Cookeroo, a New Genus of Fossil Kangaroo (Marsupialia, Macropodidae) from the Oligo-Miocene of Riversleigh, Northwestern Queensland, Australia

ABSTRACT A new genus of Oligo-Miocene kangaroo (Macropodiformes), Cookeroo, and two new species, Cookeroo bulwidarri and C. hortusensis, are described from the Riversleigh World Heritage Area, northern Australia. Species of Cookeroo are distinguished from other basal macropodids by possessing a unique combination of characters including an expanded masseteric canal confluent with the mandibular canal that extends to below m1, a sinuous i1 with enamel on the buccal surface only and with dorsal and ventral flanges present, a dentary with a marked inflection of the ventral border below m3, bilophodont molars, and an elongate third premolar. We assess the phylogenetic relationships of the genus using a combination of two previously published morphological matrices. Our analysis recovers Cookeroo species as early-branching members of a clade that also contains macropodines, sthenurines, and lagostrophines.

The bulungamayine clade was initially considered by Flannery et al. (1983) to be nested within Potoroidae. It was later placed within Macropodidae due to shared characters including long anterior cingulids on lower molars, bilophodont molars, protolophids on the lower first molar that are not laterally compressed, a deciduous premolar that is reduced or absent, and alisphenoid-parietal contact on the lateral wall of the cranium (Cooke, 1997b(Cooke, , 1999Cooke and Kear, 1999;Kear and Cooke, 2001). Several other plesiomorphic macropodid genera have also been described, including Ganguroo (Cooke, 1997b), Wakiewakie (Woodburne, 1984), Wanburoo (Cooke, 1999), and Gumardee (Flannery et al., 1983). As a result, Riversleigh represents the greatest diversity of fossil kangaroos from any fossil locality. Additional unidentified macropodid taxa have been recovered from the Riversleigh WHA. In this study, in addition to more recently discovered material, reexamination of unpublished specimens attributed to Wabularoo naughtoni by Cooke (1997a) suggests that they have been incorrectly assigned. These are described here as a new genus consisting of two new species.
Previous phylogenetic analyses have been unable to resolve the interrelationships of relatively plesiomorphic macropodids such as species of Ganguroo (Kear and Pledge, 2007;Travouillon et al., 2014). Analysis by Prideaux and Warburton (2010) places specimens of Nowidgee matrix, now synonymized with Bulungamaya delicata , at the base of the clade containing macropodines, sthenurines, and lagostrophines and with Ganguroo as a sister taxon to sthenurines. The unresolved and potentially polyphyletic nature of Bulungamayinae has led to the recommendation that it be treated as Macropodidae incertae sedis (Kear and Pledge, 2007;Prideaux and Warburton, 2010). As a result, species previously attributed to the 'Bulungamayinae' clade are here referred to as basal macropodids. Phylogenetic relationships among macropodiforms are reassessed, in light of this new material.

Materials and Measurements
Specimens used in this study were collected from several fossil sites in the Riversleigh WHA, northwestern Queensland. Specimens are housed in the Queensland Museum fossil collection (QM F), Brisbane, Australia. Specimens were measured using digital calipers. For each specimen, maximum tooth length and width of the third premolar was measured at the base of the crown. Maximum tooth length and anterior and posterior widths were measured for all molars. Measurements of width were taken across the base of the anterior and posterior lophs. All dental measurements are given in Table 1-3.

Metric Analysis
Metric analyses were done using the software PAST version 3.01 (Hammer et al., 2001). Coefficients of variation (CVs) were calculated for dental measurements in Cookeroo hortusensis, gen. et sp. nov., in order to determine whether the amount of variation present within the species is consistent with that expected for a mixed sex marsupial population. Travouillon et al. (2014) calculated CVs for the extant macropodids Thylogale stigmatica (3.54-12.5) and T. thetis (4. 98-11.16). These values are used here as the expected range of variation for dental measurements of macropodiform species. The sample size of C. bulwidarri, sp. nov., is too small for CVs to provide meaningful results. Bivariate plots (length vs. anterior, or posterior, width of the third premolar and all molars) were made for both species of Cookeroo and taxa identified on the basis of phylogenetic analysis (see below) as those likely to be closely related to Cookeroo. Either anterior or posterior width was used for bivariate plots depending on which measurement was able to best distinguish between species. Principal component analysis (PCA) was performed using log-transformed dental measurements for all cheek teeth, in order to determine which variables best distinguish between the species.

Phylogenetic Analysis
Previous attempts to analyze the phylogenetic relationships of primitive macropodiforms using a matrix by Kear and Pledge (2007) have resulted in highly unresolved phylogenetic trees (see Cooke, 1996;Kear et al., 2001aKear et al., , 2001bKear, 2002; for parent matrices), particularly for basal macropodids such as Ganguroo species . A second taxon-character matrix by Prideaux and Warburton (2010) has since been developed to investigate the interrelationships of bilophodont macropodids. We suggest that combining the Kear and Pledge (2007) and Prideaux and Warburton (2010) matrices may resolve the phylogeny of relatively plesiomorphic macropodiforms by increasing both taxon and character sampling. In this study, the two matrices were therefore combined, with the matrix of Prideaux and Warburton (2010) used as a base for our matrix because it includes more representatives of all macropodid subfamilies, including the recently recognized subfamily, Lagostrophinae.
Preliminary phylogenetic analysis using the combined Prideaux and Warburton (2010) and Kear and Pledge (2007) matrices resulted in the balbarid family being nested in Macropodidae when all characters were unconstrained. This may be because parsimony analysis without constraints only allowed for a single evolution of a bilophodont dentition, which is hypothesized to have evolved independently in balbarids and macropodids (Cooke, 1997b(Cooke, , 1997c. In this study, three characters were therefore alternately treated as unconstrained or constrained as irreversible: characters 23 (molar morphology), 38 (p3 morphology), and 116 (height of third premolar). Character 23 is constrained in order to represent the unidirectional evolution of bilophodonty from a bunolophodont ancestral form that is hypothesized to have occurred twice in relatively plesiomorphic macropodiforms (Cooke, 1997b(Cooke, , 1997c. Character 38 is constrained to reflect the unidirectional evolution of the premolar in macropodiforms Abbreviations: AW, anterior width (mm); dP, deciduous premolar; L, tooth length (mm); M, molar; P, premolar; PW, posterior width (mm). Abbreviations: AW, anterior width (mm); dp, deciduous premolar; L, tooth length (mm); m, molar; p, premolar; PW, posterior width (mm). from plagiaulacoid to sectorial. Character 116 is constrained to reflect the unidirectional evolution of third premolar from premolars that extend above the level of the premolar to teeth that extend to the height of the third molar. A parsimony analysis was performed using PAUP 4.0b10 (Swofford, 2002) for both the unconstrained matrix and a matrix tree with three characters constrained as irreversible. A twostage heuristic search was conducted, with the initial search comprising 1000 replicates, saving 10 trees per replicate, followed by a second search within the saved trees. Most parsimonious trees from the heuristic search were summarized as a strict consensus tree. Bootstrap values were calculated using 1000 replicates. Decay indices were also calculated using TreeRot.v3 (Sorenson and Franzosa, 2007). In order to test whether topologies for the constrained and unconstrained matrices differ significantly, a Templeton test (Templeton, 1983) was conducted. Comparison of these topologies will allow also for the identification of potentially homoplastic characters within the matrix. Type Species-Cookeroo bulwidarri, sp. nov. Generic Diagnosis-Macropodids with the combination of anteriorly expanded masseteric canal confluent with the mandibular canal that extends to below m1; sinuous i1 with enamel on the buccal surface only and with dorsal and ventral flanges present; dentary with a marked inflection of the ventral border below m3; shallow buccinators sulcus; elongate third premolar; bilophodont molars; m4 not greatly reduced in size; no protostylid on m1; no posterior cingulid on lower molars; slight postmetacristid and postentocristid on m1; parietal-alisphenoid contact; and StD present on M1.
Etymology-Named after Bernard Cooke who has made highly significant contributions to the field of kangaroo evolution and palaeobiology, and especially for his important contributions to Riversleigh fossil research.
Age and Distribution-All specimens are from White Hunter Site, Riversleigh WHA, northwestern Queensland. This site is interpreted to be part of Riversleigh's Faunal Zone A, which is hypothesized to be late Oligocene in age (Archer et al., 1989(Archer et al., , 1997Travouillon et al., 2006Travouillon et al., , 2011Arena et al., 2015).
Species Etymology-Named bulwidarri, meaning 'white' in the Waanyi language, for the type locality, White Hunter Site.
Species Diagnosis-This species differs from Cookeroo hortusensis in having the following combination of features: upper molars having distinct contact between the postparacrista and premetacrista forming a centrocrista; P3 with eight cusps and associated transcristae; poorly developed jugal crest; upper cheekteeth that are not buccally convex in the occlusal plane; less developed basioccipital crest; and more cuspate upper molars.

Description
Nasal and Premaxilla-In dorsal view, only a small triangular portion of the nasal is preserved on the left side, preventing a substantial description of this bone (Fig. 1). In lateral view, the small portion of the premaxillary-maxillary suture preserved extends ventrally to contribute to the posterior margin of the canine alveolus. In ventral view, the canine alveolus is larger than those for the three incisors. The sizes of the incisive alveoli indicate that I1 is likely to have been the largest incisor, followed by I3 and then I2. Incisive fenestrae are present on the ventral side of the premaxilla; the right side is crushed and poorly preserved. The incisive fenestrae are ovoid, long, and extend from level with the posterior end of I3 to level with the premaxillarymaxillary suture. A long diastema is evident between the P3 and C1 alveoli (approximately 20 mm from anterior of canine alveolus to anterior of P3).
Maxilla and Palatine-The infraorbital foramen is slightly elliptical, opens anteriorly, and is located above the anterior end of P3 (Fig. 1). The masseteric process is poorly developed. The suborbital shelf of the maxilla is flat and broader anteriorly than it is posteriorly. There is a deep infraorbital fossa within the anterior portion of the suborbital shelf. The sphenopalatine foramen is positioned on the suborbital shelf posterior to the infraorbital canal. Both the sphenopalatine foramen and infraorbital canal are oval in shape. The margin of the suborbital shelf continues past the maxilla-palatine suture onto the palatine. It is unclear how much of the palatine contributes to the lateral wall of the orbit. In ventral view, the maxillopalatine fenestrae are not well preserved; however, the anterior margin is preserved in the maxilla, level with M1, and the palatine contributes to its posterior margin from the anterior of M3 to posterior of M4.
Lacrimal-Very little of the lacrimal is preserved. In lateral view, a single lacrimal foramen occurs on the anterior margin of the orbit (if a second was present, typical of macropodiforms, it is not preserved here), small and posteriorly directed. The posterior extent of the lacrimal is not preserved.
Frontal, Parietal, and Interparietal-The left frontal is partly preserved and is remarkably flat. A small sagittal crest is present at the parietal-parietal suture (Fig. 1). Although faint, there is an evident contact between the parietal and alisphenoid on the lateral wall of the cranium. The interparietal-supraoccipital sutures form a well-defined crest. The parietal contacts the squamosal, forming the lateral walls of the neurocranium.
Zygomatic Arch-The jugal-maxilla suture begins anteriorly, level with the lacrimal and is posteroventrally directed. A crest that extends posteriorly from the jugal-maxilla suture along the jugal is present on the lateral surface. The zygomatic arch transitions smoothly into the facial region. The anterior portion of the jugal-squamosal suture is 'V'-shaped. The posterior end of the jugal is overlapped by the zygomatic process of the squamosal where it forms the anterior margin of the glenoid fossa. Although a part of the zygomatic process is not preserved, it appears to be relatively shallow. The glenoid fossa is broad and flat. The postglenoid process is small. The postglenoid foramen is located just mesial to the postglenoid process and is large and oval in shape.
The occipital condyle is of intermediate size. The basioccipital crest is well developed. The mastoid process is more massive than the paracondylar process but shorter. It is inclined ventrolaterally.
In lateral view, the ectotympanic is 'mushroom'-shaped in outline, with a concave dorsal wall. The external auditory meatus is enclosed dorsally by the squamosal and ventrally by the ectotympanic. The supraoccipital is remarkably flat.
Upper Dentition-The following description is based primarily on the upper dentition preserved in the holotype specimen QM F19798 (Fig. 1).
The P3 is the only premolar preserved in C. bulwidarri. No anterior premolar alveoli are present in the specimens preserved, although a P2 is present and displaced by the P3 in C. hortusensis. The P3 is long and blade-like and is slightly concave in lateral view. In occlusal view, the occlusal margin of P3 is straight. Eight cuspules are present along the occlusal margin, with the most posterior being the largest. The seven anterior cuspules support buccal and lingual transcristae that extend to the base of the crown. Only a faint lingual transcrista is associated with the eighth cuspule, and it connects to a well-developed posterolingual cusp.
The M1 is low crowned and bilophodont. It is square in occlusal outline, with rounded corners. The buccal cusps are larger and taller than the lingual cusps. The protoloph and metaloph are similar in width. The metacone is the tallest cusp, followed by the paracone, metaconule, and protocone. The protocone is more massive than the paracone. The metacone is more massive than the metaconule. The buccal margins of both the paracone and metacone are inclined steeply. The protoloph is formed by a lingual crest from the paracone that extends to meet a small but distinct preprotocrista approximately in the middle of the loph. A forelink is present anteriorly that separates the anterior cingulum and precingulum. The anterior cingulum extends from the buccal anterior margin to the forelink, whereas the precingulum extends from the forelink to the lingual anterior margin. The prominence of the precingulum and the forelink appears to have been reduced by wear. A welldefined StC is present on a ridge posterobuccal to the paracone. A less developed StD is present on the anterobuccal ridge of the metacone. The postprotocrista extends posterobuccally toward the interloph valley, forming a midlink approximately in line with the neometaconule. The postparacrista extends posteriorly to meet the premetacrista in the interloph valley, forming a centrocrista. A prominent neometaconule is present along the midline of the metaloph. A neometaconule crista, or postlink, extends toward the posterior end of the tooth from the neometaconule. The postmetacrista extends posteriorly from the metacone and meets the postmetaconule crista at the posterior margin of the tooth.
The M2 is similar in shape and morphology to M1 except for the following features: it is generally longer and wider; StC is reduced to a small crest; StD is absent; the forelink is present on the right M2 but absent on the left; and both the precingulum and anterior cingulum are longer.
The M3 is similar in morphology to M2 except as follows: the protoloph is wider than the metaloph; StC is absent; the neometaconule and postlink are less developed; and postparacrista, premetacrista, and centrocrista are less developed.
The M4 is similar in morphology to M3 except for the following: the tooth is narrower than M3; the metaloph is significantly shorter relative to the protoloph; postparacrista and premetacrista are further reduced and do not connect to form a centrocrista; and neometaconule and postlink are absent.
Lower Dentition-The lower dentition is preserved only in QM F19875 (Fig. 1). Only m1 and p2 are present. The p2 is plagiaulacoid in form and is slightly shorter than m1. The occlusal margin is shorter than the crown base of the tooth. Five cuspids and associated buccal and lingual transcristids are present for the four anterior cuspids. The lingual face of p2 is steeper than the buccal face.
The m1 has a roughly rectangular outline (wider posteriorly than anteriorly) with round corners. The hypoconid is the tallest cuspid, followed by the metaconid, entoconid, protoconid, and paraconid in order of decreasing height. The m1 is bilophodont, with the metaconid connected to the protoconid via a straight protolophid. The hypolophid is wider than the protolophid. The entoconid is connected to the hypoconid via a slightly concave hypolophid. The cristid obliqua runs anterolingually from the hypolophid and terminates in the interlophid valley posterior to the metaconid. The premetacristid is present and extends anteriorly from the metaconid. The postmetacristid extends posterolingually from the metaconid. The anterior cingulum is enclosed by the premetacristid and paracristid. A distinct paraconid is present but is not connected to the protoconid. A small precingulid is present anterior to the paracristid. A small preentoconid is present anterior to the entoconid. A short crest is also present on the buccal side of the entoconid. A short postentocristid is present on the posterior side of the entoconid.
Holotype-QM F57792, partial adult skull with left and right P3s and M1-4s and broken right M1 and 2 from Neville's Garden Site.
Age and Distribution-The holotype, QM F57792, and the paratype QM F57629, are from Neville's Garden Site, Riversleigh WHA, northwestern Queensland. The Neville's Garden deposit at Riversleigh is interpreted as being part of Riversleigh's Faunal Zone B and thus early Miocene in age (Archer et al., 1989(Archer et al., , 1997Travouillon et al., 2006Travouillon et al., , 2011. Neville's Garden is interpreted as interval B3, the latest of the Faunal Zone B intervals, by Arena et al. (2015), and it has been dated at 18.24 § 0.29 Ma by Woodhead et al. (2014), early Miocene in age. Paratypes QM F19813, QM F19585, QM F19610, and QM F20260 are from Wayne's Wok Site, whereas QM F57637 is from Camel Sputum Site and QM F20072 is from Mike's Menagerie Site. These sites are also considered part of Faunal Zone B, interval B3 (Arena et al., 2015). Other referred specimens are also from sites considered part of Faunal Zone B, interval B3: Cadbury's Kingdom Site, Camel Sputum Site, In Abeyance Site, Mike's Menagerie Site, Neville's Garden, RSO Site, Upper Site, Wayne's Wok Site (Travouillon et al., 2006;Arena et al., 2015). Creaser's Ramparts Site and Price Is Right Site are interpreted as Faunal Zone B, interval B2 or B3, whereas Dirk's Tower Site is interpreted as Faunal Zone B, B1 (Travouillon et al., 2006;Arena et al., 2015). Rat Vomit Site is interpreted as Faunal Zone B (Travouillon et al., 2006).
Species Etymology-Named hortusensis, Latin meaning 'coming from,' or 'belonging to,' 'the garden,' for the type locality, Neville's Garden Site.
Species Diagnosis-Cookeroo species with no distinct connection between the postparacrista and premetacrista on upper molars; long P3 with nine cusps and associated transcristae; welldeveloped jugal crest; smaller skull than C. bulwidarri; upper cheek teeth bowed laterally in occlusal plane; tall ventral crest on the basioccipital; and more distinct lophs.

Description
Nasal and Premaxilla-The dorsal surface of the rostrum is broad anterior to the orbital region and slightly domed (Fig. 2). In QM F57792, the posterodorsal portion of the premaxilla is a thin flange that contacts both the nasal and the maxilla. The anterior portion of the premaxilla is not preserved in QM F57792 but is partially preserved in QM F19610 (Fig. 2). The anterior portion of a canine alveolus is evident on QM F19610, but canines are not known. Three incisive alveoli are present, with I1 and I2 preserved. The alveolus for I3 is ventrally directed, whereas those for I1 and I2 are anteroventrally directed. The anterior and lateral walls of the incisive fenestrae are preserved. The C1 alveolus is broken laterally in QM F57792, making it difficult to accurately assess the length of the diastema, but the anterior-most portion of the alveolus is roughly at the same distance as it is in the skull of the other species (20 mm from anterior of canine alveolus to anterior of P3).
Maxilla and Palatine-In lateral view, the infraorbital foramen is slightly elliptical and is located anterior to the orbital margin, dorsal to the center of P3 (Fig. 2). The masseteric process is poorly developed and ventrally does not reach as far as the alveolar margin. The maxilla contributes the majority of the ventral wall of the orbit. The suborbital shelf of the maxilla is broader anteriorly than posteriorly. An infraorbital fossa is present on the anterior portion of the suborbital shelf. The infraorbital canal is oval in shape. The sphenopalatine foramen is positioned posterior to the infraorbital canal and is oval in shape. The mesial margin of the suborbital shelf extends onto the palatine. The maxillopalatine fenestrae are not well preserved, but the anterior margins are aligned with the posterior end of M1. There is little wear on M1-3, and none on M4, indicating that the holotype is most likely a young adult.
Lacrimal-The posterior wall of the lacrimal forms a portion of the anterior wall of the orbit. The ventral-most lacrimal foramen is obscured by broken bone and the dorsal-most lacrimal foramen is broken.
Frontal, Parietal, and Interparietal-The frontal region is broad in the interorbital region, not constricted as is seen in balbarids, and is relatively flat. In dorsal view, the frontal-parietal sutures run posteromedially, meeting in the center of the skull to form a wedge between the parietals. The parietal-alisphenoid sutures are preserved in QM F23837, where there is a clear parietal-alisphenoid contact. In dorsal view, there is a postorbital constriction of the skull. A small, poorly developed sagittal crest is evident along the parietal-parietal suture, which continues through the interparietal (Fig. 3). The interparietal-supraoccipital suture forms a well-defined crest.
Zygomatic Arch-The jugal extends anteriorly to level with the anterior of the lacrimal. A crest is present on the lateral surface of the jugal and extends posteriorly. Anteriorly, the zygomatic arch transitions smoothly into the facial region. The jugalmaxillary suture is posteroventrally directed. The anterior portion of the jugal-squamosal suture is 'V'-shaped, and the posterior end of the jugal is overlapped by the zygotic process of the squamosal. This process is relatively shallow with a flat lateral surface. In ventral view, the posterior portion of the jugal forms the anterior margin of the glenoid fossa. The glenoid fossa is broad and flat in shape and appears to merge smoothly with the zygomatic arc. A small postglenoid process is present. The postglenoid foramen is also present mesially.
Neurocranium-The posterior part of the neurocranium is preserved in QM F23837. The parietals and frontals appear to have formed most of the roof of the anterior end of the neurocranium. The alisphenoid contributes to the anteroventral wall of the neurocranium. The posterior end of the neurocranium is formed dorsally by the parietals and interparietals, and laterally by the squamosals.
Basicranium-The basicranium is preserved in QM F23837. The jugular foramina lie posterior to the alisphenoid and mesial to the anterior of the paracondylar process of the exoccipital. The majority of the posterior of the paracondylar and mastoid processes are not preserved. The alisphenoid contributed to the FIGURE 3. Posterior portion of the skull of Cookeroo hortusensis (paratype QM F23837) in A, dorsal view; B, right lateral view; C, ventral view; and D, left lateral view. Abbreviations: as, alisphenoid; bo, basioccipital; bs, basisphenoid; cf, caratoid foramen; eam, exit of auditory meatus; ec, ectotympanic; fr, frontal; gf, glenoid fossa; ip, interparietal; jf, jugular foramen; ma, mastoid; map, mastoid process; oc, occipital; pa, parietal; pap, paracondylar process of the exoccipital; pe, petrosal; sfo, secondary foramen ovale; smf, suprameatal foramen; sq, squamosal; zes, zygomatic epitympanic sinus. anterior face of the paracondylar process of the exoccipital. The tympanic wing of the alisphenoid is flat and poorly inflated. The carotid foramen is tear-shaped, positioned just lateral to the basisphenoid-basioccipital suture and walled laterally by the pterygoid. The secondary foramen ovale is oval and opens on the mesial side of the alisphenoid tympanic wing. The alisphenoid contributes to its posterior and lateral walls. A thin wall separates the secondary foramen ovale from the foramen of the greater petrosal nerve. A remnant of a primary foramen ovale is present anterior to the greater petrosal nerve opening, walled posteromesially by the petrosal and anterolaterally by the alisphenoid. The ceratoid foramen is present anterior to the basisphenoid-basioccipital suture, lateral to the pterygoid.
The occipital condyle is of intermediate size and extends more posteriorly than in species of Ganguroo. Two foramina are present on the occipital condyle, oval in shape, with the posteriormost twice the size of the anterior-most foramen, and both canals open posteriorly into the foramen magnum. The posterior appears to be the condylar foramen, whereas the anterior appears to be the hypoglossal foramen. The foramen magnum is oval in shape. The basioccipital has a well-developed crest, equivalent to the condition seen in Macropus, and is roughly hexagonal in outline.
The lateral wall of the ectotympanic is concave, whereas the mesial and anterior walls are slightly convex. The posterior wall of the ectotympanic is straight. The external auditory meatus is enclosed dorsally by the squamosal and ventrally by the ectotympanic. The zygomatic epitympanic sinus is enclosed entirely by the squamosal. The suprameatal foramen is located dorsal to the auditory meatus on the posterior lateral wall of the squamosal. The supraoccipital is flat.
Upper Dentition-The I1 and I2 are represented in the paratype, QM F19610. The I1 is the longest and tallest upper incisor. It is curved anteriorly in buccal view, with an anteroventrally orientated alveolus. It is chisel-like in shape. The I2 is oval in occlusal outline and blade-like. The I3 is not present in QM F19610, but the size of its alveolus suggests that it would have been larger than I2 but not I1.
In occlusal view, the molar row is slightly buccally convex (Fig. 2). In buccal view, the molar row is slightly concave. The P2 and dP3 are preserved in QM F19585 (Fig. 4). The P2 is short and blade-like, with straight lateral and occlusal margins. Six cuspules are present along the occlusal margin, each supporting buccal and lingual transcristae. The occlusal margin is shorter than and not parallel to the crown base. The P2 appears to be displaced by the eruption of the P3. No other anterior premolar alveoli are present.
The dP3 is bilophodont with a poorly developed protoloph and a well-developed metaloph. It is shorter and wider than P2 but narrower than M1. The tooth has a convex lingual margin and a straight buccal margin. The tallest cusp is the paracone, followed by the metacone, metaconule, and protocone. Both the paracone and metacone are laterally compressed. The postparacrista extends posteriorly to meet the premetacrista in the interloph valley, forming a centrocrista. The postmetacrista extends posteriorly, where it meets the postmetaconule crista. The protoloph is anteriorly convex and occlusally concave. A poorly developed postprotocrista extends posterobuccally towards the middle of the metaloph and ends in the interloph valley. A prominent neometaconule and associated postlink are present in the middle of the metaloph.
The P3 as preserved in QM F57792 is long and blade-like, with a concave occlusal edge (Fig. 2). The tooth is wider anteriorly than posteriorly and is approximately twice the length of M1. There are nine cuspules along the occlusal margin of the tooth, of which the most posterior cuspule is largest. The anterior eight cuspules support buccal and lingual transcristae, which extend to the base of the crown.
There is a well-developed lingual cuspule with associated crest on the posterior end of the occlusal blade.
The M1 is square in occlusal outline, low crowned, and bilophodont, with the metaloph and protoloph being similar in length. The lingual cusps are larger than, but not as tall as, the buccal cusps. The metacone is the tallest cusp, followed by the paracone, metaconule, and protocone. The paracone and metacone are steeply inclined on the buccal side. The protocone is more massive than the paracone. The protoloph is formed by a lingual crest that extends from the paracone to the protocone. A remnant of the preprotocrista extends anteriorly as a 'forelink' separating the short anterior cingulum and precingulum. The anterior cingulum extends from the buccal anterior margin to the forelink, whereas the precingulum extends from the forelink to the lingual anterior margin. A distinct StC is present on a ridge on the posterobuccal flank of the paracone. StD is present on the anterobuccal side of the metacone, but this cusp is not well developed and is reduced to a short crest. The postprotocrista extends posterobuccally to the interloph valley and then curves posteriorly toward the metaloph. A prominent neometaconule and postlink are present along the midline of the metaloph. There does not appear to be any connection between the postparacrista and the premetacrista as a centrocrista. The postmetacrista extends posteriorly where it meets the postmetaconule crista. In QM F20260, there is a minute cusp on the lingual side anterior to the metaconule; however, this does not appear in any other referred specimens and may therefore be a unique feature of this specimen. A short postentocristid is present on the posterior side of the entoconid.
The M2 is similar in morphology to M1 except for the following features: it is larger and wider than M1; the protoloph is wider than the metaloph in occlusal view; the tallest cusp is the metacone, followed by the metaconule, paracone, and protocone; StC is less prominent; StD is absent; and the forelink, anterior cingulum, and precingulum are more developed, whereas the postlink is less developed.
The M3 is similar in morphology to M2 except for the following features: it is significantly longer and wider than M2; the protocone is significantly taller; StC is further reduced or absent; and StD is absent. The neometaconule and postlink are further reduced. The premetacrista is shorter; the anterior cingulum and precingulum appear longer.
The M4 is similar to M3 except for the following features: it is slightly smaller than M3; the metaloph is considerably narrower; StC is absent; a small cusp is sometimes present anterolingual to the metaconule such as in QM F20260; the metacone and metaconule are shorter; and the neometaconule and postlink are not present.
Dentary-The dentary has a marked inflection of the ventral border below m3 (Fig. 5). The horizontal ramus is twisted such that the molar row is inclined buccally towards the anterior of the dentary. The buccinator sulcus is shallow and runs posteroventrally from below m1 to the posterior end of m2. The mandibular symphysis extends anteriorly from the posterior margin of p3 and is markedly rugose. The diastema is preserved in QM F57629 and is approximately the length of p3. Two mental foramina are present on the buccal surface of the dentary, one anterior to p3, below the diastemal margin, and one smaller mental foramen below the anterior of m3. The masseteric foramen is large, and the masseteric canal extends anteriorly to below m1. The ascending ramus is not entirely preserved in any specimen but is partly preserved in QM F19813. The angle formed by the occlusal margin and the ascending ramus is approximately 115 degrees. The masseteric canal is buccally expanded and confluent with the mandibular canal.
Lower Dentition-The i1 is preserved in QM F57629 and QM F57637 and is sinuous with enamel principally on the buccal side of the tooth. Enamel phalanges are present on both the dorsal and ventral sides. This condition is consistent with basal macropodids such as species of Ganguroo (Kear and Cooke, 2001). The molar row is straight in lateral and occlusal views.
The p2 and dp3 are preserved in QM F57637 (Fig. 6). The p2 is plagiaulacoid-like and is approximately the same length as m1. Five cuspids and associated buccal and lingual transcristids are present along the tooth. The occlusal margin of the tooth is shorter than the crown base. The p2 appears to be displaced by the p3. No alveoli for other anterior premolars are evident.
The dp3 is a molariform tooth but shorter than p2. The tooth is subrectangular in shape, with a taller lingual margin than buccal margin. The trigonid is laterally compressed. The paracristid is anteroposteriorly oriented. The protoconid is the tallest cusp, followed by the paraconid, entoconid, and hypoconid. The metaconid is significantly reduced and is evident by the postmetacristid extending posteriorly toward the interloph valley, which contacts the preentocristid. The entoconid is connected to the hypoconid by a hypolophid. The cristid obliqua extends anterolingually from the hypoconid, terminating in the interloph valley. The p3 of QM F19813 is elongated, gracile, and blade-like with an elliptical occlusal outline. In occlusal view, the midline curves slightly. The p3 is approximately twice the length of m1 and is wider than the anterior width of that tooth. Nine cuspids are present along the occlusal margin, with the posterior-most cuspid being the largest. The anterior eight cuspids have associated lingual and buccal transcristids. The number of cuspids in other specimens varies between seven and nine. This variation appears to be the result of dental wear, where some cuspids are worn away on some specimens as a result of mastication. The shape of p3 varies between specimens. A number of specimens, including QM F20072, QM F19585, and QM F57629, have a more buccally concave midline and appear more robust, whereas specimens such as QM F19813, QM F57634, and QM F20084 have less buccally concave molars.
connected to the protoconid via a straight protolophid. The paraconid is a minute cuspid anterolingual to the protoconid on m1. It is connected to the protoconid by a weak and sinuous paracristid. A moderately developed premetacristid is present and joins the paraconid anterobuccally, enclosing the small anterior cingulid. A very small precingulid is present buccal to the paracristid. A short postmetacristid is present but ends before reaching the interlophid valley. The entoconid is connected to the hypoconid via a concave hypolophid. The hypolophid is concave both anteriorly and occlusally. The cristid obliqua runs anterolingually from the hypolophid and terminates in the interlophid valley posterior to the protoconid. A short preentocristid is present and descends the anterior flank of the entoconid but ends before reaching the interlophid valley.
The m2 is similar in morphology to m1 but differs as follows: it is longer and wider in occlusal outline; the protolophid is concave both anteriorly and occlusally; the protolophid and hypolophid are similar in width; the anterior cingulid and precingulid are wider; and there is no postmetacristid or preentocristid.
The m3 is similar to m2 but differs as follows: the tooth is longer and wider in occlusal outline; the hypolophid is narrower than the protolophid; and the paraconid is more distinct.
The m4 is similar to m3 in outline but differs as follows: it has a considerably narrower hypolophid; and the paraconid is less developed. The m4 is not significantly smaller than m3.

Coefficient of Variation Analysis
Coefficients of variation for tooth measurements of Cookeroo hortusensis range between 3.90 and 8.85 (Appendix 1). The least variable tooth was the first lower molar (3.90-7.09). The most variable molar was the fourth upper molar (6.96-8.85). The CVs are generally consistent with the range of CVs reported by Travouillon et al. (2014) for a mixed sex population of the macropodids Thylogale stigmatica (3.54-12.5) and T. thetis (4.98-11.16).

Bivariate Plots
Species included in size comparisons were Bulungamaya delicata, Ganguroo bilamina, and Wabularoo naughtoni. Ganguroo bites was also included in plots of lower molar size (only lowers are known for this species). Bivariate plots of upper tooth dimensions are presented in Figure 7. There is some overlap between premolar length in C. hortusensis and C. bulwidarri, but the premolar sample of C. bulwidarri (two specimens) is not sufficient to draw conclusions. Furthermore, the species cannot be distinguished by molar size. The upper molars and premolar of both species of Cookeroo are significantly longer and wider than those of B. delicata and species of Ganguroo. Cookeroo hortusensis and C. bulwidarri differ from W. naughtoni in having a narrower premolar but have similarly sized molars.
The lower dentition shows a similar pattern (Fig. 8) in that both species of Cookeroo are distinguished from B. delicata and species of Ganguroo in having longer and wider molars and premolar. Cookeroo hortusensis and C. bulwidarri cannot be compared using measurements of lower molars because only the first molar of C. bulwidarri is known. Wabularoo naughtoni generally has wider lower premolars and longer molars than Cookeroo species.

Principal Component Analysis
In the PCA of upper tooth dimensions (Appendix 2), Component 1 accounts for 68.36% of variance, whereas Component 2 accounts for 12.29% (Fig. 9A). The length and width of P3 appear to be the most useful measurements for distinguishing between the species included in the analysis. Cookeroo specimens fall outside the range of species of Ganguroo and B. delicata. Larger specimens of Cookeroo fall in a similar size range to those of W. naughtoni. In the PCA of lower tooth dimensions (Appendix 3), Components 1 and 2 account for 68.21% and 12.35% of variance, respectively (Fig. 9). The length and width of the p3 are the most useful measurements to distinguish between species. Cookeroo specimens fall outside the range of other species with the exception of W. naughtoni, which is similar in size to larger Cookeroo specimens.

Phylogenetic Analysis
Parsimony analysis of the unconstrained matrix resulted in 540 most parsimonious trees (tree length D 583, consistency index D 0.3533, retention index D 0.7523), with a strict consensus presented in Figure 10. The outgroup species, Hypsiprymnodontidae (including the propleopine clade), and two balbarid species, Nambaroo saltavus and Nambaroo tarrinyeri, form a polytomy at the base of the tree. The outgroup species cluster together to the exclusion of all other species in bootstrap analysis (bootstrap D 77). However, this is not reflected in the consensus tree. Other FIGURE 6. Juvenile dentary of Cookeroo hortusensis (paratype QM F57637) in A, occlusal view (stereo); B, buccal view; and C, lingual view. Abbreviations: i1, lower first incisor; p2, lower second premolar; dP3, deciduous lower third premolar; m1-4, lower first to fourth molars; maf, masseteric fossa; mf, mental foramen; p3, lower third premolar; s, symphysis.   balbarid species are nested within Macropodidae with no support. Sthenurinae forms a distinct clade (bootstrap D 54). However, it is positioned at the base of the clade that includes macropodines and lagostrophines in a polytomy with Cookeroo and Wabularoo. Species of Wabularoo do not cluster together. Wabularoo prideauxi is positioned at the base of the sthenurine clade, although with no bootstrap support. Lagostrophinae is nested within Macropodinae; however, the positions of all subfamilies within Macropodidae are unresolved in bootstrap analysis.
Our analysis of the data with three characters constrained as irreversible resulted in 450 most parsimonious trees (tree length D 587, consistency index D 0.3509, retention index D 0.7945), with a strict consensus presented in Figure 11. Hypsiprymnodontidae is unresolved at the base of the macropodiform clade. The propleopine clade is, however, recovered with low bootstrap values (bootstrap D >50). Propleopinae is recovered as a sister taxon to Balbaridae. However, this relationship is not robustly supported. The sister group relationship between Propleopinae and Balbaridae is consistent with a number of previous phylogenetic analyses (Kear and Pledge, 2007;Travouillon et al., 2014). Nambaroo is polyphyletic. Macropodidae is recovered as a clade distinct from Hypsiprymnodontidae and Balbaridae (bootstrap D 86). However, it is unclear if that clade represents a sister group to macropodids.
Within Macropodidae, Potoroidae (in so far as it includes living plus fossil species) is paraphyletic. The modern potoroines Potorous tridactylus, Bettongia penicillata, and Aepyprymnus rufescens cluster as the sister taxon to other macropodids plus Bettongia moyesi and Wakiewakie lawsoni. Fossil potoroines (Bettongia moyesi and Wakiewakie lawsoni) are recognized as the first branch of the clade that includes all other macropodids. Bettongia moyesi does not have a monophyletic relationship with extant potoroines such as B. penicillata. Interpretation of these results should, however, be treated cautiously because extant potoroines are recovered as a clade to the exclusion of fossil potoroines and other macropodids with no bootstrap support in this study.
Ngamaroo archeri and Purtia mosaicus are recovered near the base of the clade that includes macropodines, lagostrophines, and sthenurines. This is consistent with the phylogenetic analysis by Kear and Pledge (2007). Prideaux and Warburton (2010) did not include Purtia mosaicus and Wakiewakie lawsoni but did find support for the placement of N. archeri and Nowidgee matrix near the base of the macropodid clade. Specimens of Nowidgee matrix have since been reassigned to Bulungamaya delicata , which is also recovered here as a relatively plesiomorphic macropodid. The clustering of Largostrophus fasciatus and Troposodon minor as a distinct clade is also consistent with Prideaux and Warburton (2010).
Lagostrophines are recovered as a monophyletic group (bootstrap D 91); however, their position within Macropodidae is unresolved. Wanburoo hilarus and Hadronomas puckridgi do not fall within the sthenurine clade and are unresolved within a larger clade including lagostrophines, sthenurines, and macropodines. Bootstrap analysis does, however, find support (>50) for clustering of W. hilarus, H. puckridgi, and other sthenurines as a distinct clade, although this is not represented in the consensus tree. Monophyly of sthenurines Sthenurus andersoni, Simosthenurus occidentalis, and Procoptodon goliah is unequivocal (bootstrap D 100). Previous studies resolve W. hilarus and H. puckridgi as plesiomorphic members of Sthenurinae (e.g., Prideaux and Warburton, 2010). The monophyly of Sthenurinae including W. hilarus and H. puckridgi is reinforced by high bootstrap values in this study. However, the relationship is not represented in the strict consensus tree. The unresolved relationships of sthenurines and lagostrophines within Macropodidae may have resulted from the inclusion of species of Cookeroo, causing the sthenurine clade to collapse. These species appear to share more features with early sthenurines (e.g., W. hilarus and H. puckridgi) and macropodines (e.g., D. fossilis) than other basal macropodids such as species of Ganguroo. However, it should be noted that strong support for the sthenurine clade containing species of Sthenurus, Simosthenurus, and Procoptodon was recovered in bootstrap analysis. The inclusion of additional macropodine species in future analyses may assist in resolving the phylogenetic position of Cookeroo, sthenurines, lagostrophines, and macropodines.
Species of Ganguroo and Wabularoo are unresolved within the clade that includes lagostrophines, sthenurines, and macropodines. Species of Cookeroo form a clade with support from bootstrap analysis (bootstrap D >50) within that larger clade. However, the position of Cookeroo in relation to lagostrophines, sthenurines, and macropodines is unresolved. Species of Cookeroo share several features with other members of the clade that includes lagostrophines, sthenurines, and macropodines and appear to be more derived basal macropodids than Ganguroo and Wabularoo. However, the relative positions of Ganguroo, Wabularoo, and Cookeroo are unresolved in this analysis. The placement of Ganguroo and Wabularoo near the base of the clade containing sthenurines and macropodines is consistent with previous studies (Kear and Pledge, 2007;. The analysis by Prideaux and Warburton (2010), however, placed lagostrophines as a sister clade to Ganguroo and all other sthenurines and macropodines. Macropodinae is recovered as a monophyletic group, but with no support from bootstrap analysis. Some clades within the macropodines are resolved, such as Macropus (bootstrap D 88) and tree kangaroos (boostrap D 99).
A Templeton test comparing the constrained and unconstrained consensus trees found a statistically significant difference (P D 0.0006, n D 75) between the two trees ( Table 4). The statistically significant difference between the analyses can be interpreted as the result of a large number of homoplastic characters within Macropodiformes, particularly dental characters, which was previously acknowledged by Black et al. (2014). Low support for the position of some clades in the consensus tree may therefore be the result of reliance upon these homoplastic characters. Although this could be supplemented by other data sources such as postcranial material, for many fossil species there are not sufficient postcranial remains to adequately address the problem. The polyphyly of the balbarid clade and the position of the lagostrophine clade in the unconstrained tree are resolved by constraining three dental characters as irreversible. The overall topology of the constrained phylogenetic tree is also consistent with that of previous analyses (e.g., Black et al., 2014) and likewise derived low support for most internal nodes.

DISCUSSION
Cookeroo species are larger than species to which they are most similar in terms of morphological character states, such as Ganguroo bilamina. Although most similar in size to Wabularoo naughtoni, species of Cookeroo can be distinguished by their narrower premolar, a lingual cingulum on P3 that tapers anteriorly, a shallow buccinator sulcus, and a less robust dentary. Coefficients of variation for tooth measurements in C. hortusensis generally fall within the range of values for mixed-sex populations found in other macropodid species (Bartholomai, 1971; Travouillon et al., 2014). Cookeroo species can be distinguished from Ganguroo species by their size and in having a masseteric canal that extends to below m1, a slight postmetacristid and postentocristid on m1, and StD on M1.
Cookeroo bulwidarri from White Hunter Site in Faunal Zone A is similar in terms of molar and premolar sizes to C. hortusensis, which is known from many sites representing Faunal Zone B. Cookeroo bulwidarri differs from C. hortusensis in having eight cusps on p3 as opposed to nine in C. hortusensis, and in having a distinct connection between the postparacrista and premetacrista, forming a centrocrista. A better-developed crest on the lateral surface of the jugal and a taller ventral crest on the basioccipital are also present in C. hortusensis. The upper molar row of C. hortusensis is also bowed laterally in the occlusal plane, whereas in C. bulwidarri the upper molar row is relatively straighter. These morphological differences suggest that specimens from Faunal Zones A and B represent separate species of Cookeroo. The loss of a distinct connection between the postparacrista and premetacrista forming a centrocrista in specimens from Faunal Zone B suggests that C. bulwidarri may well be ancestral to C. hortusensis. There are no autapomorphies in C. bulwidarri that would prohibit this possibility.
Some features of Cookeroo are plesiomorphic for macropodoids. For example, the smooth transition between the zygoma and cheek region and poorly developed masseteric process in Cookeroo species are consistent with more basal macropodiforms such as hypsiprymnodontids and potoroines (Kear and Cooke, 2001). However, Cookeroo species also share a number of derived features with macropodines, such as the parietal and alisphenoid contact on the lateral wall of the cranium and an alisphenoid contribution to the posterior wall of the secondary foramen ovale (Kear and Cooke, 2001). The postglenoid foramen is similar in size to that of macropodines such as Thylogale thetis. The basioccipital crest is well developed and similar to the basioccipital, as scored in previous studies for macropodines such as Macropus and Petrogale (e.g., Kear and Pledge, 2007;Prideaux and Warburton, 2010). Further, m4 is not significantly smaller compared with m3, a condition consistently found in macropodids but not in potoroids (Kear and Cooke, 2001). This suggests that Cookeroo species may be more closely related to derived macropodids than relatively plesiomorphic macropodiforms.
Cookeroo species are similar to other Miocene bilophodont species regarded to be macropodids, such as Wanburoo hilarus (Cooke, 1999), Hadronomas puckridgi (Woodburne, 1967), and Dorcopsoides fossilis (Woodburne, 1967), in having a distinct postmetaconule and postmetacrista, a shallow symphyseal plate, a reduced or absent postentocristid and posthypocristid, a smooth transition between the zygoma and cheek, and p3 with a posterolingual cusp or ridge. Species of Cookeroo and D. fossilis differ from H. puckridgi (Woodburne, 1967) and W. hilarus (Cooke, 1999) in having small ectoglenoid and postglenoid processes, a shallow buccinator sulcus, and a masseteric canal that extends below the anterior cheekteeth. However, like H. puckridgi and W. hilarus, species of Cookeroo exhibit a sinuous lower incisor, neometaconule, and postlink (Woodburne, 1967;Cooke, 1999;Kear and Pledge, 2007). These features support a close relationship between Cookeroo species and other derived macropodids from the early Miocene. However, species of Cookeroo differ in having a narrow and anteriorly tapered P3, unlike the condition seen in D. fossilis where the lingual cingulum is very fine (Woodburne, 1967;Kear and Pledge, 2007). Further, the lingual cingulum of H. puckridgi and W. hilarus is broader compared with that structure in species of Cookeroo.
The genus Cookeroo represents two additional macropodid species from the Oligo-Miocene at Riversleigh. The description of these species may have implications for previous interpretations of species richness of macropodiforms such as that by Cooke (1997a). Since Cooke (1997a), multiple other macropodiform species have also been described Bates et al., 2014;Travouillon et al., 2014;Cooke et al., 2015;Travouillon et al., 2015), which may affect the interpretation of trends in species diversity within macropodiforms. Understanding of species richness and trends in species diversity of macropodiforms throughout the Oligo-Miocene is a crucial first step towards understanding the evolution of early macropodiforms and their response to significant environmental changes during the Oligo-Miocene. APPENDIX 1. Unviariate statistics for teeth of Cookeroo hortusensis. Abbreviations: AW D anterior width; L D tooth length; m D lower molar; M D upper molar; Max D maximum measurement; Min D minimum measurement; N D number of specimens; p D lower premolar; P D upper premolar; PW D posterior width. APPENDIX 2. Summary statistics of principal component analysis of upper tooth measurements for Cookeroo bulwidarri, C. hortusensis, Bulungamaya delicata, Ganguroo bilaminar, and Wabularoo naughtoni (rounded to two decimal places).