh will be made on this basis and implications for vegetation structure in the Miocene of Riversleigh will be discussed in future publications. Acknowledgements Vital support for the Riversleigh Fossil Project has been provided by the Australian Research Council, Department of Environment and Heritage, Environment Australia, the University of New South Wales, the Queensland Museum, the Australian Museum, Mount Isa City Council, Outback at Isa, Xstrata, CREATE, and many private supporters and volunteers as well as staff and postgraduate students of the University of New South Wales. We thank D. Arena R. Beck, K. Black, P. Brewer, A. Gillespie, H. Godthelp, S. Hand, B. Kear, J. Louys, K. Roberts, K. Travouillon, and V. Weisbecker for discussion, technical advice and assistance. T. Ennis and S. Ingleby kindly allowed access to comparative specimens at the Australian Museum.