THE PILKIPILDRIDAE, A NEW FAMILY ANDFOUR NEW SPECIES OF ?PETAUROID POSSUMS

(MARSUPIALIA: PHALANGERIDA)FROM THE AUSTRALIAN MIOCENE

MICHAEL ARCHER1, RICHARD H. TEDFORD2 and THOMAS H. RICH3

Collections made from the Etadunna Formation at Lake Palankarinna, Etadunna Station in northern South Australia fromthe Namba Formation at Lake Pinpa, Lake Yanda and Lake Tarkarooloo, Frome Downs Station, northern South Australia andfrom the newly discovered Miocene freshwater limestones of Riversleigh Station, northwestern Queensland, have revealedmaterial referable to four previously unknown small Miocene possums. These cannot be referred to a known family-level groupof diprotodont marsupials and are described herein as Pilkipildra handi (Namba Formation), Pilkipildra taylori (EtadunnaFormation), Djilgaringa gil/espiei (Riversleigh) and Djilgaringa thompsoni (Namba Formation). All four of these possums areplaced in a new family of marsupials, the Pilkipildridae. All appear to have been omnivores but some were probably morespecialised as seed or hard fruit-eaters. They appear to have been Phalanger-like and/or Petaurus-Iike in diet. Structurally,pilkipildrids, most resemble petaurids but they also demonstrate resemblances to phalangerids, miralinids and ektopodontids.Although the precise phylogenetic interrelationships of the pilkipildrids within Phalangerida are uncertain, the tentativeconclusion is that Pilkipildridae is the sister-group of Petauridae.

Key Words: Pilkipildridae; Pilkipildra handi; Pilkipildra taylori; Djilgaringa gil/espiei; Djilgaringa thompsoni; Phalangerida;Petauroidea; Marsupialia; Miocene.

Pages 607-27 in POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION ed by M. Archer. Surrey Beatty & Sons and the

Royal Zoological Society of New South Wales: Sydney, 1987.

demonstrate the distinction of these possumsfrom all others.

Finally, a dentary fragment with M2 was foundby Rich and others in 1984 at Lake Yanda(Namba Formation; Yanda Local Fauna), SouthAustralia (Fig. 1).

The purpose of this paper is to describe andphylogenetically analyse this new material.Although these new possums are rare in all faunasin which they occur, they provide an opportunityto relate the local faunas of northern and centralAustralia and to test current concepts of their rela-tive ages (e.g., Woodburne et at. 1985; Archer andFlannery 1985).

Dental terminology used in this paper followsArcher (1978, 1984), Tedford and Woodburne(1987) and Woodburneet al. (1987). Nomenclatureof higher taxa follows Aplin and Archer ( 1987; thisVolume). Stratigraphic and faunal names forSouth Australia follow Woodburne et at. (1985).

INTRODUCTION

IN 1972, Archer, Woodburne and others dis-covered an isolated M2 of a very unusual possumfrom the Middle Miocene Tedford Locality(Etadunna Formation; Ditjimanka Local Fauna)at Lake Palankarinna, South Australia (Fig. 1).

In 1973, Tedford and others discovered adentary of a clearly similar form from the MiddleMiocene sediments at Lake Pinpa (NambaFormation; Pinpa Fauna, Billeroo Creek Site),South Australia (Fig. 1). Another South Aust-ralian isolated MJ. referable to this Pinpa formwas collected by Rich and others in 1976 at LakeTarkarooloo (Namba Formation; TarkaroolooLocal Fauna) (Fig. 1).

Early reference to the Tedford locality formwas given as a "Burramyid similar to species ofCercartetus" (e.g., Rich et at. 1982) and to thePinpa taxon as a "large Cercartetus-Iikeburramyid" (Tedford et at. 1977).

Discovery by Archer and others in 1984 at theMiocene Riversleigh fossil sites (Fig. I, unnamedformation; Last Minute Site, Last Minute LocalFauna) of a dentary of yet another speciessimilar to the central Australian forms wasfollowed by the finding at Riversleigh (unnamedformation; Gag Site, Dwornamor Local Fauna)of the first upper molars referable to one of thesepeculiar possums. These teeth further serve to

SYSTEMATICS

Subclass Marsupialia (Illiger, 1811)Superorder Diprotodontia Owen, 1866Suborder Phalangerida Aplin and Archer, 1987Superfamily Petauroidea (Gill, 1872)

Family Pilkipildridae newType gen1L5: PilkiPildra is selected as the type

genus of the family.

lSchool of Zoology, University of New South Wales, PO. Box I, Kensington, New South Wales, Australia 2033.'Department of Vertebrate Palaeontology , Atnerican Museum of Natural History, Central Park West at 79th Street, New York, New York, USA.sMuseum of Victoria, Russell Street, Melbourne, Victoria, Australia 3000.

608 POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION

Fig. I. Localities and local faunas from which pilkipildrids have been recovered: A, Frome Downs Station, Pinpa and TarkaroolooLocal Faunas, Pilkipildra handi gen. and sp. n.; B, Etadunna Station, Ditjimanka Local Fauna, Pilkipildra taylori gen. andsp. n.; C, Riversleigh Station, Last Minute and Dwornamor Local Faunas, Djilgaringa gillespiei gen. and sp. n.; D, FromeDowns Station, Yanda Local Fauna, Djilgaringa thompsoni gen. and sp. n. (Line drawings by J .Taylor).

ARCHE~, TEDFORD, and RICH: THE PILKIPILDRIDAE 609

and deep dentary , a paraconid on M2' prominentmetastylids, extensively crenulated crownsurfaces and a transversely compressed trigonid

onM2.

They differ from tarsipedids in all aspects ofgross dental and dentary morphology .

They differ from miralinids (see Woodburne,Pledge and Archer 1987) in having transverserows of small cuspules on the upper molars,reduced paracones and metacones, a short anddeep dentary, steeply inclined Ii' wide P 3' non-linear M2 trigonid and large metastylids.

They differ from macropodoids in not havinga masseteric foramen or a masseteric insertioninto the body of the dentary and in having a shortand deep dentary , transverse rows of cuspules onthe upper molars, low-crowned bunodont molarsposterior to M2 and distinct trigonid basins on theanterior lower molars.

Djilgaringa new

Genotypic species: Djilgaringa gillespiei is the typespecies of the genus.

Additional species: Djilgaringa thompsoni.

Distribution: Last Minute Local Fauna and GagLocal Fauna, unnamed freshwater limestones,Riversleigh Station, northwestern Queensland;Yanda Local Fauna, Namba Formation, LakeYanda, Frome Downs Station, South Australia.

Age: Medial Miocene, ca 14-12 ma(Woodburne et at. 1985 and Archer and Flannery

1985).

Etymology and gender: Djilga means "baby" andGaringa means "possum" in the Wanji languageas spoken by the late I vy George of RiversleighStation northwestern Queensland. The name isin reference to the small size and presumed shortfaces of these possums. The genus is here givenfeminine gender.

Generic diagnosis: Species of Djilgaringa differfrom the species of other pilkipildrid genera inthe following combination of features: largersectorial and more buccally deflected p ; P lacksposterobuccal and posterolingual cinguia; the M2trigonid is higher; the M2 is proportionatelyshorter and wider; the crown base of M2 is pro-portionately more swollen; the talonid l>asin onM~ is less well-developed; termination of thecnstid obliqua of M2 occurs at the posterobuccalbase of the trigonid; the M2 has a poorly-developed paracristid; and there is no notch inthe anterior cingulum to accommodate theposterior edge of P '!.

Additional genera: Djilgaringa.

Range: The four new and only known speciesin the family range stratigraphically, geo-graphically and temporally as follows: Pilkipildrahandi, Namba Formation, Frome Downs Station,South Australia, medial Miocene; Djilgaringathompsoni, Namba Formation, Frome DownsStation, South Australia, medial Miocene; Djil-garinga gillespiei, unnamed freshwater lime-stones, Riversleigh Station, northwesternQueensland, medial Miocene; PilkiPildra taylori,Etadunna Formation, Etadunna Station, SouthAustralia, medial Miocene.

Etymology ofthefamily name: Pilki is a Dieri wordmeaning "different" and pildra is a Dieri wordmeaning "possum" (Reuther 1901; as translatedby Scherer and published in 1981). The DieriTribe occupied the Tirari Desert in which occursLake Palankarinna where the first pilkipildridfossil was discovered in 1972.

Family diagnosis: Pilkipildrids are phalangeri-dans that differ from ektopodontids in havingdistinguishable protocones and metaconules, alarge buccally deflected P , a very steep molargradient, a distinct trigoni~ and cristid obliqua onM2' and in lack.ing the pronounced transverserows of cuspules that are the dominant featuresof the lower molars of ektopodontids.

Pilkipildrids differ from phalangerids inhaving short and deep dentaries, a steeplyinclined Ii' reduced and cuspule-like paraconesand metacones, a very steep molar gradient,short and relatively wide molars, pronouncedmetastylids, poorly-developed transverse ridgeson the molars, poorly-indented medial buccaland lingual flanks of the molars, reduced enechelon crest relationships, a distinct protostylidon M2' an extremely reduced M2 metaconid, asteeply inclined M2 trigonid, a simple or reducedcristid obliqua on all molars with no "kink" andvery low crown relief.

They differ from petaurids in having a verylarge and sectorial P 3' transverse rows of cuspuleson the upper molars, extremely reduced butdistinct paracones and metacones and pro-nounced metastylids.

They differ from pseudocheirids in their lackof distinct selenodonty, steep molar gradient,reduction of en echelon crest relationships, reduc-tion of the paracone and metacone and trans-verse rows of cuspules on the upper molars.

They differ from burramyids in their short anddeep dentary , extensively crenulated molars,transverse rows of cuspules on the upper molarsand trigonid basins on the anterior lower molars.

They differ from acrobatids in having M5' alarge sectorial and buccally oriented P 3' uppermolars with transverse rows of cuspules, a short

POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION610

Djilgaringa gillespiei n. sp.

(Figs 2- 7)Holotype: Queensland Museum fossil collec-

tions F13028, a broken right dentary with P 3' M2-5and alveoli for II' P 2' ?P I and a partial alveolus for?12.: !his specimen was collected by Archer et al.in May, 1984.

Type locality and age of the holotype: The holotypewas collected at Last Minute Site, from anunnamed freshwater limestone deposit (possiblya lateral equivalent of the Carl Creek Limestone )on Riversleigh Station, northwestern Queens-land. The deposit is interpreted here to beapproximately medi(1.l Miocene in age.

Paratype and its locality: A single isolated upperright molar (Queensland Museum fossil collec-tion No. FI4371), possibly an M3, appears torepresent this taxon. It was retrieved byH. Godthelp from blocks collected by Archer andothers in May, 1983, from Gag Site, anotherlocality in the same unnamed but presumablymiddle Miocene freshwater limestone deposits ofRiversleigh Station.

Etymology: This species is named in honour ofAnna Gillespie, one of the principal AustralianResearch Grants Committee preparators whoworked on the Riversleigh Project at the Uni-versity of New South Wales unti11986. She wasalso responsible for the discovery duringpreparation of many of the new Riversleigh taxa.

Species diagnosis: This species differs from Djil-garinga thompsoni (new; see below) in that the P 3was probably more buccally out-turned, the M2(the only tooth of D. thompsoni known) is pro-portionately longer and narrower, the talonid isproportionately shorter, there is a protostylid,the anteroventral face of the trigonid is lesssteeply inclined, the metaconid is better-developed, the cristid obliqua has distinctanterior and posterior parts and the crenulationsare coarser .

Description: The dentary is deep and short. Thebuccal side is badly damaged and reveals littleinformation. The lingual side preserves apart ofthe symphysis which was clearly very steep,possibly rising at about 60° to the horizontalramus. None of the dental foramina are pre-served other than a possible small mentalforamen on the buccal side below M2. A branch ofthe dental canal is breached at the anterior end,immediately adjacent to the steeply risingalveolus for II. The main portion of the dentalcanal terminates in a large trabecular septumwhich forms the posterior wall of the II alveolus.There is only a very slight ectalveolar projectionof the dentary bey<:>nd M5 along the tooth row.

The II alveolus suggests that this tooth wasvery robust. The alveolus rises at an angle ofabout 45° to the horizontal ramus and thus

suggests a semi-erect posture for this tooth, aposture in keeping with the apparently very shortand deep dentary .

There are three alveoli or alveolar remnantsbetween p 3 and the alveolus for Ii. We suspectthat each represents a distinct tooth becausethere are no suggestions from the septal areasbetween the alveoli that a crown bridged any ofthe septa. The anterior alveolar remnant is in aposition suggestive of 12, such as occurs in manyphalaneridans (e.g., burramyids, phalangerids,petaurids etc.). The next alveolus in sequenceprobably represented a single rooted PI althoughthe homology of teeth in this position even inliving phalangeridans is in doubt (Archer 1984).The alveolus anterior to p 3 probably representedp 2. The relationships of tfie alveoli for this toothand p 3 indicate that p 2 occurred at the base of theanterobuccal corner of the p 3 crown. There is nowear facet in this position on the base of the p 3crown thus suggesting that p 2 was a very smalltooth that functioned at the level of the gum.

The p 3 is a massive sectorial tooth, beingmatched in height only by the trigonid ofM2. Thewhole crown is obliquely set with respect to themolar row, much as in the manner of the p 3 ofsome potoroids and burramyids. There was onerelatively narrow cylindrical anterior root andeither one very wide posterior root or twoposterior roots of uncertain width. The base ofthe crown is very swollen posteriorly and narrowsanteriorly. At a position level with the junction ofthe anterior and posterior roots, the base of thecrown is slightly indented on the buccal side. Thetall sectorial ridge is obliquely set within thecrown such that it forms approximately a 45°angle with respect to the molar row. There arethree cuspules along the crest corresponding tothe three prominent vertical ridges on the buccalside and three of the four prominent ridges onthe lingual side. The fourth ridge on the lingualside occurs between the first and the second. Ashort basally situated ridgelet occurs on thebuccal side anterior to the base of the secondprominent ridge. The anterior end of the crownis smoothly curved and anteriorly concave. Theposterior end is a descending extension of theocclusal cutting crest. It descends abruptly awayfrom the main cutting crest, at an angle of about45° to the posterior end of the crown. Theposterior half of this descending crest underliesthe comparably inclined trigonid of M2. Theanterior height of the crown markedly exceedsthe posterior height. There are no basal cingulaor accessory basal cusps.

The M has a high trigonid and broadertalonid. nere is a short but wide anterior rootand along and wider posterior rooL Theprotoconid is displaced lingually such that itoccurs lingual to the midpoint of the transverse

612 POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION

Fig. 3. Djilgaringa gillespiei, holotype, QM F13028, from Last Minute Site, Riversleigh Station, north-western Queensland. Stereophotographs of right dentary with alveoli for P 2 (and ?P J andP 3-M5 (anterior end up). (Photographs by H. Godthelp).

width of the trigonid. The paraconid isrepresented by the anterior end of the anteriorlyoriented preparacrista. The metaconid appearsto be a slight bump along the crest extendingposterolingually from the protoconid. Of thesethree cusps, the protoconid is just taller than themetaconid which is about the same height as theparaconid. There appears to be a protostylidadpressed to the buccal flank of the protoconidbelow the level of the principal trigonid cusps.The paracristid appears to be continuous withthe lingual end of a short, somewhat poorly-defined anterior cingulum. The cristid beyondthe metaconid and metacristid is a postmeta-cristid. It is more steeply inclined than the meta-cristid and connects the metaconid to the lingual

end of the midvalley. The postmetacristid isserrate, reflecting the pronounced enamelcrenulations that occur on its posterobuccalflank. The protostylid has two minor cristids, oneof which descends the trigonid flank in ananterior direction, to intersect the anteriorcingulum, and the other of which abruptlydescends the flank of the protostylid to mergewith the flank of the trigonid buccal to the cristidobliqua. There is also a distinct vertical cristidthat descends the buccal flank of the protostylidto intersect the buccal cingulum of the trigonid.The poorly-defined anterobuccal basal cingulumpasses around the buccal side of the trigonid butextends no further posteriorly than the mid-valley. The whole trigonid rise steeply and rests

ARCHER. TEDFORD, and RICH: THE PILKIPILDRIDAE

against the posterior flank of p 3. The trigonidcrest is continuous with, but at an angle to, thecutting crest of p 3. The two almost certainlyfunction as a unit to oppose p3. The talonid ofM2has two cusps, a buccal hypoconid and a lingualentoconid. The whole of the talonid basin andposterior flank of the trigonid are stronglycrenulated. These crenulations rise towards andinterrupt the cristid obliqua, postmetacristid andpre-entocristid. The crenulations all converge toa point in the middle of the talonidt>asin, midwayalong the transverse midvalley. The serratecristid obliqua extends from the hypoconid to thebase of the trigonid near its buccal flank,posterior to the protostylid. A prominent trans-verse fissure demarcates its anterior end.Another prominent transverse fissure occursmidway along its length. The cristid obliqua sodivided has a descending posterior segment andan ascending anterior segment. There is noactual transverse hypocristid but the talonidbasin and its crenulations are raised along a trans-verse line between the entoconid and hypoconid.With wear, this rise could have functioned as aserrated transverse ridge. There is a semilunarposterior crest linking the posthypocristid andpostentocristid around the posterior end of thetalonid basin. There may be a small pre-entostylidmidway along the pre-entocristid. The base ofthe buccal side of the crown is topographicallymuch lower than the base of the lingual side. Thewhole of the tall,' almost vertical buccal face of thetalonid and parts of the buccal side of thetrigonid exhibit what appear to be rudimenta~thegotic facets. This suggests that the M2 and pwere vertically honed against these surfaces. Thelingual side of the crown does not show this sortof wear. Thegotic facets are also developed in thetransverse groove of the poseterior cingulum, onthe posterobuccal flank of the entoconid, on thecrenulations ascending the posterior flank of thetrigonid and on the posterior flank of thehypoconid.

The M3 differs from the M2 as follows. Most ofthe differences pertain to the markedly differenttrigonid. The trigonid and talonid are approxi-mately the same height and width. The anteriorroot is longer than it is on M2. The whole crownis lower. The protoconid is anterobuccallysituated. There is no distinct paraconid. A pre-protocristid connects the protoconid to the frontof the tooth and merges with or becomes theanterior cingulum. Two distinct crests connect tothe large lingual metaconid, a postmetacristidand a premetacristid. The latter is continuouswith the anterior cingulum. There is a low trans-verse ridge raised among the crenulations thatlinks the protoconid to the metaconid. This ridgeis dorsally concave and highest where it contactsthe metaconid. With wear this ridge couldbecome a functional serrated transverse lophid.Of the trigonid cusps, the metaconid is by far thetallest, the smaller protoconid being subequal inheight to the talonid cusps. There is a smalltrigonid basin defined by the transverse ridgebetween the protoconid and metaconid and bythe anterior cingulum. This basin is crenulated asis the talonid basin. The cristid obliqua isabruptly truncated at the midvalley by a pro-nounced depression at the buccal end of the mid-valley. A short postprotocristid which extendsposterolingually from the protoconid alsoterminates at this depression. On the lingual sideof the crown, there is a large metastylid blockingthe lingual end of the transverse median valley. Ithas short, inclined anterior and posterior connec-tions to the postmetacristid and pre-entocristid.The cristid obliqua posterior to the midvalleybuccal depression is divided into two sections (asit is in M2) the anterior one of which is short andsurmounts a small discrete swelling (thehomologous swelling in M2 is relatively muchlarger). The posterior segment of the cristidobliqua is anteroposteriorly oriented (whereas itis more oblique in M2). There are no buccal

Fig. 4. Djilgaringa gillespiei, holotype, QM F13028, from Last Minute Site, Riversleigh Station, northwestern Queensland.Composite scanning electron microscope photographs of lingual obliqua view of alveoli for ?P I and P 2 (or only for P 2)' andP 3-M5. (Photographs by H. Godthelp).

POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTIOr.614

occlusal surface of ~5 faces dorsolingually whilethe same surface of M2 faces dorsobuccally.

The only known upper molar (FI4371)appears to be an M3 on the basis of size andocclusal patterns. There are only two distinctprincipal cusps, the protocone and metaconule.Assuming that there is a paracone and ametacone, the position of these cusps is in somedoubt. Tentatively, we have assumed that theyare not on the well-developed buccal margin ofthe crown. This is because there are two smallcusps lingual to the buccal crest in positionsanalogous to the paracones and metacones ofwynyardiids and other vombatimorphians andsome plesiomorphic petauroids (e.g., pseudo-cheirids). The protocone is the largest cusp,although the protocone, metaconule and buccalcrests of the crown are subequal in height. Thebuccal margin of the crown is formed by two adja-cent occlusally convex crests. We think that thislarge crest represents hyperdevelopment of abuccal cingulum. Between the peak of theposterobuccal crest and the protocone, occurthree cuspules aligned in a transverse row.Similarly, between the peak of the posterobuccalcrest and the metaconule, there are also threealigned cuspules. Of the anterior row of cuspules,the most buccal is associated with minor cristaethat appear to be homologues of a preparacristaand a postparacrista. For this reason, weinterpret this cuspule to be the paracone. Thereare no comparable cristae associated with themost buccal cuspule in the posterior row. Theshort preparacrista extends anterobuccally tointersect the anterobuccal crest at the antero-buccal corner of the tooth. The longer postpara-crista extends posteriorly from the paracone andthen swings posterobuccally to intersect the

cingula around the trigonid or on the antero-buccal face of the trigonid.

The M4 resembles M3 except as follows. Thecrown is lower in height, narrower and shorter .The talonid is distinctly narrower than thetrigonid. The meta<.onid, although still thehighest cusp of the tooth, is proportionatelyshorter in height than that cusp on M .Thetrigonid is only just taller than the talonia. Thetransverse trigonid ridge is less well-developedand the trigonid basin less distinct. There is virtu-ally no development of the transverse talonidridge. Instead, a distinct sinuous but basicallylongitudinal median fissure divides the whole ofthe talonid into lingual and buccal sections. Theanterior segment of the cristid obliqua is veryshort and in fact little more than a slightlyenlarged buccal end of one of the talonid basincrenulations.

The M5 is similar to M4 except as follows. Thecrown is smaller in all dimensions. The talonid ismuch narrower than the trigonid. Themetaconid and protoconid are subequal in heightand just taller than the talonid cusps. Theentoconid is extremely reduced and theposterolingual corner of the tooth compressed.There is no differentiation of the occlusal surfaceinto a trigonid or talonid basin, the crenulationsforming a single pattern over the whole surface.The metastylid is only just distinguished as aslight prominence on the postmetacristid. Thereis no distinguishable cristid obliqua. Thehypoconid is linked on the buccal side of thetooth to the protoconid via the buccal ends oftransverse crenulations. Much of the occlusalsurface of the crown exhibits small horizontalthegotic facets that have sharpened the edges ofthe crenulations and crown margins. There isalso a curious elongate depression worn in themiddle of the rear half of the crown. Possibly it isthe result of occlusion with the buccal edge ofM5.However, no analogous facet occurs on any of theother lower molars.

Meristic gradients in the lower tooth row are asfollows. Apart from the protoconid, all principalcusps decrease in size posteriorly. Theprotoconid of M2 is approximated to the tallmetaconid but the protoconids of M3 to M5 alsodecrease in size posteriorly. The metastylidsincrease in size from M2 to M3 and then decreaseto M5. The molars decrease in length, width andcrown height from MI to M4.: .The whole toothrow is gently arcuate with M2 and M5 beingtopographically the highest posItioned teeth andM3 the lowest. This results in an occlusallyconcave tooth row when viewed from a lateralpoint of view. Accordingly it may be presumedthat the curvature of the upper tooth row was inthe opposite direction. There is also a sinuoustwist in the crown orientations such that the

Fig. 5. Djilgaringa gillespiei. holotype, QM F13028, from LastMinute Site, Riversleigh Station, northwesternQueensland. SEM detail of occlusal view of alveoli for?P I and P 2 (or only for P 2)' and P 3-M2 (plus part of

trigonid of M3). (Photograph by H. Godthelp).

ARCHER, TEDFORD, and RICH: HE PILKIPILDRIDAE 615

Fig. 6. Djilgaringa gillespiei, paratype, QM Fl4371 , RM3, from Gag Site, Riversleigh Station, northwesternQueensland. Stereophotographs of occlusal view. (Photographs by H. Godthelp).

posterobuccal crest. A preprotocrista extendsanteriorly and then swings buccally to becomethe anterior cingulum. This cingulum is con-tinuous around the anterobuccal end of thecrown with the hyperdeveloped buccalcingulum. The postprotocrista extendsposteriorly then swings abruptly buccally andterminates anterior to the premetaconulecrista.The premetaconulecrista extends anteriorly andthen swings anterobuccally to terminate oppositethe postprotocrista. The postmetaconulecristaextends posteriorly and then swings buccally tobecome the posterior cingulum. This cingulum iscontinuous around the posterobuccal cornerwith the buccal cingulum. A deep obliquelyoriented fissure interrupts the crown from nearits lingual edge to pass between the pre-metaconulecrista and postprotocrista therebydetermining the boundary between these twocrests. A similar transverse fissure extends infrom the buccal margin of the crown to divide thebuccal cingulum into its convex crests. Both thelingual and buccal fissures lead to the centre ofthe crown, as do the crenulations covering mostof the occlusal surface of the crown. There is thusa "point-focus" for the crenulations as there areon the lower molars. A small transverse crest linksthe protocone to the most lingual of the threecuspules in the anterior transverse row. Betweenthis cuspule and the paracone is the third andlonger isolated cuspule. With extreme caution,we refer to this medial cuspule as a paraconule.There are two small transverse crests that leavethe buccal tip of the metaconule. The postero-most of these terminates in the slight swellingthat represents the most lingual of the three

cuspules in the posterior transverse row. Theanteromost of the two buccal metaconule crestsbifurcates into crenulations that run to the centreof the crown. The isolated median cuspule on theposterior transverse crest is (with extremecaution) referred to here as the neometaconule.The surface of the crown is dominated by thelarge central basin bounded on the anterior andposterior sides by the transverse ridges. The lowpoint of the occlusal surface of the crown is in themiddle of this basin. Thegotic facets are notapparent on the occlusal surface but the crown isalmost unworn. The lingual side of the crownexhibits very gentle crenulations on the flanks ofthe protocone and metaconule but there is nolingual cingulum. The anterolingual flank of theprotocone exhibits a sharp corner extendingfrom the tip of the cusp to the base of the crown.The crown has three flattened roots. There is onetransversely compressed and elongate lingualroot and one posterior and one anterior buccalroot both of which are very wide and antero-posteriorly compressed.

Discussion: Clarification should be given hereof the reasons for concluding that the paratypeand holotype are conspecific. First, both speci-mens represent a similar-sized species ofphalangeridan with very low-crowned, similarlycrenulated, bunoquadritubercular posteriormolars. Second, both represent a very rare taxonin assemblages of otherwise relatively well-represented species. The probability that the twospecimens represent different but similar-sized,similar-shaped, rare and yet different taxa seemsremote. However, the most compelling reason

POSSUMS AND OPOSSUMS: STUDIES IN LVOLUTION616

--r-

"'1

~.37

I11.20

I19.6 15"26

I

1.'12

-1- 3.21

i~2.99-1

J7~ r9-L~-1

Fig. 7. Measurements (mm) of Djilgaringa gillespiei. Tooth lengths and widths are maximum dimensions of crowns. Paratype.lower left. Holotype, above (buccal view) and right (occlusal view).

chitinous insects such as beetles or even smallvertebrates.

for concluding that they represent the sametaxon comes from consideration of occlusal

patterns.Djilgaringa thompsoni new

(Figs 8-9)The paratype occludes perfectly with theholotype ..even to the extent that the unusualmetastylid of the M3 perfectly occludes along theanterior flank of the distinctive transverse"crista" formed by the transverse alignmentof the protocone, buccal protocone crest,protoconule and reduced metacone. Similarly,the metaconule perfectly occludes within theshort but distinctive trigonid basin of M4"

Because this taxon is the only one for whichupper teeth are known, it is the best one to use inattempts to determine occlusal mechanics. con-sidering the transverse alignment and similar sizeof the principal upper cusps and cuspules andthe incipient or slightly developed transversecristids of the lower molars, it seems probablethat there was a significant transverse grindingphase in the chewing cycle.

The elaborate crenulations covering the sur-face of the crown would have acted as a veryeffective grating device during a transversegrinding phase as well as added to the durabilityof the crown surface. The wide bases of thecrowns, much narrower occlusal surfaces andshort, deep dentaries suggest that, whatever wasbeing comminuted, it required considerableforce. The powerfully built shearing premolarand combined M2 paracristid- P 3 shearing crestunit also suggest that this possum was cuttingsomething hard and resilient. Possibly it wasseeds with a hard husk or seed coat, very

Holotype: The holotype and only known speci-men is South Australian Museum SAM P24918,collected by a joint expedition of the RoyalAustralian Electrical and Mechanical Engineers(RAEME) of the Australian Army and theMuseum of Victoria on 23 August, 1984. It is aleft M2 in a fragment of a dentary that also retainsparts of the alveoli of C1 ' P ?2' P 3 and M3.

Type locality: The holotype was collected fromthe upper 10 centimetres of lake surface sedi-ment, approximately 5 metres east of the westernedge of Lake Yanda (31001 'S, 140°19'E),approximately 63 km northeast of Frome DownsStation Homestead, Frome Downs Station, SouthAustralia.

Local Fauna and Age: The faunal assemblagecollected at the type locality has been named theLake Yanda Local Fauna. The unconsolidatedsurface material comprising the sample collectedfor processing was not demonstrably in situ. Itcould represent: 1, an erosional wash from thewestern edge of the Lake Yanda outcrops andhence from an unknown horizon within theNamba Formation; or 2, a natural lag depositremaining in situ after the finer sediments fromthe originally fossiliferous layer were removed.

Tedford et al. ( 1977) suggest that this fossil-iferous deposit at Lake Yanda is part of the lowerunnamed member of the Middle Miocene

2 '69

'12.24

"-J

ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE 617

Fig. 8. Djilgaringa thompsoni, holotype, SAM P24918,left M2' from Lake Yanda, Frome DownsStation, South Australia. Upper, SEM stereophotographs of occlusal view (anteriortowards top). Lower, SEM photograph of oblique buccal view (anterior to left). The palearea on the buccal flank of the hypoconid is damaged enamel. (Photographs byH. Godthelp).

Tertiary Australian mammals. She has also beena mainstay in the preparation and curation ofthe Tertiary collections of the Museum ofVictoria.

Species diagnosiS': This species differs from Djil-gan:nga gilles~iei in that 1he P ~ was prob~bly lessoblIquely oriented, the M2 IS proporuonatelyshorter and wider, the talomd is proportionatelylonger, there is no protostylid, the anteroventralface of the trigonid is more steeply inclined,the metaconid is less distinct, the cristid obliquaappears to consist of a single uninterruptedcristid and the crenulations are more delicate.

Namba Formation and that the Yanda LocalFauna is probably equivalent to the Pinpa LocalFauna. Rich et at. (1982), however, suggest thatthe fossiliferous unit may be at the base of theupper unnamed member of the Namba Forma-tion. Further comments about this locality aregiven in Archer (1982: 404) but the precisestratigraphic position of the Yanda Local Faunawithin the Namba Formation is at present uncer-tain.

Etymology: This species is named in honour ofBetty Thompson who was involved in the col-lection of the holotype as well as many other

J'OSSUMS AND OPOSSUMS: STUDIES IN EVOLU [ON618

"\

2.7T

3'44-1- 3'06

Fig. 9. Measurements (mm) of the holotype of Djilgaringathompsoni. Tooth lengths and widths are maximumdimensions of crown.

The Yanda species is relatively morespecialised than the Riversleigh species in itsshorter, stockier molars. However, there areother characters that suggest that the Riversleighspecies is relatively more derived. For example,its p 3 is markedly oblique and was probablyhyperdeveloped with respect to the Yandaspecies, its p appears to have had only a singleroot, its posfuypocristid is rounded and unlikethe shape more typical of other phalangeridansand the protoconid is more buccal in position.Each species, therefore, appears to beautapomorphically specialised and unsuitable asan ancestor for the other. They are regardedhere as each other's sister-group.

Pilkipildra new

Genotypic species: Pilkipildra handi.

Additional species: Pilkipildra taylori.

Distribution: Tedford Locality, EtadunnaFormation, Etadunna Station, South Australia;Billeroo Creek (Site 3) and Tom D's Quarry onthe western side of Lake Tarkarooloo, NambaFormation, Frome Downs Station, South Aust-ralia.

Age: Medial Miocene (Woodburneet al. 1986).

Etymology and gender: The etymology of Pil-kiPildra is the same as that given above for thefamily. The genus is here given feminine gender .

Generic diagnosis: Species of Pilkipildra differfrom those of the only other pilkipildrid genus,Djilgaringa, as follows: they have a smaller andless buccally out-turned p 3; posterobuccal andposterolingual cingula on p 3; lower trigonids onM2; proportionately longer and narrower M2;less swollen M2 crown base; continuation of thecristid obliqua up onto the posterior flank of thetrigonid; a relatively well-developed paracristidon M2; a notch in the anterobasal cingulum of M2for receipt of the posterior edge of p3; a lessserrate and buccally poorly-developed transversecristid on the talonid of M2; less well-developedtransverse cristids on the talonids of M3-4; lesscrenulated crowns; a single pronounced vertical

Description: This description will confine itselfto aspects that differ from those of Djilgaringagillespiei.

The disposition of the alveoli in the dentaryfragment indicate that the premolar row wasmuch less irregular than that of D. gillespiei. Theposterior (and possibly only) root of p 2 occursanterior to the anterior root of p 3' ratDer thanlingual to it. There is also a suggestion from thealveolar margins that p 2 was two-rooted with amuch smaller anterior root. This conclusion isreached because of the sharp rise in the alveolarborders between what is therefore interpretedhere to be an anterior and a posterior root. In D.gillesPiei, p 2 appears to have had only one rootbecause a comparable rise in the alveolar marginis not evident.

The M of D. thompsoni differs from that toothin Do gillespiei as follows. It is proportionatelyshorter and wider. Its talonid is also proportion-ately wider and shorter than the trigonid. Thecristid obliqua is not interrupted by a transversefissure. The posthypocristid forms a gentle angleas it passes lingually to become the posteriorcingulum (in contrast to the continuous curvepresent in Do gillesPiei). The posterior flank of thetalonid is vertical such that the posteriorcingulum is the most posterior extension of thecrown (in contrast to Do gillespiei where the mostposterior portion is the posteriorly swollen baseof the cro~n). Despite the proportionately shorttalonid, the posterior cingular "basin" (the areabetween the posthypocristid-postentocristid andthe poorly-developed transverse ridge) is largerthan it is in D. gillespiei. The trigonid is propor-tionately more massive but lacks any trace of aprotostylid or buccal vertical crest on the flank ofthe protoconid. There are no coarse verticalcrenulations on the buccal flank of the trigonidand no anterobuccal basal cingulum. The para-cristid is proportionately .longer. The protoconidis situated in a more buccal position. Themetaconid is not distinct and appears to besmaller than that cusp in Do gillesPiei.

Discussion: This species is unusual but not onlybecause of its peculiar morpholoy. On the apicalbuccal face of the trigonid, just below theprotoconid, there is a thegotic facet that does notappear to have a counterpart in Do gillesPiei. It dis-plays some microstriations that are parallel tothose which occur on other parts of the M2 andwhich indicate an oblique vertical shearing phaseas the teeth approach the centric position. How-ever, in this facet there are many more micro-striations at right angles to these which suggestthat the trigonid of M2 was being forced in ananteroventral direction to hone the postero-lingual flank of p3o

A wear facet in a similar position on M2 of D.gillespiei does not appear to be a thegotic facet, itssurface being irregular rather than polished andcovered by pits and multidirectional scorings.

ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE 619

Etymology: This species is named after SuzanneHand (formerly of Macquarie University andnow of the University of New South Wales) who,besides being a research colleague and the prin-cipal research worker to make sense of themyriad bat fossils from Riversleigh, has spentfour field seasons in the Riversleigh area (sup-ported in part by ARCS funds) and two in SouthAustralia helping to collect many of the newmammals that have been found in these areas.

crenulation on the posterior face of the M2trigonid below the metaconid; and proportion-ately shorter and wider posterior molars.

DiagnosiS': This species differs from PilkiPildrataylori (which is known only from an isolated M2)as follows: the M2 is shorter; the trigonid is pro-portionately shorter and wider; the posteriorcingular basin of the talonid is shorter and lesswell developed on the buccal side; the posthypo-cristid is less crenulate; the vertical crest belowthe metaconid is less extensively developed; thebuccal segment of the transverse talonid cristid isless well developed; the anterior basal notch isnarrower because the buccal side of the notch isless well developed; and the postentocristid isbetter developed.

Description: This description is based on a com-parison with the only other reasonably completepilkipildrid dentary known, the holotype of Djil-garinga gillesPiei and, in large part, will be anannotation of the differences between the two.

Pilkipildra handi new(Figs 10-12)

Holotype: The holotype is South AustralianMuseum palaeontoloical No, P27909, a leftdentary fragment with damaged P 3 and M2'complete M3,.1. and alveoli or parts of alveolirepresenting LI' P ?1' P ?2 and M5'

Type locality: The holotype was collected at Bil-leroo Creek, Site 3 (31°9'S, 140°14'E), FromeDowns Station, South Australia. It was obtainedfrom low outcrops of the Namba Formationexposed on the north side of the Creek 2.4 kmnortheast of the north end of Lake Pinpa. Thesediments at this locality consisted of green clay-stones overlaying dolomitic claystones,

Paratype and locality: An almost unworn isolatedleft M2 (NMV P157570) appears to represent thistaxon. It was collected in 1976 by a joint expedi-tion of the Queensland Museum, Museum ofVictoria and Monash University from Tom 0'sQuarry (31°8'30"S, 140°6'20"E), on the westernside of Lake Tarkarooloo, Frome Downs Station,South Australia,

Local faunas and age range: the Billeroo Creektype locality h':ls produced an unnamed localfauna evidently roughly equivalent to the PinpaLocal Fauna in terms of stratigraphic level withinthe Namba Formation; Tom O's Quarry appearsto be stratigraphically above the sites at LakePinpa and Billeroo Creek, The Namba Forma-tion is interpreted (Woodburne et al, 1985) to bemedial Miocene in age.

The dentary is broken posterior to the alveolifor M5 and anterior to the posterior wall of the Iialveolus. There are at least two small alveolianterior to p 3. Although their homology is uncer-tain, their similarity to the alveoli in the holotypeof D. gillespiei suggests that they represent asingle-rooted PI and p 2' The root of the lower

(')3

Fig. 10. PilkipildrQ handi, holotype, ~ AM P27909, from Lake Pinpa, Frome Downs Station, South Australia,left dentary with alveolifor P 2 and Ms plus P 3-M .P 3 and M2 are significantly damaged. A, lingual view (anterior towards bottom). B, buccal view(anterior towards top). a, stereophotographs of occlusal view (anterior towards top). (Photographs by ~. Tarka).

620 ;POSSUMS AND OPOSSUMS: S- DIES IN EVOLUTION

Fig. 11. PilkipildrQ handi, paratype; NMV P157570, LM2' from Lake Tarkarooloo, Frome Downs Station. South Australia.Left, SEM photograph of occlusal view (anterior towards top). Right, SEM photograph of buccal view (anterior toleft). (Photographs by H. Godthelp).

paratype. Otherwise there are no cingula on thebuccal, lingual or anterior sides of the crown.There is no protostylid. The buccal face of thetrigonid is smooth (rather than verticallycrenulated), although there is a prominent verti-cal swelling that extends ventrally beyond thesteeply sloping metacristid. The cristid obliquaruns smoothly in a sinuous curve from thehypoconid to the posterior flank of the trigonidin a position immediately lingual to theposterobuccal end of the metacristid. An obliquewear facet connects the end of the two cristids.Apart from the single prominent vertical crenu-lation that extends down the posterior flank ofthe trigonid from the metaconid, there are noprominent vertical crenulations on this part ofthe trigonid. The metaconid is relatively promi-nent and curved. It extends anteriorly from theprotoconid and then, about halfway along itslength, it swings anterolingually and continues tothe basal anterior end of the crown where itforms the lingual flank of the anterior basalnotch. The lingual flank of the trigonid is con-cave because of the curvature of the paracristid.The hypocristid is sinuous rather than evenlycurved. It gently swings posterolinually and thentranversely from the hypoconid until it reachesthe mid-point along its length where it bends for-ward, then backwards and then curves antero-lingually to gradually merge with the postento-cristid. This inflexion in the hypocristid corres-ponds with the anterior end of M3 and probablyhelped to reduce independent transverse move-ments of the teeth. The talonid basin is lesscrenulated than it is in the species of Djilgaringa. Thetalonid basin is proportionately longer, narrower ,deeper and less crenulated. The transverse talonidcristid is not serrate and is less well developed than inthe species of Djilgaringa. It extends buccally fromthe entoconid to the midline of the talonid whereit vanishes at the base of the lingual flank of thehypoconid. The posterior cingular pocket

incisor was very large judging from the dimen-sions of the alveolus that remains. It extendedposteriorly to at least the level of M2. The mentalforamen occurs at about mid-depth down thedentary and opens just behind the anterior rootof p 3. A groove begins at this opening and con-tinues steeply upward across the anterior root ofp 3 to the alveolar margin. A tiny posteriorlyopening foramen occurs between M2 and M3 onthe buccal side of the dentary at about mid-depth.

The p 3. is damaged but enough remains todemonstrate its distinctive shape. It was narrowerthan the p 3 of D. gillesPiei and had posterobuccaland posterolingual cingula. A steep but well-developed thegotic facet extends along thebuccal side of the posterior remnant of themedian longitudinal crest and down onto theposterobuccal cingular crest. The remnants ofthe buccal and lingual flanks of the crown suggestthat there was a basal swelling and that the shear-ing portion of the tooth was narrower than thatregion of the p 3 in D. gillespiei. The remnantslopes of the tooth suggest that it was shorter-crowned and there is no indication of verticalribs. Altogether, it appears to have been a moreslender, shorter and less sectorially specialisedtooth.

There are two M2's known (the holotype andthe paratype). The anterior end of M2just over-laps the posterolingual basal cingulum of the p 3(rather than the posterior part of the medianlongitudinal crest). The base of the trigonid isonly gently inclined with respect to the base of thetalonid. The M2 crown is relatively long for thetooth row, in part because the trigonid is lesscompressed anteroposteriorly. The anteriorbasal portion of the crown is developed into twoswellings that define a small median basal notch.A tiny anterobuccal basal cuspule or cingularremnant occurs on the anterobuccal flank of theparacristid near the anterior notch on the

521ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE

suggests that the bite was not as powerful. Thesignificance of the difference in orientation ofthe p 3 is unclear although it must reflect adifferently shaped p3 and perhaps a less roundedface. The dentary is certainly less arcuate andsuggests that the lower jaw was relatively longer .The smaller and less extensive crenulations onthe molar surfaces and the relatively moreprominent and non-serrated cristids suggest thatP. handi spent relatively less time transverselygrinding hard foods and more time shearing andgrinding softer foods. The relative reductionof the cingular pocket on the talonid suggeststhat the metaconule of the as yet unknownupper molars was smaller (or less distinct}and that a less significant break occurredbetween the premetaconulecrista and the post-protocrista. The presence of the trigonid basin,however, suggests that the metaconule waspresent even if it was smaller than that cusp in p .handi.

(between the transverse talonid cristid and theposthypocristid) is relatively small and restrictedto the lingual half of the talonid. The walls of thispocket are not crenulated.

The M is only represented in the holotype.Comparea with the M3 of Djilgaringa gillespiei it isproportionately shorter and wider althoughlonger in absolute dimensions. The basal buccalinflexion between the trigonid and talonid occursin a more posterior position. There is noposterior cingular pocket and no transverse cris-tid on the talonid. The talonid basin is relativelyless crenulated. The trigonid basin is not clearlycrenulated. The metastylid is smaller. Thehypocristid is less curved and gives the posteriorend of the tooth a more rectangular (rather thenround) appearance.

The M4 is only known in the holotype. Com-pared with that tooth in D. gillespiei, it is relativelymuch shorter, wider and less crenulated. There isno clear longitudinal groove in the talonid. Themetastylid is very poorly developed. The trigonidbasin is not crenulated.

The M5 is unknown but its anterior alveolus inthe holotype suggests that it was also proportion-ately wider than that tooth in D. gillespiei.

Discussion: This species is the best-known of thegenus. The relatively unworn isolated LM2(NMV P157570) from Lake Tarkarooloo addsimportant information lacking in the holotype.Fortunately, enough of M remains in theholotype to demonstrate the aimost indistinuish-able shape and size of the paratype and hence itsconspecificity with the holotype.

Compared with the dentition of Djilgaringa gil-lespiei, the teeth of PilkiPildra handi suggest thatthis possum used its cheekteeth in a slightlydifferent way. The smaller and less swollen P 3

~2.52i

T4.51

~13"40

(

3.54

r

4.37

~

3,42

~

2,87

1/

Pilkipildra taylori new(Figs 13-14)

Holotype: The holotype and only known speci-men is South Australian Museum Palaeonto-logical collections No. P27910. It is an isolated

LM2.Type locality: The holotype was collected by

Woodburne, Archer et al. in 1972 from TedfordLocality, Lake Palankarinna, Etadunna Station,South Australia.

Local fauna and age: Tedford Locality is thetype locality for the Ditjimanka Local Faunawhich is interpreted (Woodburne et al. 1985) tobe medial Miocene in age.

Etymology: This species is named in honour ofJenny Taylor who spent almost seven years as adistinguished part-time preparator, sorter ,photographer, drafts person and artist connectedwith the vertebrate palaeontological researchprogramme at the University of New SouthWales.

Diagnosis: PilkiPildra taylori differs from P.handi as follows: M2 is longer; the trigonid isproportionately longer and narrower; theposterior cingular basin of the talonid is longerand better developed on the buccal side; the post-hypocristid is more crenulate; the vertical crestbelow the metaconid is better developed; thebuccal segment of the transverse talonid cristid isbetter developed; the anterior basal notch iswider; and the postentocristid is less well

developed.Description: The holotype is similar to the M2 of

P. handi except for the features noted above inthe diagnosis. Overall, it is a larger tooth than theM2 of P. handi. It also exhibits a slightly corru-gated buccal protoconid flank in contrast to thesmooth buccal flank of P. handi.

IIFig. 12. Measurements (mm) of Pilkipildra handi. All

measurements are maximum crown widths, lengthsand heights. Left, paratype. Right, holotype.

622 ;POSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION

Although it is unfortunate thatP. taylori is onlyknown from M2' this tooth is the most distinctiveof the lower molars and, judging by the differ-ences in M2 morphology in other pilkipildrids, agood basis for recognition of species.

Considering the many tonnes of matrix thathave been processed from Tedford Locality since1971, there seems little point in putting offdescription of P. taylori on the chance thatadditional material would turn up in the future.

Fig. 13. Pilkipildra taylori, holotype, SAM P27910, LM2' fromLake Palankarinna, Etadunna Station, SouthAustralia. A, SEM photograph of occlusal view(anterior towards top). B, line drawing of same tooth:occlusal (upper left); buccal (upper right); anterior(lower left); posterior (lower centre); lingual (lowerright). Bar scale = 2 mm. The occlusal view in 13B

is more directly overhead; the occlusal view of 13A isan oblique view from a more posterior position.(Photograph by H. Godthelp; drawings by M. Wood-burne).

2.94

4'77

~

DISCUSSION

Four aspects of pilkipildrids invite discussion:their interfamilial relationships; their intra-familial relationships; their ecological role; andtheir chronostratigraphic significance.

The interfamilial relationships of the Pilki-pildridae within Diprotodontia are somewhatambiguous. There are three most reasonable,hypotheses: I, either the pilkipildrids are mostclosely related to the Petauridae (Fig. 15A; thehypothesis favoured here); or 2, they are moreclosely related to a cluster of families thatincludes the Phalangeridae, Ektopodontidae andMiralinidae (Fig. 15B-C); or 3, they form a dis-tinct major group of phalangeridan possums,one which is no more closely related tophalangerid-like forms than it is to petaurid-Iikeforms (Fig. 15D).

Pilkipildrids and petaurids share manyfeatures of molar morpholoy including thefollowing: 1, bunodonty; 2, quadritubercularmolar shape; 3, extensive fine crenulations; 4,point focus of the crenulations of the lowermolars into the centre of the crown; 5, a relativelysimple cristid obliqua; 6, a comparably modifiedtrigonid on M2; 7, a similar trigonid basin on theM3-4; 8, low transverse cristids on the talonids ofM2-3 and trigonids ofM3-4; 9, reduced dimensionsand almost flat occlusal surface of M. ; 10, a simi-lar dentary shape; 11, relatively s~all medialindentations on the buccal and lingual surfaces ofthe lower molars; 12, narrow posterior halves ofthe upper molars; 13, a strongly curved paracris-tid on M2; and 14, reduction of en echelon cristid

relationships.However, pilkipildrids also share with

phalangerids, miralinids end ektopodontidsseveral features (Fig. 15B-C) including: 1, atendency to elaborate conules and crenulationsinto cuspidate transverse crests; 2, presence of avertical cristid below the metaconid on theposterior flank of the trigonid; 3, quadri-tubercular molars; 4, point focus of the crenula-tions of the lower molars; 5, a comparably modi-fied trigonid on M2 (ektopodontids differ and areunique here); 6, afunctional M2 paracristid-P 3shearing unit (again ektopodontias differ here);7, a very large and sometimes buccally deflected

~3.06

Fig. 14. Measurements (mm) of Pilkipildra taylori. Allmeasurements are maximum crown length andwidths.

Discussion: The two species of PilkiPildra aremuch more similar to each other than are thespecies of Djilgaringa leaving little doubt that theyare congeneric. On the other hand, they exhibitsuch different proportions that there can also belittle doubt that they represent distinct-species.

623ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE

p 3 (and again ektopodontids display anautapomorphic condition here, one that appearsto converge on that seen in petaurids); 8, cusps ofthe upper molars arranged into transverse rowswith the buccal cusps being subequal in height tothe lingual cusps; 9, reduction of the paraconeand metacone; 10, development of trigonidbasins on M3-4. with their associated transversetrigonid cristids (phalangerids have also donethis on all of the lower molars and on the talonidsas well, features not present in the lower molarsof pilkipildrids or petaurids).

The third major phylogenetic hypothesis, anunresolved trichotomy (Fig. 15D), depends for itsdefence on the inability to confidently recognisedichotomies.

Choice between the three phylogenetichypotheses must depend to a certain extent onpreconceptions about which dental features aremore reliable indicators of relationship. Atpresent, we have no other guide such as basi-cranial morphology .

Perhaps the best evidence for phalangeridrelationship is the upper molar morphology. Theincipient but incomplete transverse lophs and theevidently reduced metacone and paracone arefeatures found in some phalangerids (albeit themore derived taxa; see Flannery et at. 1987) butare not known to occur in petaurids. Added tothis is the well-developed phalangerid-like M2paracristid- P 3 shearing crest unit. Again, nothinglike this develops in petaurids but its absencecould be the result of the marked reduction of P 3in petaurids. This possibility is significantbecause a comparable functional unit occursin some macropodoids, miralinids, someburramyids and even thylacoleonids. Although itis possible that this two-tooth shearing-unitevolved independently in these groups, it is alsopossible that it represents a symplesiomorphicfeature of diprotodontians.

suggest that the similarities of upper molarmorphology between phalangerids and pilki-pildrids are based on symplesiomorphy. In turn,that suggests that the albeit low and poorly-formed transverse crests of petaurid uppermolars are autapomorphic developments of thetransverse rows of low cusps that would havebeen present in a petaurid-pilkipildrid commonancestor. Phylogenetic transition of this sort mayhave occurred several times in diprotodontmarsupials such as in phalangerids, macro-podoids and possibly wynyardiids (e.g., seeArcher 1976; Tedford et at. 1977; Rich andArcher 1978; Archer 1984 and Pledge 1987).

Accepting petaurids to be the group mostclosely related to pilkipildrids has other con-sequences. Petaurids are clearly related topseudocheirids among families of phalangeri-dans with living representatives (e.g., see McKay1984). They may also be related to acrobatids andtarsipedids (Aplin and Archer 1987) but theevidence for this is ambiguous. If pseudocheiridsare accepted as the sister-group of the pilki-pildrid-petaurid sister-group, they may serve asthe outgroup for interpretation of the probableplesiomorphic states of polymorphic charactersin the pilkipildrid-petaurid sister-group. In thiscase, the following features of pilkipildridswould probably be autapomorphic ratherthan symplesiomorphic pilkipildrid-petauridfeatures: The steeply inclined Ii; the large, some-times out-turned and sometimes serrated P ; theprominent and isolated metastylids (which do notappear to be the same as the metastylids ofpseudocheirids in that they are not part of ordefined by a postmetacristid); the very narrowposterior half of M5 (this feature occurs also inmiralinids); the gently sloped lingual flanks ofthe upper molar; the well-rounded corners of theupper molars; the very prominent, tall andswollen buccal cusps of the upper molars; and thevery reduced paracones and metacones.

Similarly, autapomorphic petaurid dentalfeatures would include: marked reduction of P 3and incorporation (with loss of identity) of theintermediate conules of th~ upper molars intolow transverse crests.

Other non-pseudocheirid-like features inpilkipildrids and petaurids are noted above as

hypothetical synapomorphies.

We should consider whether the Pilkipildridaemight be paraphyletic and more properly includewithin it the Petauridae. In other words, isjustoneof the pilkipildrids or pilkipildrid g~nera thesister-groupof Petauridae? Although possible, itseems improbable. Among the diagnosticfeatures of the Pilkipildridae are the large P 3 andisolated metastylids, features that do not appearto be plesiomorphic within the Petauroidea (they

The best evidence for petaurid relationshipseems to be the strikingly similar morpholoy ofthe posterior lower molars, These are very low-crowned, have reduced or absent transversetalonid cristids, a very reduced M5, a longitudinalmedian groove on the posterior lower molars,reduction of en echelon crest relationships,barely indented buccal and lingual margins onthe lower molars and marked transverse con-striction of the posterior half of M5' Thesefeatures are unknown in phalange rids which allhave well-developed transverse crests on all ofthe lower molars, large M5s, strongly indentedcrown margins and marked en echelon crestrelationships, particularly in the lower molars,

Tentatively , pending discovery of more com-plete material, we are inclined to favour thehypothesis of petaurid relationship and to

624 J'OSSUMS AND OPOSSUMS: STUDIES IN EVOLUTION

~ wMacropodoipea

~

Ma(ropodoiQea

y BA

.(V'

f2f (

fj)

Fig. 15. Most-supported cladograms involving the Pilki-pildridae. Character states for ektopodontids arebased, where possible, on Chunia cf C. illuminata Wood-burne and Clemens, 1986 (e.g., AMNH ,95584), themost plesiomorphic of known forms. Those forphalangerids are based on the "simple" -crownedforms such as Phalanger ursin1L5 (see Flannery, Archerand Maynes 1987). Those for petaurids are based onPetaurus alL5tralis (e.g., AM M9550). Those forpilkipildrids are mainly Pilkipildra handi. Those forpseudocheirids are based on the "simple" species ofPseudocheirus such as P. peregrinlL5. Tarsipedidae, etc.,refers to the apparent relationship between tarsi-pedids with acrobatids (see Baverstock, Birrell andKrieg 1987). Potential synapomorphies andautapomorphies (the hypothetical phylogenetic sig-nificance of these features depends on the position ofthe taxon in the alternate cladograms) are as follows: I,a combined shearing unit involving the M2 trigonidand the posterior crest of P 3; 2, the ectotympanic issolidly fused to the surrounding bones of the middleear; 3, principal crests (the prehypocristid/postproto-cristid, preentocristid/postmetacristid and pre-metaconulecrista/postprotocrista oppositional pairs)meet in an en echelon pattern, there is a tendency toenlarge the conules (intermediate cuspules betweenthe principle cusps) and the metacone is enlarged; 4,the cristid obliqua has a "kink" near its anterior end(after leaving the hypoconid, the cristid obliqua turnssharply lingually, sometimes even intersecting a shortanterolingual crest from the hypoconid, before meet-ing the posterolingual end of the postprotocristid), theconules are arranged into high transverse ridges onthe upper and lower molars, the ancestral selenodontoat tern is modified into a bunodont semi-selenodont

condition and there is commonly a notch in theanterobuccal corner of the anterior cingulum of thelower molars (the condition varies intraspecifically butthe tendancy occurs only with the three families forwhich it appears to be a synapomorphy); 5, the trans-verse crests are relatively smooth rather than denticu-late, the P3 is markedly out-turned and the lowermolars exhibit a pronounced transverse constrictionbetween the trigonid and talonid; 6, there is an elab-oration of conules on the lower molars, an elaborationof the M2 parastyle, markedly serrate transverse crests,a modification of the "kink" in the cristid obliqua suchthat a small part becomes lingually isolated from therest of the crest and the P3 is inturned (i.e., the crownis oriented anterolingually); 7, there is a steep molargradient reducing posteriorly, a very narrow talonidon M5' a highly compressed M2 trigonid with a distinctanterobuccal swelling, a very narrow P 3 and steep andbuccally displaced paracone crests on M2; 8, many con-ules comprise horizontal transverse crests, premolarsize is reduced and the molars are short and wide; 9,biochemical affinity of living members (e.g., Kirsch1977 and Baverstock, Birrell and Krieg 1987) andreduction of MI; 10, the protoconid is displaced to thelingual side of the trigonid of M2' the cristid obliqua ofM2 extends to the top of the metaconid, thehypoconids and posterior protoconids are markedlyangular and the postprotocristids extend across thetooth to the lingual side of the crown; 11, molecularaffinity of living forms (e.g., Baverstock, Birrell andKrieg 1987); 12, selenodonty is reduced and there is anelaboration of a bunodont quadritubercular pattern,the posterior molars are very low crowned, thereis a steep molar gradient which reduces posteriorly,the M. is reduced with a verv reduced talonid width.

ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE 625

Fig. 15 cont. -the transverse cristids on the talonids of theposterior molars are reduced, there is a longitudinalmedian groove on the posterior lower molars, the enechelon crest relationships are reduced as are themedian transverse constrictions on the lower molars;13, the II is steeply inclined, the P3 is large, the meta-stylids are prominent and isolated, the lingual flanks ofthe upper molars are small and vertically sloped as arethe buccal flanks of the lower molars, the comers of theupper molars are well rounded, the M3 talonid isnarrow, the buccal cusps of the upper molars areprominent, tall and swollen and the paracones andmetacones are very reduced; 14, there is a markedreduction of P3 and a loss of distinction of the conulesof the upper molars; 15, there is a tendency to developtransverse crests with cuspules, the cristid obliqua ofM2 runs to the posterior flank of the protoconid (orprotostylid) rather than the metaconid or between theprotoconid and metaconid and there is a small verticalcristid on the lingual posterior flank of the trigonid ofM2; 16, see 9; 17, there are prominent and isolatedmetastylids, a loss of the "kink" in the cristid obliqua, aloss of the anterobuccal notch in the lower posteriormolars, well-rounded corners on the upper molars, thetalonid of M has a very narrow talonid, the molarsshow a mar[ed decline in size posteriorly and theparacones and metacones are reduced; 18, see 10; 19,see 12 and 14 except that here the low transverse molarcrests and <;ristids are autapomorphic; 20, the M5 isvery reduced, the buccal sides of the lower and lingualsides of the upper molars have gentle slopes, the uppermolars have rounded comers and there are semi-Iophsformed by small cuspules; 21, see 17 but here the well-rounded comers of the upper molars aresymplesiomorphic; 22, there are prominent andisolated metastylids, reduced posterior molars,markedly reduced and low-crowned M5, a verticallingual cristid on the posterior flank of the M trigonid,well-rounded'comers on the upper molars, low trans-verse crests on the lower molars, transverse rows oflowcusps on the upper molars, reduced paracones andmetacones, prominent anterobuccal cusps on theupper molars, a steeply inclined II and gently inclinedli!lgual flanks on the upper molars and buccal slopes ofthe lower molars; 23, the cristid obliqua has a "kink" atits anterior end and there is a tendency to develop anotch in the anterobuccal comer of the posterior lowermolars.

do not appear to be any potential synapomorphicstates that would support a hypothesis that thepilkipildrids were actually the sister-group ofjustone subgroup of the Petauridae (e.g., speciesof Gymnobelideus, Dactylopsila-Dactylonax orPetaurus). There are, in contrast, shared-derivedstates that support the hypothesis of petauridmonophyly including the gross reduction of P3and the loss of distinction of the conules on theupper molars.

Relationships among the known pilkipildridsare less obscure. Clearly there are two clades,each presented here as a new genus (Fig. 16). Ofthese, the species of PilkiPildra appear to be themore plesiomorphic if Petauridae is accepted asthe sister:.group of Pilkipildridae. They retainwhat appears to be the plesiomorphic states forlower molar morphology (e.g., less well-developed transverse talonid cristids). Of thespecies of Djilgaringa, D. thompsoni appears to bethe more plesiomorphic in its lack of a protostylidon M2' its apparently two-rooted P 2' its smaller P 3and in its less elaborate crenulations. The differ-ences between the species of Pilkipildra are lesssignificant but the short proportions, smalltalonid cingular basin and poorly developedtransverse talonid cristid of P. handi are morePetaurus-Iike than the corresponding states in P.taylori.

The intrafamilial phylogenetic relationshipsmay have chronostratigraphic significance. First,it is clear that Pilkipildra handi (Pinpa Local Faunaand Tarkarooloo Local Fauna) and P. taylori(Ditjimanka Local Fauna) are very similar to eachother and probably comparable in age. There isno way of confidently determining which (ifeither) of the two is the older, although P. handiappears to be structurally a bit more plesio-morphic.

In contrast, Djitgaringa gillesPiei (Riversleighlocal faunas) exhibits several features thatsuggest it is more derived than D. thompsoni (theYanda Local Fauna) and hence that it is probablyyounger in age. This hypothesis is supported byaccumulating evidence that suggests closerchronologic ties between the Riversleigh Gag andLast Minute local faunas and the KutjamarpuLocal Fauna of South Australia (Archer in prep.),the latter generally being regarded (e.g.,Woodburne et at. 1985) as younger in age thanthe Ditjimanka, Pinpa, Tarkarooloo and Yandalocal faunas. It should be pointed out here thatother more recently discovered Riversleigh localfaunas (e.g., Mike's Menagerie) appear to havecloser ties to the older local faunas of centralAustralia (e.g., the Tarkarooloo Local Fauna) onthe basis of other taxa (e.g., species ofNamitamadeta). However, these apparently olderRiversleigh local faunas have yet to producepilkipildrids.

are absent in petaurids and pseudocheirids) and,therefore, evidence against the hypothesis ofPilkipildridae being a paraphyletic taxon. Addi-tional autapomorphic and non-petaurid featuresin pilkipildrids that argue against inclusion of thePetauridae within the Pilkipildridae include therelatively short anteroposteriorly compressedtrigon ids on the lower molars, the relatively wideand short lower molars and the highlyautapomorphic morphology of the upper molarsincluding the reduced paracone and metaconeand relatively very deep transverse fissure whichbisects the crown. In all of these characters,all known petaurids display plesiomorphicconditions.

As a final consideration regarding interfamilialrelationships of the pilkipildrids, we shouldconsider the possibility that the Petauridae astreated here may be paraphyletic because thePilkipildridae may actually belong within thatfamily rather than outside it as its sister-group.Against this view, however, is the fact that there

626 POSSUMS AND OPOSSUMS: STUDIES IN EVOLU [ON

Q. thompsoni Q. qiltespiei

""' /

>Jf

y~~ ~

Petauridaey

Pilkipildridae

~

~

,(1)'

Fig. 16. The most-supported cladogram of hypotheticalintrafamilial relationships within Pilkipildridae.Potential synapomorphies are as follows: 1, seecharacter-states No.12 in Fig. 15; 2, see character-states No.14 in Fig. 15; 3, the II is steeply inclined, theP3 is large, there are prominent and isolated meta-stylids, the lingual flanks of the upper and buccalflanks of the lower molars are gently inclined, thecorners of the upper molars are well rounded, the Mgtalonid is narrow, the buccal cusps of the upper molarsare prominent, tall and swollen, there are very reducedparacones and metacones, transverse rows of cuspulesare developed on the upper molars and there are veryshort trigonid basins; 4, there are posterobuccal andposterolingual cingula on P g' there is a well-developednotch on the anterobasal cingulum of M..2 and there isan overall reduction of crenulations; -', the talonidbasin on M is well developed as is the vertical cristid onthe posterolingual flank of the trionid of M2' there is arelatively well-developed posterior transverse cristidon the talonid of M2; 6, the trigonid of M2 is short andwide, the M2 is shortened and the postentocristid isreduced; 7, the P3 is large and obliquely oriented,there is a loss of the posterior cingula, the M trigonidis elevated, the molars are short and wide, the crownbases of the lower molars are more swollen, the cristidobliqua and paracristid on M2 are shortened and thenotch on the anterobasal flanK of M! is reduced; 8, theM2 is proportionately short and wide, the M2 trigonidis very steeply inclined and the metaconid of M2 isbarely distinct; 9, the P 9 is very obliquely oriented, thetalonid is proportionately short, the M2 has a smallprotostylid, the cristid obliqua is divided into anteriorand posterior segments and the crenulations of thecrowns are pronounced.

Overall, the upper and lower crown mor-phology of pilkipildrids is reminiscent of, but lesseleborate than, that exhibited by some of theomnivorous phalangerids (e.g., Phalangermaculatus) and petaurids (e.g., Petaurus breviceps).Perhaps pilkipildrids were phalangerid- andpetaurid-Iike in diet; i.e., opportunistic omni-vores. Their small size (relative to phalangerids;comparable-sized petaurids being as yetunknown from this time) suggest that they mayhave exploited the smaller range of this omni-vorous habitus which was otherwise unoccupied

by phalangerids.

research possible. Archer's 1983, 1984 and 1985fieldwork at Riversleigh was supported by theAustralian Research Grants Scheme (Grant No.E7915864). Additional aid was provided by theQueensland Museum and anonymous donors.Woodburne's and Archer's work in South Aust-ralia was supported by a National Science Found-ation Grant to Woodburne (Grant No. GB-35488) and Australian Research Grants Schemesupport to Archer, T. Rich and P. Rich (GrantNo. E76/15122). Tedford's work in the FromeBasin was supported by a National Science Foun-dation Grant (Grant No. GB-18273X1). Otherfinancial support for South Australian work wasprovided to Archer by the University of NewSouth Wales and the Queensland Museum and toRich by the National Georaphic Society (GrantNo.1562).

Invaluable field and other support wasprovided by many colleagues (in alphabeticalorder): K. Aplin, E. Archer, R. Beale, R. Brown,F. Bussat, G. Clayton, C. Cleeland,J. Courtenay,T. Flannery , A. Gillespie, H. Godthelp, K.Gollan, K. Grimes, C. Hann, S. Hand, W. Head,Q.Jones, R. Langford, P. Lawson, D. Marina, M.Macdonald, W. Moore, P. Morrison, M. Plane, N .Pledge, P. Rich, I. Stewart, J. Taylor, E.Thompson, R. Wells, M. Whitelaw, M. Wood-burne and many undergraduate students of theUniversity of New South Wales.

Particular thanks are extended to Ken Grimes(and the Geological Survey of Queensland) forhis part in the discovery of the new Riversleighsites, to John Courtenay (and the Burke Shire)for his role as co-ordinator of Probe ScientificExpeditions in organising most of the logisticsand support for the 1984-85 Riversleigh expedi-tions and for helping to discover many of the bestsites and to Henk Godthelp for his vital role indiscovering and preparing (via A.R.G.S. support)most of the Riversleigh material. Critical logisticsupport for the Riversleigh work was alsoreceived from the Wallaby Task Force of the 35thSquadron of the RAAF based in Townsvilleduring 1984-85. They performed the seeminglymiraculous task of shifting 12 tonnes of pilki-pildrid-bearing rocks from the remote andalmost inaccessible outback of Riversleigh Stationto the nearest point of road transportation. TheBurke Shire Council then provided much-appreciated support in transporting the rocks toMount Isa. Personnel of the 4th Base Workshop,RAEME, Bandiana, Victoria, not only providedthe logistic support essential to reach Lake Yandabut also had a crucial role in the collection of thefossil vertebrates from this remote area of centralAustralia.

The manuscript was typed by T. Deguara andthe figures prepared by J. Taylor, P. Watkins andH. Godthelp (University of New South Wales).Photographs were taken by H. Godthelp and R.Arnett (University of New South Wales) and C.Tarka (American Museum of Natural History).

ACKNOWLEDGEMENTS

The greatest debt of gratitude we owe is to thevarious funding bodies that have made this

627ARCHER, TEDFORD, and RICH: THE PILKIPILDRIDAE

REFERENCES

APLiN, K. AND ARCHER, M., 1987. Recent advances in mar-supial systematics with a new syncretic classification. Pp.xv-Ixxii in "Possums and opossums: studies in evolution"ed by M. Archer. Surrey Beatty & Sons and the RoyalZoological Society of New South Wales: Sydney.

ARCHER, M., 1976. Phascolarctid origins and the potential ofthe selenodont molar in the evolution of diprotodontmarsupials.Mem. QdMw. 17: 367-71.

ARCHER, M., 1978. The nature of the molar-premolar bound-ary in marsupials and a reinterpretation of the homologyof marsupial cheek-tooth. Mem. Qd Mw. 18: 157-64.

ARCHER, M., 1982. Review of the dasyurid (Marsupialia) fossilrecord, integration of date bearing on phylogeneticinterpretation, and suprageneric classification. Pp. 397-443 in "Carnivorous marsupials" ed by M. Archer. RoyalZoological Society of New South Wales: Sydney.

ARCHER, M., 1984. The Australian marsupial radiation. Pp.633-808 in "Vertebrate zoogeography and evolution inAustralasia" ed by M. Archer and G. Clayton. HesperianPress: Penh.

ARCHER, M. AND FLANNERY, T. F ., 1985. Revision of theextinct gigantic rat kangaroos (Potoroidae: Marsupialia),with description of a new Miocene genus and speciesand a new Pleistocene species of Propleopu.s. J. Paleo. 89:1131-49.

BAVERSTOCK, P. R., 1984. The molecular relationships ofAustralasian possums and gliders. Pp. 1-8 in "Possumsand gliders" ed by A. P. Smith and I. D. Hume. SurreyBeatty & Sons and the Australian Mammal Society:

Sydney.

BAVERSTOCK, P. R., BIRRELL,J.AND KRIEG, M., 1987. Albuminimmunologic relationships among Australian possums: aprogress report. Pp. 229-34 in "Possums and opossums:studies in evolution" ed by M. Archer. Surrey Beatty &Sons and the Royal Zoological Society of New SouthWales: Sydney.

FLANNERY, T. E., ARCHER, M. AND MAYNES, G., 1987. Thephylogenetic relationships of living phalangerids(Phalangeroidea: Marsupialia) with a suggested newtaxonomy. Pp. 477-506 in "Possums and opossums:studies in evolution" ed by M. Archer. Surrey Beatty &Sons and the Royal Zoological Society of New SouthWales: Sydney.

McKAY, G. M., 1984. Cytogenetic relationships of possumsand gliders. Pp. 9-16 in "Possums and gliders" ed by A. P .Smith and I. D. Hume. Surrey Beatty & Sons and theAustralian Mammal Society: Sydney.

PLEDGE, N. S., 1987. Muramura williamsi, a new genus andspecies of ?wynyardiid (Marsupialia: Vombatoidea) fromthe Middle Miocene Etadunna Formation of South Aust-ralia. Pp. 393-400 in "Posums and opossums: studies inevolution" ed by M. Archer. Surrey Beatty & Sons andthe Royal Zoological Society of New South Wales:

Sydney.

REUTHER, J. G., 1901. "The Diari". Translated (as "A DiariDictionary") by Rev. P. A. Scherer. Awtralian Institute ofAboriginal Studies Microfiche No.2 (1981).

RICH, T. H. AND ARCHER, M., 1979. Namilamadeta snideri, anew diprotodontan (Marsupialia, Vombatoidea) fromthe medial Miocene of South Australia. Alcheringa 3:

197-208,

RIcH,T. H., 1982. Monotremes, placentals, and marsupials:the record in Australia and its biases. Pp. 386-477 in "Thefossil vertebrate record of Australasia" ed by P. V. Richand E. M. Thompson. Monash University Offset Print-ing Unit: Clayton.

RIcH,T. H., ARCHER, M., PLANE, M., FLANNERY, T., PLEDGE,N., HAND, S. AND RICH, P., 1982. Australian Tertiarymammal localities. Pp. 525-72 in "The fossil vertebraterecord of Australasia" ed by P. V. Rich and E. M.Thompson. Monash University Offset Printing Unit:

Clayton.

TEDFORD, R. H., ARCHER, M., BARTHOWMAI, A., PLANE, M.D., PLEDGE, N. S., RICH, T. H., RICH, P. V. AND WELi$, R.T., 1977. The discovery of Miocene vertebrates, LakeFrome area, South Australia. Bur. Min. Res. J. Awt. Geol.Geophys. 2: 53-57.

WOODBURNE, M. 0., TEDFORD, R. H., ARCHER, M., TURNBULL,W. D., PLANE, M. D. AND LUNDELIUS, E. L., 1985. Bio-chronology of the continental mammal record of Aust-ralia and New Guinea. Spec. Publ., S. Awt. Dept. Mines andEnergy 8: 347-63.

WOODBURNE, M. 0., TEDFORD, R. H. AND ARCHER, M., 1987.New Miocene ringtail possums (Marsupialia:Pseudocheiridae) from South Australia. Pp. 639-79 in"Possums and opossums: studies in evolution" ed by M.Archer. Surrey Beatty & Sons and the Royal ZoologicalSociety of New South Wales: Sydney.