MURDOCH RESEARCH REPOSITORY

Transcription

1 MURDOCH RESEARCH REPOSITORY This is the author s final version of the work, as accepted for publication following peer review but without the publisher s layout or pagination. The definitive version is available at Warburton, N. and Prideaux, G. (2010) Functional pedal morphology of the extinct tree-kangaroo Bohra (Diprotodontia: Macropodidae). In: Coulson, G. and Eldridge, M., (eds.) Macropods: The Biology of Kangaroos, Wallabies and Rat- Kangaroos. Collingwood, Victoria, CSIRO Publishing, pp Copyright: 2010 CSIRO Publishing It is posted here for your personal use. No further distribution is permitted.

2 Functional pedal morphology of the extinct tree-kangaroo Bohra (Diprotodontia: Macropodineae) NATALIE M. WARBURTON 1,2 AND GAVIN J. PRIDEAUX 1,3 ABSTRACT Species in the extinct genus Bohra have been allied with living tree-kangaroos (Dendrolagus) on the basis of marked similarities in isolated craniodental and hind limb elements. Prompted by recent discoveries of the first near-complete Bohra skeletons in Pleistocene deposits in caves beneath the central Nullarbor Plain, south-central Australia, we compared the pedal morphology of tree-kangaroos with a range of terrestrial macropodines. Our objectives were to more clearly detail the key functional attributes of the Dendrolagus pes and to assess the likely arboreal adeptness of Bohra. Overall proportions of the calcaneum, talus, cuboid and metatarsals, as well as the morphology of their articular facets, suggest that the Bohra pes was specialised for enhanced mobility and flexibility, and thus well adapted to the functional demands of an arboreal environment. The presence of these Pleistocene species on the Nullarbor Plain indicates that tree-kangaroos had far broader geographic and climatic ranges than hitherto anticipated, and also demonstrates that the Treeless Plain was not always treeless. Key words: macropodines, Pleistocene, arboreal adaptations, Nullarbor Plain, palaeoecology 1 Department of Earth and Planetary Sciences, Western Australian Museum, Locked Bag 49, Welshpool DC, Western Australia, Australia School of Veterinary & Biomedical Sciences, Murdoch University, Murdoch, WA 6150, Australia; n.warburton@murdoch.edu.au 3 School of Biological Sciences, Flinders University, Bedford Park, SA 5042, Australia; gavin.prideaux@flinders.edu.au 1

3 INTRODUCTION MACROPODOIDS (kangaroos and their kin) descended from an arboreal, possum-like ancestor in the Palaeogene Period, possibly during the Eocene Epoch (Burk et al. 1998). Following a postulated origin in the middle Miocene (16 11 million years ago), members of the family Macropodineae (kangaroos and wallabies) radiated into a broad array of terrestrial herbivore niches. The hind limb morphology of macropodines is among the most specialised of all mammals, reflecting their unique method of saltatorial locomotion (e.g., Alexander and Vernon 1975; Flannery 1982). The pes (hind foot) is characterised by loss of the hallux, marked reduction of the syndactylous metatarsals II III, and enlargement and elongation of metatarsals IV V. Articulations between tarsal elements have been modified to enhance the hinge action at the ankle while restricting rotational movements (e.g., Szalay 1994). Given this evolutionary trajectory and the functional adaptive constraints posed by the terrestrial macropodine bauplan, it is remarkable that one lineage (tree-kangaroos) returned to the arboreal realm during the late Miocene (11 5 million years ago), dispersing into New Guinea before its isolation from the Australian mainland (e.g., Hall 2001). Compared with terrestrial macropodines, extant tree-kangaroos (species of Dendrolagus) have a shorter, broader pes characterised by stout tarsal and metatarsal bones (Fig. 1). Articular surfaces have been modified for increased flexibility, allowing for greater balance and mobility in an arboreal environment (Flannery 1982; Flannery and Szalay 1982). Of the ten extant species of Dendrolagus, two inhabit tropical forest in northeastern Queensland, while the other eight occupy generally limited distributions in different parts of New Guinea (Flannery et al. 1996). Dendrolagus has been divided into a long-footed group, comprising both Australian species (D. bennettianus, D. lumholtzi) plus D. inustus, and a short-footed group including the remaining New Guinea species (Groves 1982). The long-footed group is 2

4 apparently the more plesiomorphic (Groves 1982; Flannery et al. 1996). Bohra paulae was raised by Flannery and Szalay (1982) to describe a large fossil treekangaroo represented by a few isolated hind limb elements from Wellington Caves in central eastern New South Wales. The fossils were collected in the late nineteenth century, but retained no detailed provenance data. Their age is, thus, uncertain; deposits in these caves range from early Pliocene to late Pleistocene (Dawson and Augee 1997; Dawson et al. 1999). Dawson (2004) described a second species, B. wilkinsonorum, on the basis of a partial maxilla from the late Pliocene Chinchilla Local Fauna (LF) of southeastern Queensland, while Hocknull (2005) identified a calcaneum attributable to the genus from both the Chinchilla LF and the middle Pleistocene Mount Etna cave fills. Two new Pleistocene species of Bohra have recently been identified on the basis of more complete specimens, including cranial and postcranial elements, collected from caves beneath the Nullarbor Plain of south-central Australia (Prideaux et al. 2007; Prideaux and Warburton 2008). Flannery and Szalay (1982) identified several features in common between Bohra and Dendrolagus, including a relatively smoothed calcaneal-cuboid articulation, the transverse orientation of the talar head, the transversely expanded tuber calcanei and the reduced length of the fibular crest of the tibia. This suggested at least a partially arboreal lifestyle for B. paulae and a phylogenetic link with Dendrolagus, views confirmed by the analyses of B. wilkinsonorum (Dawson 2004) and the new Nullarbor species, Bohra illuminata and Bohra sp. nov. In this chapter we present a detailed functional analysis of the pedal elements of Bohra, prompted by the well-preserved and more complete Nullarbor specimens. Our aims are to establish what their morphology tells us about its arboreal aptitude relative to Dendrolagus. We expect this to provide part of the foundation required for the development of an empirically grounded hypothesis for the evolutionary transition from the terrestrial to the 3

5 arboreal realm. We approach this by identifying the unique pedal features of Dendrolagus and Bohra compared with terrestrial macropodoids. We then employ multivariate analyses to establish whether these differences can be quantified morphometrically, and consider the functional significance of different pedal morphologies. MATERIALS AND METHODS Taxonomic comparisons Specimens of an ecomorphologically diverse range of extant macropodines and potoroines (21 species) were examined and measured (Table 1), along with those of three species of Bohra (Table 2). Our analysis focused on the three major elements of the tarsus (calcaneum, talus, cuboid), but also included observations on the tibia and metatarsals. Taxonomic comparisons focused on the relative size and shape of musculo-tendon attachment sites and articular surfaces. As far as is possible, descriptive terminology used follows the Nomina Anatomica Veterinaria (NAV1994). For annotated illustrations of morphological features referred to in the text, refer to Prideaux and Warburton (2008). Linear measurements (recorded to 0.1 mm) were made using fine-point callipers for 36 tarsal dimensions (Appendix 1, Figures 2 4). Measurements included those which have been used in previous studies of the tarsus and those that incorporate features with proposed functional significance (Bishop 1997; Carrano 1997; Argot 2002; Sargis 2002; De Gusta and Vrba 2003; Salton and Szalay 2004). Due to the unequal number of individuals measured for each species (Table 1), we analysed mean measurements. Some sexual dimorphism in body size is apparent but because of the small sample sizes available we combined female and male specimens within groups. Because shape changes with ontogeny, measurements were 4

6 restricted to sub-adult and adult specimens (denoted by fused epiphyses). Raw numerical data is available from NMW. The three species of Bohra represented by tarsal elements were examined: B. paulae (AM F62099, AM F62100), Bohra illuminata (WAM ) and Bohra sp. nov. (WAM ). The complete set of tarsal measurements were only available for Bohra illuminata. A second analysis using a reduced dataset including only calcaneal measurements was undertaken to investigate the morphometric affinities of all three species. Data analysis Preliminary univariate analysis was used to determine basic descriptors: mean, standard deviation and coefficients of variation. Scatter plots demonstrated a strong positive regression between means and their standard deviations, and data were subsequently log 10 -transformed to accommodate this size-dependent variance. All measurements were divided by calcaneal length (CML) to standardise for size, thus, data are indices of relative size. Nine indices proved useful in discriminating taxa (Table 2). Principal components analysis (PCA) was conducted using 36 morphometric variables (Appendix 1) incorporating all extant taxa and Bohra illuminata. Additional analyses on reduced datasets investigated a) variation within macropodines alone, and b) calcaneal measurements only. The first three principal components were extracted in each case as determined by their respective eigenvalues, the percentage of variation explained in each case, and scree plot interpretation. Discriminant grouping analysis was used to determine whether grouping by habitat preference yielded obvious trends in morphological data, and to reveal the likely affinities of the species of Bohra. Habitat preference was classified according to Strahan (1995). The 5

7 dataset was reduced to a set of six uncorrelated (as determined by PCA) variables to meet the assumptions of the analysis. These variables were CAW, CTM, CPL, CCH, AWM and QTL (Appendix 1). Set tolerance limits were Analyses were conducted in Microsoft Excel using statistixl ( Institutional Abbreviations AM M, Mammal Section, Australian Museum, Sydney; AMNH M, Department of Mammalogy, American Museum of Natural History, New York; QM M and QM JM, Mammal Section, Queensland Museum, Brisbane; SAM M, Department of Mammals, South Australian Museum, Adelaide; NMV C, Mammal Section, Museum Victoria, Melbourne (formerly National Museum of Victoria); WAM M, Mammal Collection, Department of Terrestrial Vertebrates, Western Australian Museum, Perth. RESULTS Pedal morphology Calcaneum Species of Dendrolagus are united by the following calcaneal attributes compared with terrestrial macropodines: a short, very broad calcaneum; medial expansion and dorsoventral compression of the tuber calcanei; a short, broad plantar surface; increased size of the sulcus anterior to the rugose area on the plantar surface (Fig. 5). The talocalcaneal articulation is the main component lower ankle joint of macropods. It comprises an oval-shaped, concave posteromedial facet, a convex posterolateral facet, and usually a small anterior facet. In 6

8 Dendrolagus, this articulation is markedly broadened and equivalent to at least half the maximum length of the calcaneum. In addition, the lateral and medial facets are confluent and disproportionate in size, with the lateral facet larger (Fig. 5B). The lateral facet is medially constricted and laterally expanded producing a conical, rather than cylindrical, convexity when viewed dorsally. The medial facet is flatter, more obliquely aligned, and less pronounced posteriorly than in terrestrial species. In all species of Dendrolagus except D. inustus, the anterior talocalcaneal facet is lost due to the more medial placement of talar head. In D. inustus, has a large anterior facet; dorso-medial orientation, separated a short distance from the postero-medial facet. Expansion (transversely) of the sustentaculum tali in Dendrolagus is correlated with the medial extension of the medial talar facet, which results in enlargement of the entire anteromedial portion of the calcaneum (Fig. 5D). Ventrally, the sustentaculum tali is convex, with a very shallow flexor groove. The anterior end of the calcaneum is dominated by the cuboid articular facets. The calcaneal-cuboid articulation is the major component of the transverse tarsal joint. In macropods, this articulation is divided into two major facets separated by a step, which results from the medial facet extending much more anteriorly than the lateral facet. A third, ventromedian facet is variably developed across the family. In Dendrolagus, the calcanealcuboid articulation is relatively expanded and flattened, and the cuboid step is smoothed and curved in profile, and the ventromedian facet is indistinct, i.e., completely merged with dorsolateral facet (Fig. 5F). The calcaneum of Bohra is very similar in overall appearance to that of Dendrolagus, in particular D. inustus, and the Australian species, D. bennettianus and D. lumholtzi. It is stout and very broad, with enlarged articular facets and a robust, medially expanded tuber calcanei (Fig. 5). The anteromedial aspect, including the medial talar facet, is very broad and centrally 7

9 expanded, as in Dendrolagus. Overall, the calcaneal-talar articulation of Bohra most closely resembles that of D. inustus, with a large medial facet and a strongly convex, conical lateral facet (Fig. 5). The articular surface is continuous in B. paulae and Bohra sp. nov., but there is a partial separation in Bohra illuminata. In this species, the lateral facet is relatively narrow and less tapered mesially. The dorsal anteromedial facet, for articulation with the talar head, is reduced but not lost entirely. The cuboid step of Bohra is smoothed, as in Dendrolagus, and the ventromedian facet is broad and partially integrated with the dorsal facets (Fig. 5E). The medial surface of the lateral facet is medially concave, i.e., the boundary between the medial and lateral facets is arcuate. Talus The upper ankle joint is fundamentally a hinge that joins the pes to the crus (tibia and fibula). The main component of the joint is the talotibial articulation. The talus of Dendrolagus is much broader relative to its length than in terrestrial kangaroos and wallabies (Fig. 1). The increase in overall width is primarily due to the expansion of the talar head, neck and medial malleolus and, to a lesser extent, the expanded width of the articular trochlea. The talar head is ovoid and wider than it is long (although the orientation of the long axis does vary among the species; oblique in D.inustus and D bennettianus, and roughly transverse in other species). In terrestrial macropodines, the articular surface for the navicular is dorsoventrally long and narrow and rectangular in shape (Fig. 1). The trochlear crests are low and more gently sloped laterally in Dendrolagus compared with terrestrial macropodines, where the crests are high with near-vertical sides. The lateral crest is also higher than the medial crest. The ventral articular facets for the calcaneum, in particular the ovoid medial facet, is lengthened in Dendrolagus, and the two facets are confluent, as in the corresponding 8

10 calcaneal features. The talar neck is very short and broad. The astragalotibial pit on the dorsal surface of the talar neck is broad and shallow in contrast to the deep morphology observed in terrestrial macropodines. The talus of Bohra (Fig. 6A-D) is intermediate in general proportion between those of terrestrial macropodines and Dendrolagus, i.e., relatively wider than that of terrestrial macropods and relatively longer than that of Dendrolagus (Fig. 1). As in Dendrolagus, increased width of the talus in Bohra has been primarily achieved by mesial expansion of the medial malleolus and the talar head (Fig. 6B). The head is obliquely orientated as in D. inustus and D. bennettianus (Fig. 6D). Unlike Dendrolagus, however, the head is relatively anteriorly placed due to the longer neck region and reminiscent of the morphology observed in terrestrial macropodines. The trochlear crests are long, and the trochlear fossa is relatively narrow relative to Dendrolagus (Fig. 1). The height of the crests is intermediate between Dendrolagus and terrestrial species, and the lateral crest is more flattened than the medial (Fig. 6D). The flattened lateral crest is similar to Dendrolagus while the higher medial crest is a feature retained from terrestrial species. The ventral facets for articulation with the calcaneum are greatly expanded (Fig. 6C), and show a tendency toward the confluence seen in Dendrolagus. Cuboid By comparison with the very cubic morphology of terrestrial macropodines (see Fig. 4 M. fuliginosus), the cuboid of Dendrolagus is distinctly compressed dorsoventrally and anteroposteriorly (Fig 1D). This is emphasised in the short-footed species group. The step on the posterior surface of the cuboid is smoothly contoured, reflecting the morphology of the calcaneum. The lateral plantar process is much more elongate and obliquely oriented in 9

11 Dendrolagus than in the terrestrial taxa. It forms a broad shelf across the posterolateral quadrant of the plantar surface, and is bounded by the oblique peroneal sulcus anteriorly, and the deep flexor sulcus medially. The medial plantar process is flattened in D. lumholtzi and D. bennettianus, and is barely distinguishable in D. inustus, D. dorianus and D. matschiei. The cuboid of Bohra (Fig. 6E-H) is intermediate in morphology between those of Dendrolagus and terrestrial macropodines. It retains a primarily cubic appearance, but the lateral plantar process is laterally expanded (Fig. 6G), and its anterior border, together with the peroneal sulcus, is oblique in orientation. In addition, the lateral portion of the cuboid is slightly compressed when viewed dorsally (Fig. 6F), and the step between the medial and lateral calcaneal facets is smoothed. Metatarsals The metatarsals are short and robust in all species of Dendrolagus relative to those of terrestrial macropodines (Fig. 1). They are relatively longest in D. inustus (mean metatarsal IV length to proximal width: 3.9) and relatively shortest in D. matschiei (length to width: 2.8). Typically, metatarsal IV in Macropus is six to seven times longer than its proximal width. This ratio varies in other terrestrial macropodine genera from approximately four (Setonix, Petrogale) to eight times (Lagorchestes). The metatarsals of Dendrolagus also differ by being sub-equal in length and more ventrally convex in the sagittal plane, and by having relatively much broader, deeper distal epiphyses and expanded articular facets for the interosseous articulation (Table 2, Fig. 1). The proximolateral process of metatarsal V is also expanded. The head is asymmetrical. Among terrestrial macropodines, the interosseous facets of metatarsals IV and V are elongate and variably constricted mesially. In Setonix and some species of Macropus this mesial constriction is complete, but in other taxa (e.g., Petrogale, 10

12 Thylogale) the two portions are contiguous. In Dendrolagus, facet for metatarsal V is relatively large and continuous. Metatarsals IV V in Bohra are large and relatively broad in cross-section in comparison to Macropus (Fig. 1). The articulation between them resembles that of Dendrolagus in general shape and the confluence of the two facets. The shaft length of metatarsal IV is intermediate between those of Dendrolagus and Macropus, and most similar to those of Setonix and Petrogale. The Bohra metatarsal V is particularly robust compared with that of both Macropus and Petrogale. Metatarsal IV differs from that of terrestrial macropodines in the greater expansion of the cuboid articulation, the plantar tuberosity medially, and the articulation for metatarsal V laterally. Similarly, the proximolateral expansion of metatarsal V and the expansion of the articular facets closely resemble the conditions expressed in Dendrolagus. Statistical Analyses Principal Components Analysis Analysis of the dataset comprising calcaneal, talar and cuboid indices of Bohra illuminata and 21 extant macropodines, including five species of Dendrolagus (Fig. 7), demonstrated strong separation of the tree-kangaroos from the terrestrial species in the first principal component (PC1). This accounts for 57.9% of the variance. Nearly all component loadings for PC1 are positive, but not all of the positive loadings are close to unity, and PC1 is not correlated with size (CML; R 2 =0.28). Even after the removal of the tree-kangaroos and potoroines, no clear relationship between this axis and body size is evident. The components that contribute to the strong positive placement of Dendrolagus and Bohra on this axis, in 11

13 decreasing order of significance, include: AWM, CAW, ATW, ALH, CMW, QPW, APW, AMT, AMF, AHW, ALL, QML, ALM, CTW. The most significant of these measures relate to the relative widths of the tarsal elements, especially those of the articular surfaces, as well as the expansion of the lateral aspect of the trochlear articulation of the talus. The strong negative placement of the large species of Macropus is attributed to, in decreasing order of significance: QTL, CPL, QPH, CSL. These correspond to the relatively large size of the plantar features of the cuboid and calcaneum, and the length of the sustentaculum tali. The second principal component (PC2) accounts for 10.7% of the variation and separates the potoroines and Lagostrophus from the remainder of the species (Fig. 7A). Their positive placement reflects the contribution of, in decreasing order of significance: CCS, CCH, CCL, QPH, CTD, AHL, CAL, AMF. The first four of these dimensions relate to the shape of the calcaneal-cuboid articulation. The remaining four dimensions relate to the depth of the tuber calcanei, the relative shape of the talar head, the talocalcaneal articular length and the relative proportions of the medial talar facet. Dendrolagus, Bohra and the remaining macropodines are negatively placed on PC2, reflecting, in decreasing order of significance: CTM, AML, CPL, CTW, ALM. These relate to features including the relative shape of the tuber calcanei and length of the talus. PC2 is not correlated with size (CML; R 2 =0.37). PC3 accounts for 6.6% of the variation in the dataset but does not show any obvious differences between the main species groups (Fig. 7B). Tree-kangaroos, which demonstrate the greatest range and include both the highest (D. inustus; 2.667) and lowest (D. dorianus; ) score in PC3. Bohra has an intermediate score (1.668). The strongly positive components include QLL, AHL, CSL, QPH, QTW and CCV. The most significant of these is the relative length of the cuboid, the length of the talar head and the length of the sustentaculum tali, while the remaining three correspond to the relative size of the cuboid facets for the calcaneum. The strongly negative components are CCS, QAW and ALW. 12

14 Removal of potoroines and Lagostrophus from the dataset has little effect on the manner in which the macropodine species are differentiated for PC1. This accounted for 65.9% of the variation, and is not strongly correlated with size (CML; R 2 = 0.31), although nearly all of the components had positive loadings, the majority of these being close to unity. Strongly negative components, in decreasing order of significance, were: CPL, QTL, QPH, CCH, CSL. PC2 (7.8% of the variation) corresponded strongly with PC3 from the previous analysis, and resulted in scattering of the tree-kangaroo species. PC3 accounts for 7.0% of the variation and separates taxa into a strongly positive pair of Lagorchestes conspicillatus and M. irma, a strongly negative pair, comprising two of the large species of Macropus, and an intermediate group, made up of Thylogale, Petrogale, Onychogalea, Lagorchestes hirsutus and the remaining species of Macropus. The species of Dendrolagus are widely spread in PC3 while Bohra is placed at the positive end of the axis. The strongly positive components include, in decreasing order of significance: CCH, QPH, and CTD. The most strongly negative components include CSL, AHL, CTH, QTW, and CTM. Analysis of calcaneal dimensions alone for all species, plus Bohra sp. nov. and B. paulae, again showed a very similar distribution of the taxa in the first three principal components (Fig. 7C), accounting for 51.9%, 18.9% and 10.4% of the data, respectively. All three species of Bohra cluster with Dendrolagus, being strongly positive in PC1 and negative in PC2. Among them, B. paulae and Bohra sp. B lie closer to the terrestrial macropodines than any of the species of Dendrolagus. The most strongly positive loadings in PC1 are CAW, CWS, CTH, CAL, CCM, CTW, and CTM. All reflect the relative width of the calcaneal features. The strongest negative loading is for CPL. There is no correlation between PC1 and size (CML; R 2 =0.21). Discriminant Analysis 13

15 Discriminant analysis of the dataset comprising calcaneal, talar and cuboid indices of the extant species plus Bohra illuminata utilised six uncorrelated measures as determined from the Principal Components Analysis (CAW, CTM, CPL, CCH, AWM, QTL; n=17; tolerance=0.005) to plot groups based on habitat preference (Table 1). The first function explains 94.5% of the variation between the groups and demonstrates a clear separation of the groups, ranging from the terrestrial species adapted to the most open terrain, through inhabitants of denser understorey habitats, to rocky habitats, and finally arboreal specialists (Fig. 8). Bohra illuminata clusters with the species of Dendrolagus. The variable with the strongest positive loadings for function 1 is CTM followed by CCH and AWM, while the strongest negative loading is for CAW. DISCUSSION Functional morphology The short, wide tuber calcanei of Dendrolagus and Bohra provides for less torque in the sagittal plane than in terrestrial macropodines, but provides greater mechanical advantage over a wider range of positions. The disproportionate development of the lateral facet relative to the medial facet suggests a less stable joint in which unbalanced loads, experienced during variable inversion and eversion of the pes, are a dominant pattern during locomotion. A similar pattern of modification characterises other arboreal marsupials (Argot 2002). Furthermore, the anterior articulation between the calcaneum and the talar head, which helps to stabilise the lower ankle joint in terrestrial kangaroos, has been entirely lost in treekangaroos, again suggesting a wider range in rotational movements. High crests and a deep trochlea groove of the talus typically correlate with strong flexion 14

16 extension movements and restricted rotation at the upper ankle joint in mammals (Szalay 1994; Argot 2002). The vertical sides of the crests form a strong locking mechanism with the tibial and fibular malleoli, restricting movement to the sagittal plane. This morphology is exemplified in terrestrial macropodines. Reduced height of trochlear crests on the talus of Dendrolagus highlights greater mobility at the joint. The unequal height of the crests, and their smoother, flatter sloping external faces resembles the morphology of other arboreal mammals (Argot 2002), and facilitates a comparatively wider range of rotational movements. The form of the lateral crest facilitates a more mobile fibular articulation. The relatively short and broad trochlear surface of Dendrolagus further serves to enhance mobility at the talotibial articulation; the reduced length of the trochlea highlighting a shift away from restricted foreand-aft hinge movements, and the increased width of the surface suggests a spreading of forces acting at the joint over a wider range of postures. Bohra retains a more intermediate morphology of the talotibial articular surfaces, suggesting that major changes at this joint occurred later in the specialisation to an arboreal lifestyle. Presence of a deep talotibial pit on the dorsal surface of the talar neck denotes a large tibial malleolus, which prevents extreme dorsoventral flexion in terrestrial mammals. The feature is usually reduced or lost in arboreal forms (e.g., Salton and Szalay 2004). In Dendrolagus and Bohra, the talotibial pit is broad, but not as deep or steeply sided as in terrestrial kangaroos. The tree-kangaroo morphology reflects enhanced dorsoventral flexion and a greater emphasis on rotation. The more anteroventrally rounded head of Dendrolagus facilitates a wider range of possible movements and the short, robust talar neck suggests a wider range of pedal positions (e.g., Szalay 1994; Argot 2002; Salton and Szalay 2004). Bohra is less specialised in this regard, which suggests a more generalised function than in Dendrolagus, and that specialisation of this tarsal region occurred subsequent to those relating to enhanced mobility at the talocalcaneal and calcaneal-cuboid joints. 15

17 The stepped morphology of the transverse tarsal joint restricts rotational movements, thereby stabilising the pes of terrestrial macropodines during bipedal saltation (Bishop 1997). Strong development of a ventromedian facet and plantar tuberosity of the cuboid further restricts movement at the joint. Dendrolagus, and to a lesser extent Bohra, have modified the calcaneal-cuboid articulation by smoothing the step, increasing the curvature of the articular surface, and coalescing the ventromedian and dorsal facets to facilitate greater rotation (i.e., supination / pronation) and translation. The more oblique orientation and elongate morphology of the cuboid plantar tuberosity, and the flexor tendons that pass around it, aid in this movement (e.g., Gebo 1987; Argot 2002), helping overcome the phylogenetic constraints of the stepped calcaneal-cuboid articulation. While the morphology of the cuboid of Bohra is less specialised than that of Dendrolagus, the expansion of the lateral plantar process and oblique peroneal sulcus are well-developed. Overall expansion of the articular surface, particularly of the lateral facet, has been shown to facilitate a wider range of inversion movements in other mammals (Salton and Szalay 2004). The smoothed, arcuate step in the calcaneal-cuboid articulation of Bohra reflects an enhanced range of rotational movements compared with terrestrial kangaroos, but less translation than in Dendrolagus, as revealed by its more completely smoothed articulation. The functional modification of the transverse tarsal joint represents perhaps the key adaptation in the evolution of tree-kangaroo pes. The short, robust metatarsals and phalanges of Dendrolagus reflect selection for increased manoeuvrability and balance in an uneven, three-dimensional environment (Flannery 1982). Bohra also possesses relatively short and robust metatarsals, but the phalanges are not as stout proportionally as those of Dendrolagus. The enlarged distal epiphyses and expanded interosseous articular facets of the metatarsals of Dendrolagus and Bohra facilitate a greater range of movement in the metatarsal and phalangeal region than is possible in terrestrial macropodines. Similar observations have been made in other arboreal marsupials (Argot 16

18 2002). Reduced length and increased flexibility of the pes is a key adaptation in the assumption of an arboreal habitat. Morphometry Multivariate analyses of linear landmark data provide an opportunity to test whether morphological differences observed through visual comparison also separate species statistically. They allow morphological variation to be plotted graphically to help discern both large scale and fine scale patterns within and between taxa. The interaction between PC1 and PC2 spreads the terrestrial groups along a broad gradient from large species occupying open areas through to thicket inhabitants (Fig. 7). The first discriminant function accounts for 96.9% of the variation between species, and also clearly distinguishes, irrespective of body size, species adapted to open terrain, inhabitants of denser understorey habitats and rocky habitats, and tree-kangaroos (Fig. 8). Within the PCA, the potoroines and Lagostrophus are distinct from the terrestrial macropodines, suggesting a clear deviation in tarsal morphology between them and the macropodines. Further quantitative analyses incorporating larger samples sizes and a wider spread of terrestrial taxa may be expected to help to resolve the fine-scale shape changes of the tarsus and its relationship to the functional and ecological evolution of terrestrial macropodines. Principal components analysis (PCA) confirms that the three species of Bohra represented by pedal elements have a quantitatively distinctive calcaneal form. The species plot closer to those of Dendrolagus than to any of the terrestrial macropodines, although their exact positions differ slightly. B. paulae plots relatively closer to the terrestrial kangaroos in calcaneal morphometry than the other two species, Bohra sp. nov. is intermediate between B paulae and Dendrolagus, and Bohra illuminata lies within the bounds of the Dendrolagus 17

19 cluster (Fig. 7C). When measurements of the calcaneum, talus and cuboid are analysed together, Bohra illuminata remains clustered with Dendrolagus (Fig. 7A B). Thus, PCA strongly confirms the qualitative comparisons: the tarsus of Bohra illuminata was as well adapted to the arboreal realm as the extant, rainforest-dwelling tree-kangaroos. Bohra sp. nov. and B. paulae differ only by being displaced along one axis of the PC1 plot due to a slightly longer tuber calcanei, but the similarities with Bohra illuminata strongly outweigh the differences. By contrast, substantially more variation in tarsal morphology is observed between treekangaroo species for PC3, which accounts for 6.6 % of the total variation in the entire group (Fig. 7B). Whether this reflects functional or phylogenetic patterns within the tree-kangaroo group remains unresolved. There is strong separation between the short-footed New Guinea species (D. dorianus, D. matschiei) and the long-footed Australian species (D. bennettianus, D. lumholtzi, Bohra illuminata). More subtle morphometric differences are also observed between species within these groups. Understanding the functional or phylogenetic basis for this variation must await a more detailed analysis incorporating a broader range of extant species. Palaeoecology The concept of tree-kangaroos as prior inhabitants of the now treeless Nullarbor Plain initially seems to beggar belief. However, given that most other taxa in the late Cenozoic faunal assemblages of Chinchilla (southeastern Queensland) and Wellington Caves (central eastern New South Wales) are more suggestive of open woodland and grassland than forest, it is perhaps not quite as astonishing as it first seems. Flannery and Szalay (1982) and Dawson (2004) were well aware that Bohra paulae and B. wilkinsonorum may not have inhabited 18

20 forest, although they justifiably left open the possibility of a preference for remnant riparian forest. The flat limestone karst terrain of the Nullarbor Plain is too permeable to support surface streams, and during the middle Pleistocene, as now, it is likely that animals incapable of surviving without free water utilised ephemeral rock holes. Palaeoecological and stable isotopic analyses of the diverse Nullarbor fossil fauna (70 species, including 23 kangaroos) point to a similar climate to today ( mm mean annual rainfall), but a much more diverse vegetation structure than the modern shrub steppe (Prideaux et al. 2007). Our analyses show that species of Bohra had a very similar level of pedal mobility and hind limb strength as Dendrolagus species. Even though they may have weighed up to twice as much as their extant relatives, the species of Bohra were clearly adept at moving within trees. Bohra differed from Dendrolagus in overall build, possessing a larger body relative to the size of the head, and longer hind limbs relative to both forelimb length and body size (Prideaux and Warburton 2008). So while the species of Bohra were taxonomically and functionally tree-kangaroos, their body proportions were not quite as derived relative to their terrestrial ancestor as modern Dendrolagus. Within the scleromorphic woodland/shrubland mosaic inferred for the middle Pleistocene environment of the Nullarbor Plain (Prideaux et al. 2007), the two species of Bohra may have utilised a range of small tree species now largely restricted in this region to remnant stands on the periphery of the plain. These include Myoporum platycarpum, Pittosporum phylliraeoides, Santalum acuminatum, S. spicatum, Alectryon oleifolius and Acacia papyrocarpa, plants with palatable leaves and/or fleshy fruit now highly sought after by stock and rabbits (Mitchell and Wilcox 1998). CONCLUSIONS 19

21 Pedal attributes shared by Bohra and Dendrolagus to the exclusion of all other macropodines reflect enhanced mobility of the pes for greater flexibility and balance in a three-dimensional environment. These include marked expansion of the tuber calcanei, enlargement and confluence of the posterior facets for the calcaneal-talar articulation, expansion and smoothing of the facets of the calcaneal-cuboid articulation, broadening of the talus, deepening of the talar head, flattening of the cuboid, and the sub-equal length and stoutness of metatarsals IV V. Together, these facilitate an expanded range of inversion/eversion and abduction/adduction movements. Based on this evidence, the species of Bohra were adapted to moving in an arboreal environment. Still, the pes of Bohra was not quite as specialised as that of Dendrolagus, which suggests that Bohra is the more plesiomorphic taxon. Species of Bohra bear closer resemblance to the long-footed species of Dendrolagus than the short-footed species, and it is probable that the similarities represent symplesiomorphies, although confirmation awaits completion of our phylogenetic analysis of the group. ACKNOWLEDGEMENTS We thank the following individuals and institutions for the loan of or access to comparative specimens: Norah Cooper, Western Australian Museum; Robert Jones, Australian Museum; Scott Hocknull, Steve Van Dyck, Heather Janetski and Kristen Spring, Queensland Museum; Robert Palmer, Australian National Wildlife Collection; David Stemmer, South Australian Museum; Robert Voss, American Museum of Natural History. Jane Prince and Phil Withers advised on statistical analyses. This study was supported by the Rio Tinto WA Future Fund and an anonymous donation to the Western Australian Museum Foundation. 20

22 REFERENCES Alexander, R. M. and Vernon, A The mechanics of hopping by kangaroos. J. Zool. 177: Argot, C Functional adaptive analysis of the hindlimb anatomy of extant marsupials and the paleobiology of the Paleocene marsupials Mayulestes ferox and Pucadelphys andinus. J. Morphol. 253: Bishop, N. J Functional anatomy of the macropodine pes. Proc. Linn. Soc. NSW 117: Burk, A., Westerman, M. and Springer, M. S The phylogenetic position of the musky rat-kangaroo and the evolution of bipedal hopping in kangaroos (Macropodineae: Diprotodontia). Syst. Biol. 47: Carrano, M. T Morphological indicators of foot posture in mammals: a statistical and biomechanical analysis. Zool. J. Linn. Soc. 121: Dawson, L A new Pliocene tree kangaroo species (Marsupialia, Macropodinae) from the Chinchilla Local Fauna, southeastern Queensland. Alcheringa 28: Dawson, L. and Augee, M. L The late Quaternary sediments and fossil vertebrate fauna from Cathedral Cave, Wellington Caves, New South Wales. Proc. Linn. Soc. NSW 117: Dawson, L., Muirhead, J. & Wroe, S The Big Sink Local Fauna: a lower Pliocene mammalian fauna from the Wellington Caves complex, Wellington, New South Wales. Rec. West. Aust. Mus., Suppl. 57: DeGusta, D. and Vrba, E. S A method for inferring paleohabitats from the functional morphology of bovid astragali. J. Archaeol. Sci. 30: Flannery, T. F Hindlimb structure and evolution in the kangaroos (Marsupialia: 21

23 Macropodoidea). Pp in The Fossil Vertebrate Record of Australasia ed. by Rich, P. V and Thompson, E. M. Monash University Press: Melbourne. Flannery, T. F., Martin, R., and Szalay, A Tree-kangaroos: A Curious Natural History. Reed Books: Melbourne. Flannery, T. F. and Szalay, F. S Bohra paulae, a new giant fossil tree kangaroo (Marsupialia: Macropodineae) from New South Wales, Australia. Aust. Mammal. 5: Gebo, D. L Functional anatomy of the tarsier foot. Am. J. Phys. Anthrop. 73: Groves, C. P The systematics of tree kangaroos (Dendrolagus; Marsupialia, Macropodineae). Aust. Mammal. 5: Hall, R Cenozoic reconstructions of SE Asia and the SW Pacific: changing patterns of land and sea. Pp in Faunal and Floral Migration and Evolution in SE Asia Australasia ed. by Metcalfe, I., Smith, J. M. B., Morwood, M. and Davidson, I. D. A. A. Balkema: Lisse. Hocknull, S. A Additional specimens of Bohra (Marsupialia: Macropodineae) from the Pliocene of Queensland. Mem. Qd Mus. 51: 26. Mitchell, A. A. and Wilcox, D. G Arid Shrubland Plants of Western Australia. University of Western Australia Press: Perth. Prideaux, G. J., Long, J. A., Ayliffe, L. K., Hellstrom, J. C., Pillans, B., Boles, W. E., Hutchinson, M. N., Roberts, R. G., Cupper, M. L., Arnold, L. J., Devine, P. D. and Warburton, N. M An arid-adapted middle Pleistocene vertebrate fauna from southcentral Australia. Nature 445: Prideaux, G. J. and Warburton, N. M A new Pleistocene tree-kangaroo (Diprotodontia: Macropodineae) from the Nullarbor Plain of south-central Australia. J. Vert. Paleo. 28: in press. 22

24 Salton, J. A. and Szalay, F. S The tarsal complex of Afro-Malagasy Tenrecoidea: a search for phylogenetically meaningful characters. J. Mammal. Evol. 11: Sargis, E. J Functional morphology of the hindlimb of tupaiids (Mammalia, Scandentia) and its phylogenetic implications. J. Morphol. 254: Strahan, R. (ed.) The Mammals of Australia. Reed Books, Sydney. Szalay, F. S Evolutionary History of the Marsupials and an Analysis of Osteological Characters. Cambridge University Press: New York. 23

25 Table 1. Species used to investigate pedal morphology and their preferred terrain (from Strahan 1995). n = number of specimens used for Principal Components and Discriminant Analyses. See Materials and Methods for institutional abbreviations. Species Preferred habitat Specimens examined n Aepyprymnus rufescens Mixed WAM M Bettongia lesueur Mixed WAM M3644, M4392, M15488, M Bettongia penicillata Woodlands WAM M13405, M14961, M19064, M19521, M Dendrolagus bennettianus Arboreal NMV C7125, C7123; SAM M Dendrolagus dorianus Arboreal AM M Dendrolagus inustus Arboreal 1 Dendrolagus lumholtzi Arboreal QM M9333, JM13924, JM3860, JM6667; AMNH M65257, M Dendrolagus matschiei Arboreal WAM M21013, M Lagorchestes conspicillatus Grasslands WAM M7032, M19854, M19855, M Lagorchestes hirsutus Grasslands WAM M7965, M Lagostrophus fasciatus Mixed WAM M3625, M3626; M4393, M5658, M6303, M8006, M Macropus eugenii Thickets WAM M6573, M6737, M7654, M8357, M7093, M4162, M Macropus fuliginosus Woodlands WAM M6750, M3328, M13015, M13451, M Macropus irma Woodlands WAM M19864, M19581, M19869, M Macropus robustus Woodlands WAM M14955, M15848, M19223, M19554, M Macropus rufus Grasslands WAM M4953, M6749, M Onychogalea unguifera Grasslands WAM M6942, M11622, M13969, M17142, M Petrogale brachyotis Rocky outcrops WAM M19066, M19540, M Potorous tridactylus Thickets WAM M Setonix brachyurus Thickets WAM M6792, M6795, M13404, M19866, M13936, M Thylogale billardierii Thickets Prideaux private collection 2 24

26 Table 2. Observed range of calcaneal and metatarsal ratios (mean and sample size in parentheses). Ratio Dendrolagus Terrestrial species Bohra CMW: CML (0.69; 13) (0.50; 70) (0.66; 4) CAW: CML (0.54; 13) (0.40; 70) (0.53; 4) CTW: CML (0.41; 13) (0.30; 70) (0.43; 4) CTM: CML (0.32; 13) (0.18; 70) (0.33; 4) CPL: CML (0.71; 13) (0.63; 70) (0.74; 4) CCM: CML (0.21; 13) (0.18; 70) (0.25; 4) CCS: CML (0.06; 13) (0.11; 70) (0.07; 4) Length metatarsal IV: metatarsal V (1.07; 9) (1.12; 15) 1.11 (1) Metatarsal IV distal epiphysis width: length (0.25; 9) (0.15; 16) 0.22 (2) 25

27 FIGURE CAPTIONS Fig. 1. Right pes of four macropodines in dorsal view. A: Macropus fuliginosus; B: Thylogale billardierii; C: Bohra sp. nov. (WAM , reversed, incomplete); D: Dendrolagus bennettianus. Scaled to same tarsus width. 26

28 Fig. 2. Right calcaneum of Macropus fuliginosus, illustrating features that reflect functional attributes and dimensions measured. A: dorsal view; B: plantar view; C: anterior view. Fig. 3. Right talus of Macropus fuliginosus, illustrating features that reflect functional attributes and dimensions measured. A: dorsal view; B: ventral view; C: medial view. 27

29 Fig. 4. Right cuboid of Macropus fuliginosus, illustrating features that reflect functional attributes and dimensions measured. A: dorsal view; B: plantar view. Fig. 5. Left calcanea of Bohra sp. nov. and Dendrolagus bennettianus. A B: dorsal view; C D: plantar view; E F: anterior view. A D scaled to same length; E F scaled to same width. 28

30 Fig. 6. Right tarsal bones of Bohra illuminata. A-D: talus, medial view (A), dorsal view (B), ventral view (C), anterior view (D); E-H: cuboid, posterior view (E), dorsal view (F), ventral view (G), anterior view (H). 29

31 Fig. 7A Fig. 7B Fig. 7C Fig. 7. Principal Components Analysis of tarsal morphological variables. A B: calcaneum, 30

32 talus and cuboid dimensions, PC2 versus PC1 (A), PC3 versus PC1 (B); C: calcaneum dimensions, PC2 versus PC1. 31

33 Fig. 8 Fig. 8. Discriminant Analysis of tarsal variables within macropodines (excludes potoroines and Lagostrophus). 32

34 APPENDIX 1 Abbreviation Definition AHH AHL AHW ALH ALL ALM ALW AMF AMH AMM AML APW ATW AWM CAL CAW CCH CCM CCL CCV CCS CML CMW CPL CSL CTD CTH CTM CTW QAW QLL QML QPH QPW QTL QTW Anteroposterior height of talar head measured at 90 to mediolateral length. Dorsoventral length of talar head measured on medial aspect from anterior margin to posterior margin of navicular facet. Maximum mediolateral width of talar head. Maximum height of lateral trochlear crest of talus from basal plane to crest apex. Maximum length of the lateral trochlear crest of talus in sagittal plane. Maximum medial anteroposterior length measured from posterior plantar tubercle to anterior margin of talar head. Width of lateral facet measured parallel to coronal plane of talus. Width of medial facet measured parallel to coronal plane of talus. Maximum height of medial trochlear crest of talus from basal plane to crest apex. Minimum median height of talus from basal plane to lowest point on trochlea. Maximum length of medial trochlear crest of talus in sagittal plane. Distance between the posterior medial and lateral margins of the trochlear facet taken parallel to the coronal talar plane. Maximum distance from anterior medial and lateral margins of trochlear facets measured parallel to coronal plane of talus. Maximum talar width in coronal plane from lateral tubercle to medial margin. Distance between anteriormost margin of dorsal calcaneal surface measured from level with anterior facet for talus to anterior margin of posterolateral facet. Maximum linear distance on calcaneum between medial and lateral margins of posterior talar articular facets measured along long axis of articulation. Height of cuboid articulation on calcaneum measured from dorsal to plantar margin. Maximum width of dorsomedial facet of cuboid articulation on calcaneum. Maximum width of dorsolateral facet of cuboid articulation on calcaneum. Maximum width of ventromedian facet of cuboid articulation on calcaneum. Projected height of cuboid step on calcaneum measured in sagittal plane. Maximum linear length of calcaneum in sagittal plane measured from posteriormost tip of tuber calcanei epiphysis to anteriormost point of cuboid articular facet. Maximum linear calcaneal width in transverse plane. Maximum length of rugose plantar surface of calcaneum measured parallel to sagittal plane. Length of sustentaculum tali of calcaneum. Minimum dorsoventral height of tuber calcanei measured at 90 to sagittal plane (perpendicular to CTM). Maximum dorsoventral height of tuber calcanei measured at 90 to sagittal plane (Perpendicular to CTW). Minimum mediolateral width of tuber calcanei. Maximum mediolateral width of tuber calcanei measured as maximum width of epiphysis. Maximum width of dorsal anterior aspect of cuboid. Distance on cuboid from most dorsal lateral point of metatarsal facet to most dorsal lateral point of calcaneal facet. Distance on cuboid from most dorsal medial point of metatarsal facet to most dorsal medial point of calcaneal facet. Maximum height of posterior aspect of cuboid corresponding to dorsoventral plane of calcaneal facet, perpendicular to QPW. Maximum width of posterior aspect of cuboid corresponding to transverse plane of calcaneal facet. Maximum length of posterior plantar process of cuboid measured parallel to sagittal plane. Maximum width of posterior plantar process of cuboid measured parallel to transverse plane. 33