New craniodental remains of Wakaleo alcootaensis (Diprotodontia: Thylacoleonidae) a carnivorous marsupial from the late Miocene Alcoota Local Fauna of the Northern Territory, Australia

New jaws and teeth referable to the rare thylacoleonid marsupial Wakaleo alcootaensis are figured and described. The species is the geologically youngest known member of the genus and is only known from the late Miocene Alcoota Local Fauna of the Northern Territory, Australia. A revised diagnosis of the species is presented which is found to be morphologically distinct from its congeners. W. alcootaensis can be distinguished from other species of Wakaleo by its greater size, deeply recessed masseteric fossa, more steeply angled I1, loss of P2, greater P3 to M1 ratio and loss of M3. Several characters of W. alcootaensis, including the increase in size, steeply angled I1, increase of the relative size of P3, and reduction of the molar row are present in at least some species of Thylacoleo. Phylogenetic analysis suggests that these character states are convergences and that there was parallel evolution in these two thylacoleonid lineages.


INTRODUCTION
Thylacoleonids, or 'marsupial lions' are a group of small to large-bodied diprotodontian marsupials that range from the size of small house cat in Priscileo roskellyae (Wroe et al., 2003) up to the size of a lion in Thylacoleo carnifex (Wroe et al., 1999;Wroe et al., 2003). They are characterised by the development of the third premolar pair into large shearing blades. Although there has been much debate about the diet of these creatures in the past, it is now largely accepted that Sir Richard Owen was correct in 1859 when he described the eponymous Thylacoleo carnifex as "one of the fellest and most destructive of predatory beasts" (Owen, 1859, pg. 319).
Wakaleo is a genus of thylacoleonids that ranges from the late Oligocene through to the late Miocene (Gillespie, 2007). It differs from the Plio-Pleistocene genus, Thylacoleo, by having highly reduced to absent anterior premolars, a P 3 that broadens posteriorly, a largest of these. Unfortunately the species has remained extremely rare and poorly known. Indeed, anatomical knowledge of the species is so poor that it could only be diagnosed by its larger size relative to other species of Wakaleo, leaving open the question of its validity as a distinct taxon even though this question has not been raised in the literature. The species was established for a single cranial fragment that was unfortunately badly damaged while trenching around a plastered block of dense bone bed material (Archer & Rich, 1982). Very few other specimens of this species have been found. One of them is a dentary fragment bearing two molars (UCMP 65621) that was recovered during an initial investigation of Alcoota in 1962 (Fig. 2). This specimen was initially described as a possible giant perameloid (Woodburne, 1967). G Prideaux (pers. comm., 2012) first suggested that this specimen was actually a misidentified Wakaleo specimen, a reidentification that is drawings of (A-C) respectively. Note that specimen in (A-C) has been whitened with ammonium chloride. Abbreviations: ew, remnant enamel well; m1, first lower molar; m2, second lower molar; ta, talonid; tc, transverse crest; tr, trigonid. Scale bar = 20 mm. supported in the present work. The only other positively attributable specimens that came to light prior to 2013 are a few isolated teeth and a few postcranial elements, none of which have been described in the scientific literature.
During the 2013 field season a new pit was opened at Alcoota, on the same stratigraphic level as the other pits that quarry the Alcoota Local Fauna. This new pit, named 'Shattered Dreams' , proved to be exceptionally densely packed with fragmented bones, interspersed with occasional complete, or near complete specimens. Not only was the volume of fossil bone extraordinarily high but so was the diversity, with virtually all known taxa from the Alcoota Local Fauna recovered from an area of less than two square meters. Included among these was a dentary belonging to W. alcootaensis . This is the first substantial cranial specimen of this species found since the holotype was recovered 39 years previously. W. alcootaensis is now known from all of the main quarries of the Alcoota Local Fauna.
In this paper all craniodental material of W. alcootaensis identified subsequent to the description of the holotype is described and illustrated. The diagnosis of W. alcootaensis is revised and new morphological character traits are identified that place the diagnosis on a firmer footing.

GEOLOGICAL SETTING
The known fossils of W. alcootaensis all come from a dense bone bed in the lower part of the Waite Formation, cropping out on Alcoota Station, 110 km NE of Alice Springs in south Abbreviations: bcp, base of the coronoid process; i1a, alveolus for first lower incisor; m1a, alveolus for first lower molar; m2a, alveolus for second lower molar; mfo, masseteric fossa; p3, third lower premolar; pmf, posterior mental foramen; s, symphyseal surface. Scale bar = 50 mm.
central Northern Territory (Woodburne, 1967).The Waite Formation is a late Cenozoic sequence of fluviatile beds filling the Waite Basin, a small intermontane basin, surrounded by crystalline rocks of the Arunta Block (Woodburne, 1967). The Waite Formation consists of a basal series of overbank silts that were previously interpreted as lacustrine sediments (Woodburne, 1967) with interspersed and discontinuous carbonate-rich beds. The lower overbank beds are overlain by a coarser sequence of channel deposits, consisting of calcareous sandstones grading up into coarse red sandstones that contain a localised, (B) Interpretive drawing. Abbreviations: i1a, alveolus for first lower incisor; m1ara, alveolus for anterior root of first lower molar; m1pra, alveolus for posterior root of first lower molar; m2ara, alveolus for anterior root of second lower molar; m2pra, alveolus for posterior root of second lower molar; p3, third premolar. Scale bar = 20 mm. silty, incised channel fill that is notable for containing the Ongeva Local Fauna (Megirian, Murray & Wells, 1997). The entire sequence is capped by a layer of silcrete. The bone bed that has produced the Alcoota Local Fauna and the W. alcootaensis fossils occurs in the lower overbank deposits, within a greyish-yellow silt unit that is interpreted as a crevasse-splay. The bone bed covers an area of approximately 25,000 m 2 , although its density and thickness varies considerably within that area (Megirian, 2000). The bulk of the known fossil material has been obtained from four pits: Paine Quarry, South Pit, Main Pit and Shattered Dreams (Fig. 1). The bone bed usually lies 90 cm below the present soil surface, underneath a reddish, weathered horizon (Murray & Megirian, 1992). It contains the unsorted but disarticulated and jumbled remains of many hundreds, if not thousands, of animals that appear to have perished in a mass death event, probably caused by severe drought (Murray & Vickers-Rich, 2004). Most of the in situ bones appear to be complete and show no signs of weathering prior to burial. Nevertheless the bones have undergone extensive fracturing due to the movements of the unconsolidated, clay-rich sediment that hosts them (Murray & Megirian, 1992).
Despite the apparently complete condition of most of the in situ bones, the known remains of W. alcootaensis are highly fragmented. In the case of the holotype the damage can be explained by the unfortunate circumstances of its discovery (Archer & Rich, 1982). In the case of the two dentary specimens it appears that both had weathered out of the primary bone bed and were broken up during their passage through the mobile cracking clays of the soil horizon that overlies the site.
It is thought that the fauna is late Miocene in age based on stage of evolution correlation using diprotodontid marsupials, (Stirton, Woodburne & Plane, 1967;Murray & Megirian, 1992), and its age lies between 5 and 12 ma (Megirian et al., 2010).

Terminology
Serial designation of the cheek dentition follows Flower (1867) and Luckett (1993). Standard nomenclature for mammalian tooth cusp anatomy is followed. Standard abbreviations for teeth are used: I, incisor; P, premolar; M, molar, with superscripts or subscripts representing upper or lower dentitions, respectively. Anterior and posterior are used as anatomical directions in the description of the dentition (instead of mesial and distal, respectively).

Measurements
Linear measurements were made with digital vernier callipers. Angular measurements were made with a protractor on a two-dimensional image taken normal to the plane of the angle being measured. The angle of the posterodorsal wall of the alveolus for I 1 was measured by affixing a wooden splint flush against this wall with a small amount of petroleum jelly and measuring the angle of the protruding section.

Cladistic analysis
The broader intrafamilial relationships of Thylacoleonidae, particularly its basal branches are not examined here as the question has been comprehensively examined by Gillespie (2007) and will form the basis of a future publication. The present analysis is designed soley to test whether the new data provided here are enough to affect the position of W. alcootaensis, particularly in light of several derived conditions that are shared with the genus Thylacoleo. Only a single basal thylacoleonid, Priscileo roskellyae, is included to help polarise character states that vary between Wakaleo and Thylacoleo. Character state scores for this taxon were restricted to those that could be determined from available published descriptions and illustrations (Gillespie, 1997). The three named and currently accepted species of both Wakaleo (W. oldfieldi, W. vanderleueri and W. alcootaensis) and Thylacoleo (T. hilli, T. crassidentatus, and T. carnifex) form the rest of the ingroup (data sources in Table 1).
Multistate characters that form obvious transformation series, such as the progressive enlargement of P 3 , were treated as ordered.
The resulting matrix was subjected to a maximum parsimony analysis in PAUP 4.0b (Swofford, 2002) using the following settings: heuristic search; random addition sequence with 500 replicates; and TBR branch-swapping algorithm. The strength of the internal nodes was tested with a decay analysis using the same settings. DIPROTODONTIA Owen, 1866VOMBATIFORMES Woodburne, 1984THYLACOLEONIDAE Gill, 1872 Wakaleo alcootaensis Archer & Rich, 1982 Holotype. NTM P1, a fragment of a left maxilla, with P 3 and fragments of M 1 and M 2 . Found adjacent to Paine Quarry (Archer & Rich, 1982).

Emended diagnosis
A species of Wakaleo distinguished from all others by: larger size (dental dimensions between 16 and 35% greater than the next largest species, W. vanderleueri, Table 2); anterior end of the masseteric fossa deeply recessed; long axis of I 1 inclined at an angle

Notes.
L, anteroposterior length of crown; W, maximum buccolingual width of the crown; DH, dentary height measured at the level of the posterior margin of M 2 ; P 3 -MF, distance from the posterior margin of the P 3 to the anterior rim of the masseteric fossa; P 3 -M2, length of the tooth row from the anterior margin of P 3 to the posterior margin of M 2 ; M 1 -M 2 , combined length of M 1 and M 2 .
greater than 50 • to the horizontal ramus of the dentary; loss of P 2 ; P 3 :M 1 ratio of approximately 1.5; loss of M 3 .

Description
Dentary. NTM P4325 (Figs. 3-5) is similar in size and shape to the larger dentaries of W. vanderleueri (e.g., NTM P87108-6). The horizontal ramus deepens anteriorly to reach a maximum depth under the midlength of the P 3 . The ventral border forms a straight line for its entire length. It is moderately thick buccolingually, with a midlength width of 14.1 mm and a mild dorsoventral convexity on the buccal surface. Species of Wakaleo bear a large anterior mental foramen on the buccal surface of the dentary adjacent to the incisor, and one or two smaller accessory mental foramina posterior to it. The external opening of the anterior mental foramen in NTM P4325 is missing due to damage but a single small posterior mental foramen is present, ventral to the posterior root of P 3 . The lingual surface of the dentary bears a shallow, narrowly triangular, digastric fossa impressed upon the posterior half of the horizontal ramus. Only the posterior end of the symphyseal surface is present, it extends posteriorly to the level of the middle of P 3 . In occlusal view the longitudinal axis of the horizontal ramus is inclined at an angle of 20 • to the symphyseal plane.
Although the anterior end of the dentary is missing, the posterodorsal wall of the alveolus for I 1 is preserved. It indicates that the incisor projected at an angle of 54 • from the longitudinal axis of the dentary. A very short diastema separates this alveolus from the alveolus for P 3 . There is no alveolus for any rudimentary teeth between I 1 and P 3 . The P 3 dominates the dentary, occupying 45% of the total length of the cheek tooth row, or 35% of the distance from the anterior margin of the masseteric fossa to the anterior end of the P 3 . Unfortunately the crown is largely broken away, preventing description of this tooth beyond its size. In occlusal view the anterior end of the P 3 can be seen to be angled lingually so that the distance between the symphyseal plane and the anterior end of P 3 is less than the distance from the symphyseal plane to the posterior end of P 3 (8.2 mm vs. 12.8 mm). Four alveolar sockets follow the P 3 in a linear row without any diastemata. Since the lower molars of Wakaleo are known to be double rooted, it is clear that there were only two molars present behind P 3 . Thus, M 3 was absent in this species. Although the buccal alveolar margin of NTM P4325 is partially eroded, it is clear that the lingual margin was higher and the alveoli were canted to face slightly buccally. This is reflected in the molar crowns of UCMP 65621 which are angled so that the occlusal surfaces face buccodorsally. In lingual view the alveolar margin slopes ventrally from posterior edge of P 3 to the second root socket of M 1 , after which it levels out and becomes roughly horizontal. The only known lower molars of W. alcootaensis are the two present in UCMP 65621 (Fig. 2). These are heavily worn and present few details. M 1 resembles the M 1 of other Wakaleo species in having a subrectangular occlusal outline, with a weak constriction separating the trigonid from the talonid. The lingual half of the trigonid is missing, including the large anterior cusp present in other Wakaleo species. However the buccal side of the trigonid bears a transverse crest that rises lingually to meet this cusp as in the M 1 of other species of Wakaleo. The anteroposterior length of the talonid is approximately equal to that of the trigonid. It has a squared-off posterior margin in occlusal view like other Wakaleo species. Although heavily worn, there is a small well of enamel remaining in the centre of the talonid basin (Woodburne, 1967). It is smooth but may be too small and worn to accurately determine if crenulations were present or absent. The trigonid and talonid are supported by a single root each. The exposed trigonid root is anteroposteriorly compressed and buccolingually expanded in cross section. The partially obscured talonid root has an anteroposteriorly thicker cross-section than the trigonid root.
M 2 is smaller with a more rounded occlusal outline. The trigonid is less strongly raised than in M 1 . It bears a low transverse ridge. Unlike M 1 the talonid of M 2 has a rounded posterior margin. The talonid is narrower than the trigonid although its length is approximately the same as that of the trigonid. As in M 1 , the talonid basin bears a small remnant well of enamel with a smooth surface.
Posterior to the molar tooth row the dentary rises quickly into the ascending ramus, with almost no postalveolar shelf. The buccal surface of base of the ascending process is deeply excavated by a sharply defined masseteric fossa. The anterior end of the masseteric fossa is recessed for several millimetres under the anterior rim forming a blind pocket. This recess is deeper than in the holotype of W. oldfieldi. The anterior margin of the ascending process is supported by a spar-like rib that is transversely broader than it is anteroposteriorly deep. This rib forms the anterodorsal margin of the masseteric fossa.
Upper dentition. Two isolated upper canines are known (Figs. 6 and 7). Each has a single root and a unicuspid crown. The crown is angled lingually relative to the root so that the apex lies level with the lingual side of the crown in occlusal view (NTM P4463, Fig. 7) or overhangs it (NTM P4462, Fig. 6). The long axis of the crown in buccal view is angled posteriorly relative to the root, it is lingually inclined in anterior and posterior views. The root is complete in NTM P4463 (Fig. 7). It is roughly banana-shaped and is approximately four times longer than the crown. It tapers to point at its base and expands to a maximum thickness of 10 mm, at 17 mm from the base. It is gently constricted below the base of the Notes. L, maximum anteroposterior length of the crown; L(roots), minimum anteroposterior length, measured at the constriction below the crown; W, maximum buccolingual width of the crown. Notes. L, maximum anteroposterior length of the crown; W, buccolingual width of the crown at its base; H, height of the crown from base to apex; RootL, length of the root from its tip to the base of the crown.
crown, forming a neck that is slightly narrower than the base of the crown in buccal and lingual view. The crown is spade-shaped in buccal view with a bluntly-rounded apex. The height of the crown is approximately equal its anteroposterior basal length. The anterior and posterior margins bear carinae that extend from the base to the apex, meeting at its tip and dividing the crown into distinct buccal and lingual faces. The buccal face is more distinctly convex in transverse section than the flattened lingual face. The measurements of these canine crowns (Table 4) match those reported for W. vanderleueri (Gillespie, 2007), and are smaller than the canine alveolar dimensions of CPC 26604. Given that all other dental specimens of W. alcootaensis show that it had dimensions in excess of those of W. vanderleueri, it would appear that the canines of W. alcootaensis were reduced relative to its other teeth in comparison to the former species.
The second upper molar of the holotype is only represented by broken roots in the alveolus but an isolated left M 2 is now known (Fig. 8). It is slightly larger than the M 2 of the holotype. Note that the anteroposterior length of the buccal side appears to be significantly greater than the measurement reported in Archer & Rich (1982), but this is because the crown is wider than the roots. The crown of M 2 is missing in the holotype and the length measurement was obtained from the distance between the anterior and posterior roots . When the same measurement is taken on NTM P4328, the difference in length between the two specimens is less than 12% (Table 3). The crown is distinctly trigonal and tritubercular. The occlusal outline of the tooth is nearly equilateral with its buccolingual width similar to its anteroposterior length. This differs from the M 2 of W. vanderleueri in which the width is distinctly greater than the length. Note that because of the equilateral nature of NTM P4328, no side is significantly longer than any other and this difference cannot be explained by a misinterpretation in the orientation of the tooth. A very weak flexus causes a slight emargination on the anterolingual side of the crown in occlusal view, possibly where the posterior margin of M 1 impinged upon M 2 as it does in W. vanderleueri (Murray, Wells & Plane, 1987, Fig. 8). A small depression, containing some grains of adherent matrix lies close to the anterobuccal margin, just anterior and slightly buccal to the metacone, and is interpreted as a vestigial stylar basin. A low cusp is developed at each corner of the crown. These three cusps are the paracone, metacone and protocone. There is no trace of a metaconule. Of these three cusps the paracone is the tallest, represented by a low peaked ridge. The peak is inset from the buccal margin, and the crown is expanded laterally between the roots and the peak of the paracone. However abrasion of the enamel along the buccal margin means that it is not possible to see if the lateral bulge above the paracone is as well developed as it is in W. vanderleueri (Gillespie, 2007). The height of the cusp is not as great relative to the other cusps as it is in W. vanderleueri where the paracone forms a tall peak, even in worn specimens (e.g., NTM P87103-9). Indeed the entire buccal margin is of a similar depth to the lingual margin, unlike the condition in W. oldfieldi and W. vanderleueri where the buccal margin is distinctly deeper than the lingual margin (Gillespie, 2007). Both the protocone and the metacone of NTM P4328 have been worn virtually flat. Low rounded crests connect each cusp and define a triangular, smooth trigon basin that dominates the occlusal surface of the tooth. No crenulations are present in this basin. The deepest point of the basin is slightly off-centre and located closer to the posterolingual rim than the other two sides. A poorly defined shallow trough that extends along the inside of the anterolingual rim of the trigon basin is interpreted as a feature caused by wear. The crown is supported by three subequal roots developed at each of the corners of the trigon. The roots are directed posteriorly and lingually relative to the plane of the crown as in Wakaleo vanderleuri (NTM P87103-9). The root orientation and presence of a vestigial stylar shelf indicate that this equilateral and almost triradially symmetric tooth has been correctly oriented.

Referral of the new material to Wakaleo alcootaensis
The dentaries described here can be referred to Thylacoleonidae on the basis of the enlarged P 3 (NTM P4325), strong reduction in the size of the posterior molars (UCMP 65621 and NTM P4325) and posterior narrowing of the lower molars (UCMP 65621). Within Thylacoleonidae the specimens can be referred to Wakaleo by the loss of the anterior premolars (NTM P4325). Both of these specimens exceed the size of W. vanderleueri and W. oldfieldi and have apparently lost M 3 , excluding them from either species. They can be referred to W. alcootaensis on the basis of matching large size and co-occurrence with the holotype.
The simple, flattened tritubercular and triangular upper molar (NTM P4328), lacking a metaconule and with only a vestigial trace of a stylar shelf, strongly resembles the more posterior upper molars of other species of Wakaleo as opposed to the more rectangular molars of Priscileo and Thylacoleo. As in the dentaries, the matching size of this specimen (Table 3) and its co-occurrence with the holotype of W. alcootaensis, indicate that it can be referred to this species.
As yet, no descriptions of the canines of other Wakaleo species have been published. Nonetheless canines are known for an unnamed, primitive species of Wakaleo and W. vanderleueri both of which have been described in an unpublished PhD thesis (Gillespie, 2007). Those specimens agree with the canines described here in all salient features, including the rounded apex, gentle recurvature and carinate anterior and posterior edges dividing a more convex buccal face from a flatter lingual face. No other mammal known from the Alcoota Local Fauna has a tooth with this combination of features. Thus these isolated specimens can be referred to Wakaleo. They can be referred to the species W. alcootaensis on the basis of co-occurrence.
It remains a far simpler explanation of the data that all of the large-sized Wakaleo fossils at Alcoota belong to a single species rather than to posit multiple large-bodied thylacoleonid taxa for which there is no evidence. In other assemblages where there are two co-occurring thylacoleonid species there is always a large size difference between them. For example the small house cat sized Priscileo roskellyae co-occurs with the leopard sized Wakaleo oldfieldi in the middle Miocene Cleft of Ages Local Fauna of Riversleigh (Archer et al., 2006;Gillespie, 2007) and Thylacoleo hilli has dental dimensions half those of T. crassidentatus with which it co-occurs in the Pliocene Bow Local Fauna (Archer & Dawson, 1982). Furthermore, most of the specimens discussed here show characteristics that are diagnostic of the genus Wakaleo. Since no more than one Wakaleo species is ever present in any one local fauna (Gillespie, 2007), this observation adds further support to the hypothesis that the entire sample belongs to a single species.

Diagnostic characters of W. alcootaensis
Wakaleo alcootaensis was originally diagnosed as distinct from W. oldfieldi and W. vanderleueri on the basis of size, with the P 3 reaching approximately twice the length of W. oldfieldi (Archer & Rich, 1982) and maxillary dimensions that are about 30% larger than those of W. vanderleueri (Murray & Megirian, 1990). Murray & Megirian (1990) suggested that apart from its larger size, W. alcootaensis lacks any significant morphological differences from the smaller, older species of Wakaleo. With the addition of further specimens, including lower jaws, the diagnosis of W. alcootaensis can be expanded. Several characters can now be seen to differentiate the admittedly meagre W. alcootaensis material from the other two named species of the genus (Fig. 9). These are listed briefly above and, given that some interpretation is required, discussed in more detail here: Larger size. As can be seen from the measurements in Table 2 and the discussion on variation below, the size of all known specimens of W. alcootaensis exceeds the known range of W. vanderleueri and W. oldfieldi in almost all dimensions. The only measurement for which NTM P4325 falls within the range of W. vanderleueri is dentary height (measured as the dorso-ventral height of the dentary at the posterior end of M 2 ), indicating that the species, or at least this individual, was somewhat slender jawed when compared to the most robust individuals of W. vanderleueri (e.g., NTM P87108-6).
Deeply recessed masseteric fossa. The anterior margin of the masseteric fossa of W. vanderleueri varies from a gentle change in slope of the buccal surface of the dentary resulting in a bevelled margin (e.g., NTM P87108-6) to a sharply impressed fossa with the anterior margin forming low walls perpendicular to the buccal surface. In NTM P4325 the nature of the fossa resembles the latter condition but the anterior end of the fossa is recessed under its rim, forming a blind pocket (Fig. 3A). The same condition was described in the holotype of W. oldfieldi (Clemens & Plane, 1974), although inspection of this specimen by the author reveals that the recess is barely developed and much shallower than it is in NTM P4325.
Steeply-angled I 1 . The basal section of the lower incisor of the holotype of W. oldfieldi projects anterodorsally at an angle of 27 • from the long axis of the horizontal ramus of the dentary before the apical region of the tooth curves dorsally (Fig. 10). The only I 1 of W. vanderleueri in place in a jaw (NTM P87108-5) is similarly procumbent with an angle of 31 • . In other specimens where the I 1 is missing (NTM P9273-3, P85553-4, P8695-97, P87108-6) the angle of the posterodorsal wall of the alveolus can be measured. This angle ranges from 30 to 38 • in these specimens with a mean of 34.8 • , indicating the posterodorsal wall of the alveolus is an acceptable proxy for the angle of procumbency of I 1 .
In contrast the posterodorsal wall of the alveolus for I 1 of W. alcootaensis (NTM P4325) is far more steeply-angled at close to 54 • from the long axis of the horizontal ramus (Fig. 10).
Loss of P 2 . The holotype of W. oldfieldi and all specimens of W. vanderleueri that preserve the bone between I 1 and P 3 retain a rudimentary single cusped tooth, or an alveolus for such a tooth (Clemens & Plane, 1974;Megirian, 1986; NTM P927-3, NTM P8695-97, NTM P-87108-6). This tooth is usually identified as P 2 (e.g., Megirian, 1986;Murray, Wells & Plane, 1987) although other identifications, such as P 1 or a canine, are possible. Although the presence of an upper anterior premolar is variable within W. vanderleueri (Murray & Megirian, 1990) it would appear that a lower tooth in this position is invariably present. The Alcoota dentary lacks any alveolus between I 1 and P 3 (Figs. 3A and 3C) indicating the complete loss of all lower cheek teeth anterior to P 3 .
Larger P 3 relative to M 1 . The P 3 :M 1 length ratio for W. oldfieldi is 1.19 in the holotype and 1.16 in a specimen from Riversleigh (Gillespie et al., 2014). This ratio ranges from 1.18 to 1.40 in the Bullock Creek sample of W. vanderleueri (SAM P17925, NTM P2970-26, NTM P87108-5, NTM P87108-6, NTM P85553-4). The precise ratio in NTM P4325 cannot be obtained because M 1 is missing and its length has to be taken from that of its alveolus. If this is done, a ratio of 1.50 is obtained. Thus, even allowing for estimation errors, it is clear that W. alcootaensis has a distinctly larger P 3 to M 1 ratio than W. oldfieldi and one that lies outside the range of variation seen in W. vanderleueri. Although the length of M 1 has to be estimated in NTM P4325 it is possible to compare the size of P 3 with a measureable proxy for total jaw size. If P 3 is compared to the distance from the anterior margin of the alveolus for P 3 to the anteriormost point of the masseteric fossa, similar results to the comparison of P 3 and M 1 are obtained. P 3 is 30.7% of the jaw size proxy in the holotype of W. oldfieldi, while it ranges from 31.5% to 36.0% in W. vanderleueri and is 36.4% in NTM P4325. Thus W. alcootaensis has a distinctly enlarged P 3 in comparison with W. oldfieldi and a slightly enlarged P 3 in comparison with W. vanderleueri.
Loss of M 3 The lower jaws of W. oldfieldi and W. vanderleueri bear three double-rooted molars behind the enlarged P 3 (Clemens & Plane, 1974;Megirian, 1986;Gillespie et al., 2014). The new Alcoota dentary (NTM P4325) bears just four sockets (Fig. 5), indicating only two double-rooted molars. An alternative interpretation was suggested during the review of this paper. In this interpretation the tall peak of alveolar bone observed at the anterior end of NTM P4325 is taken to mark the boundary between the last premolar and the first molar. This allows a linear row of five sockets for the roots of the molar teeth which would presumably be interpreted as receiving two double rooted molars and a posterior single rooted molar. However such an interpretation can be dismissed because the broken roots in the first two alveolar sockets form a contiguous broken surface over the peak of alveolar bone (Figs. 4C,5 and 11), indicating conclusively that they are the anterior and posterior roots of the same large premolar. Furthermore the tall peak of alveolar bone matches precisely the peak that occurs between the anterior and posterior roots of P 3 in W. vanderleueri (Fig. 11). That the four remaining alveolar sockets equate to two double rooted molar teeth is supported by the presence of two roots in all known Wakaleo lower molars, including the reduced M 3 of W. vanderleueri (Clemens & Plane, 1974;Gillespie et al., 2014). This interpretation is further supported by the dentary fragment UCMP 65621, which preserves its last two molars. These molars would appear to be homologous with M 1 and M 2 of W. vanderleueri, indicating that M 3 was absent in this specimen as well. The posterior molar of UCMP 65621 is identified as M 2 rather than M 3 because it is much larger than the reduced M 3 of other Wakaleo species both in terms of absolute size and relative size compared to the preceding molar. It also retains distinct trigonid and talonid moieties unlike the M 3 of W. oldfieldi or W. vanderleueri. In W. oldfieldi the talonid basin occupies most of the occlusal surface of the tooth, with the trigonid reduced to a raised anterior edge or absent altogether, while in W. vanderleueri reduction of M 3 has proceeded to the point that it is a simple basinless nubbin.

Variation within Wakaleo alcootaensis
As there is very little overlap between the new specimens and the holotype any discussion of variation within W. alcootaensis is restricted to size variation. The holotype has dental measurements that are about one third larger than those of W. vanderleueri. Similarly the new dentary has dental measurements that range from 16 to 35% greater than the mean value for W. vanderleueri (Table 2). However the dentary fragment UCMP 65621 is not so large, with the combined length of M 1 and M 2 only exceeding the mean value for W. vanderleueri by just over 10% (Table 2). In contrast, the isolated M 2 is slightly larger than that of the holotype of W. alcootaensis, although the difference is less than 15% of the linear measurements. These observations indicate that, like W. vanderleueri, W. alcootaensis displayed a modest range of size variation.

Evolution within Thylacoleonidae
Members of the genus Wakaleo, including W. alcootaensis, display several derived features not seen in species of Thylacoleo such as loss of the first premolar in the upper and lower jaws, presence of an anterolingual cuspule on the third upper premolar and triangular upper molars (Gillespie, 2007). These suggest that Wakaleo forms a clade to the exclusion of Thylacoleo as suggested by Clemens & Plane (1974). Nonetheless W. alcootaensis displays several derived states not present in earlier Wakaleo species but are present in Thylacoleo. These include: larger size; steeply-angled lower incisors; increased size of P 3 relative to other teeth; and reduction in the number of molar teeth. To test whether or not these character states are sufficient to remove W. alcootaensis from Wakaleo, or to nest Thylacoleo within Wakaleo as the sister taxon of W. alcootaensis, a cladistic analysis was performed. The search produced three most parsimonious trees with a length of 63 steps. The strict consensus of these trees upholds Wakalaeo as a clade including W. alcootaensis (Fig. 12).  Of the features shared between W. alcootaensis and Thylacoleo that were included in the analysis (that is all except absolute size) all were optimised as convergences between W. alcootaensis and a subset of Thylacoleo (T. crassidentatus + T. carnifex), or T. carnifex alone, at least in delayed transformation optimisation. None of them were found to be synapomorphies linking W. alcootaensis to Thylacoleo. The enlargement of P 3 and the loss of M 3 were interpreted as synapomorphies of Thylacoleonidae that were reversed in W. oldfieldi and W. vanderleueri when acctran optimisation was in place. This optimisation, although equally parsimonious within the narrow parameters of the present analysis, is incongruent with stratigraphy and is almost certainly an artefact of the high amounts of missing data for basal thylacoleonids. As new data for Priscileo roskellyae and other basal thylacoleonids become available it is likely that the ambiguity will be resolved in favour of the deltran optimisation. Thus the interpretation that the similarities between W. alcootaensis and Thylacoleo are convergent is supported by the analysis, although the strength of this support is lessened by missing data. This indicates that there has probably been a certain amount of iterative evolution in Thylacoleonidae with some character traits evolving in the late Miocene of the Wakaleo clade and again, independently, in the Plio-Pleistocene Thylacoleo clade. What selective force may be driving this convergence is unknown, although the increased size of both W. alcootaensis and later Thylacoleo, relative to other thylacoleonids hints that it may be a specialisation towards hypercarnivory and increasing prey size.
The new anatomical information and the phylogenetic analysis also allow us to revisit the position of W. alcootaensis within Wakaleo. Previous hypotheses had suggested that Wakaleo consisted of a single anagenetic lineage passing from W. oldfieldi to W. vanderleueri and finally W. alcootaensis. This hypothesis is supported by the stratigraphic succession of these taxa and the apparent morphoclinal trends that they exhibit. 'Apparent' is an appropriate qualifier because the incompleteness of both W. oldfieldi and W. alcootaensis meant that no single anatomical structure could be traced through all three species. With the addition of upper jaw material for W. oldfieldi (Gillespie et al., 2014) and lower jaw material for W. alcootaensis (this paper) these morphoclinal trends can be re-examined. The following character trends are found to be congruent with an anagenetic lineage: increasing absolute size; increasing P 3 to M 1 ratio; and progressive reduction and eventual loss of M 3 . We might also add an increasingly steeply inclined I 1 if it can be shown that like the holotype other W. oldfieldi individuals have a highly procumbent I 1 set at a lower angle to those of W. vanderleueri. However other characters are incongruent with this morphoclinal trend. Incongruent characters include the buccal height of M 2 relative to its lingual height and the excavation of the anterior margin of the masseteric fossa. W. oldfieldi and W. vanderleueri show increasing height of the buccal side of M 2 relative to the lingual side so there is a steep buccolingual gradient across the tooth. In contrast the buccal side of the M 2 of W. alcootaensis is barely any taller than the lingual side (NTM P4328). The masseteric fossa is recessed under its anterior rim in W. oldfieldi and W. alcootaensis whereas there is no recess in W. vanderleueri. These characters may be simply represent small-scale reversals within an anagentic lineage or may be indicative of a more complex branching arrangement within Wakaleo.
Although simple cladistics analysis is incapable of testing for anagenesis, with each operational taxonomic unit treated as a terminal branch, we can expect that anagenetic lineages appear as a pectinate arrangement with the constituent taxa appearing in sequence. This does not occur within the Wakaleo clade of the present analysis, casting some doubt upon the hypothesis of an anagenetic lineage in this genus. According to the analysis W. alcootaensis branched off prior to the split between W. oldfieldi and W. vanderleueri. This implies a ghost lineage for W. alcootaensis extending back to the early Miocene, for which we have no physical evidence. However an examination of the synapomorphies supporting the W. oldfieldi + W. vanderleueri clade shows that they are mostly plesiomorphic characters that have been optimised as reversals in this analysis. This may well be an artefact of the poor representation of basal thylacoleonids in the analysis and will likely change with the addition of new, currently unpublished, basal thylacoleonid material (Gillespie, 2007).
In summary, W. alcootensis would appear to be correctly placed in Wakaleo, which is supported as monophyletic but an evaluation of evolution within the genus is dependent upon the addition of new data, much of which should be forthcoming.

CPC
Commonwealth  (1); two alveoli (2) (modified from characters 1 and 2 of Gillespie, 2007). Character is treated as ordered. Given the difficulty in determining the homology of the reduced anterior teeth between the first incisor and the third premolar in diprotodontians, this character is simplified here and simply codes the variable number of alveoli present between these teeth.
3. Development of a posterolingual crest on P 3 : weakly developed to absent (0); well-developed (1) (modified from character 7 in Gillespie, 2007). The character here is simplified by subsuming the state of 'weakly developed' into the plesiomorphic state.
4. Presence or absence of a weak anterobuccal crest on P 3 : absent (0); present (1) (modified from character 9 in Gillespie, 2007). The character here is treated as a simple presence or absence character, rather than distinguishing between weakly and moderately developed derived states.
6. Height of the trigonid relative to the talonid in M 1 : trigonid subequal or lower than the talonid height (0); trigonid distinctly taller than the talonid (1); trigonid more than twice the height of the talonid (2) (modified from character 12 in Gillespie, 2007). Character is ordered.
7. Width of talonid basin of M 1 : width nearly equal to width of the crown (0); width less than 70% of the width of the crown (1); width reduced to less than 30% of the width of the crown, or near absence (2) (modified from character 13 in Gillespie, 2007). Character is ordered.
29. Presence or absence of a lateral bulge of the buccal margin of the crown, adjacent to the paracone of M 2 : absent (0); present (1) (from Gillespie, 2007).
31. Angle of the long axis of I 1 to the long axis of the horizontal ramus of the dentary: less than 40 • (0); greater than 40 • (1). Character is new.
32. Development of the masseteric fossa: fossa is not recessed under the anterior margin (0); fossa is recessed under the anterior margin (1). Character is new.
34. Depth of buccal margin of M 2 in comparison to the lingual margin: buccal and lingual margins of similar height (0); buccal margin raised relative to the lingual margin (1) (from Gillespie, 2007).

APPENDIX 2. TREE DESCRIPTION
The second of two most-parsimonious-trees (where Thylacoleo hilli is resolved as the sister taxon of T. crassidentatus + T. carnifex) is described, however only ingroup clades that are present in the strict consensus tree are described. State changes are given in brackets. An asterisk denotes a state change that occurs once, without homoplasy.

Clade 1. Thylacoleonidae
Unambiguous synapomorphies. Character 5 (1 to 2, or 0 to 2)*, bunodont molar cusps. Acctran optimisation supports the state change from bunolophodont to bunodont (1 to 2), whereas deltran optimisation interprets the state change as selenodont to bunodont (0 to 2). In either case the change to bunodonty is unambiguously tied to this clade and is an unambiguous synapomorphy of Thylacoleonidae. Character 15 (0 to 1)*: posterior longitudinal blade of P 3 is nearly horizontal and bowed. Character 24 (0 to 1)*: main sectorial blade of P 3 is longitudinally bowed. Character 26 (0 to 1)*: an anteroposterior gradient on the buccal side of M 1 with a taller anterior end. Character 29 (0 to 1)*: presence of a lateral bulge of the buccal margin of M 2 adjacent to the paracone. Ambiguous synapomorphies under acctran optimisation. Character 2 (1 to 3): a P 3 to M 1 ratio of 1.5 or more. Unknown in Priscileo roskellyae and reversed in Wakaleo oldfieldi + Wakaleo vanderleueri. Deltran optimisation interprets this chacter state as a convergence between Wakaleo alcootaensis and Thylacoleo crassidentatus + Thylacoleo carnifex. With the inclusion of new data from basal thylacoleonids this ambiguity will almost certainly resolve in favour of the deltran optimisation. Character 3 (0 to 1): presence of a well-developed posterolingual crest on P 3 (reversed in Thylacoleo crassidentatus + Thylacoleo carnifex). Deltran optimisation interprets this character as a convergence between Wakaleo oldfieldi + Wakaleo vanderleueri and Thylacoleo hilli. Character 6 (0 to 1)*: trigonids of lower molars taller than the talonids. Ambiguous due to missing published information on the lower molars of Priscileo roskellyae. Deltran optimisation finds this character state change on the branch supporting Wakaleo + Thylacoleo. Character 7 (0 to 1)*: talonid basins of lower molars distinctly narrower than the crown. Ambiguous due to missing published information on the lower molars of Priscileo roskellyae. Deltran optimisation finds this character state change on the branch supporting Wakaleo + Thylacoleo. Character 8 (0 to 1)*: talonid moiety of M 1 narrower than the trigonid moiety. Ambiguous due to missing published information on the lower molars of Priscileo roskellyae. Deltran optimisation finds this character state change on the branch supporting Wakaleo + Thylacoleo. Character 9 (0 to 1)*: trigonid of M 2 much higher than its talonid. Ambiguous due to missing published information on the lower molars of Priscileo roskellyae. Deltran optimisation finds this character state change on the branch supporting Wakaleo (the state cannot be determined in Thylacoleo due to loss of the talonid in M 2 ). Character 11 (0 to 1): loss of M 3 (reversed in Wakaleo oldfieldi + Wakaleo vanderleueri). Deltran optimisation interprets this chacter state as a convergence between Wakaleo alcootaensis and Thylacoleo crassidentatus + Thylacoleo carnifex. With the inclusion of new data from basal thylacoleonids this ambiguity will almost certainly resolve in favour of the deltran optimisation. Character 12 (0 to 1)*: loss of M 4 . Ambiguous due to missing data for Priscileo roskellyae, with deltran optimisation finding this character to be a synapomorphy of Wakaleo + Thylacoleo. Given that the upper tooth row of Priscileo roskellyae retains M 4 , it is probable that the lower tooth row retained M 4 . If this is found to be the case then the ambiguity will resolve in favour of the deltran optimisation.
Ambiguous synapomorphies under acctran optimisation. Character 28 (0 to 1)*: loss of metaconule on M 2 . Ambiguous due to missing data in Thylacoleo. This character state change is a synapomorphy of Wakaleo in delayed transformation.
Ambiguous synapomorphies under deltran optimisation. Character 2 (1 to 2): a P 3 to M 1 ratio greater than 1.1. See discussion of this character above. Character 6 (0 to 1)*: trigonids of lower molars taller than the talonids. See discussion of this character above. Character 7 (0 to 1)*: talonid basins of lower molars distinctly narrower than the crown. See discussion of this character above.

Clade 3. Wakaleo
Unambiguous synapomorphies. Character 18 (0 to 1)*: presence of a lingual cuspule below the anterior cusp of P 3 . Character 20 (0 to 1)*: triangular occlusal outline of M 1 . Character 22 (0 to 1): loss of P 1 . The absence of the P 1 in two of the outgroup taxa (Nimiokoala greystanesi and Namilamadeta albivenator) renders the optimisation of this character at the base of the tree ambiguous. Nevertheless this tooth is present basally in Thylacoleonidae and its loss can be unambiguously tied to Wakaleo within Thylacoleonidae. Character 25 (0 to 1)*: posterobuccal margin of M 1 lengthened and overlapping M 2 in lateral view. Character 27 (0 to 1): occlusal outline of M 2 subtriangular to triangular.