THE AUDITORY REGION OF EARLY PALEOCENE PUCADELPHYDAE (MAMMALIA, METATHERIA) FROM TIUPAMPA, BOLIVIA, WITH PHYLOGENETIC IMPLICATIONS

Authors


Abstract

Abstract:  New petrosal bones, assigned to Pucadelphys and Andinodelphys, from the Lower Paleocene of Tiupampa, Bolivia, are described. These remains provide new information on the anatomy of the ear region of these taxa. The re-examination of characters from the petrosal and basicranium shed light on the phylogenetic relationships of the three Tiupampan genera known from complete cranial remains (i.e. Mayulestes, Pucadelphys and Andinodelphys). The combination of dental, general cranial and basicranial characters led to two alternative hypotheses. The first is that borhyaenoids (including Mayulestes) are nested within Notometatheria. Pucadelphyds (i.e. Pucadelphys and Andinodelphys) are the sister group of a clade comprising MHNC 8369 (one isolated petrosal from Tiupampa) and Marsupialia. The second favours the paraphyly of ‘borhyaenoids’ (i.e. the exclusion of Mayulestes from borhyaenoids) and the polyphyly of ‘Notometatheria’. In this case, Mayulestes and borhyaenids represent the stem group of a clade including Asiatic, American and Australian metatherians. This analysis of combined datasets (dental, general cranial and basicranial) highlighted contradictory information in the dental and cranial characters, serving to emphasize that in a large anatomical complex like an entire skull mosaic evolution of the characters is likely.

The locality of Tiupampa, Bolivia, has yielded a diverse, well-documented fauna of early Paleocene metatherians and eutherians (Muizon 1998). It represents the oldest Tertiary mammalian assemblage known from South America and therefore sheds new light on our understanding of the origin and early radiations of therians on that continent.

Twelve species of metatherians have been described previously from Tiupampa (Marshall and Muizon 1988; Muizon 1992; Muizon and Cifelli 2001). Of these, three are represented by complete skulls and complete to partial skeletons (Pucadelphys andinus, Andinodelphys cochabambensis, Mayulestes ferox), which are the oldest known American metatherians. These exceptional specimens were reported by Muizon (1994) and the anatomy of the skulls of Pucadelphys and Mayulestes has been thoroughly described by Marshall and Muizon (1995) and Muizon (1998). Although the skull of Andinodelphys is still being studied by one of us (CM), we describe here the petrosal of this taxon and compare it to petrosals of Pucadelphys and Mayulestes. This paper also sheds light on the phylogenetic relationships of the three genera based on dental, cranial and basicranial characters.

It has been proposed that Andinodelphys was the sister taxon of all Australian taxa (Marshall et al. 1990; Woodburne and Case 1996), especially on the basis of the presence of twinned cusps in the C position. However, the features proposed by these authors to support this relationship are highly variable among metatherians (Kirsch et al. 1997; Godthelp et al. 1999). Complete skulls and mandibles of Andinodelphys discovered in 1996 allow a better understanding of this taxon. These specimens, among others, show that a twinned cusp in the C position is not always present in Andinodelphys. Muizon (1992) and Muizon et al. (1997) argued that Andinodelphys shared significant features with Pucadelphys and Mayulestes, especially the presence of a medial process on the squamosal. Because some characters of the molar structure relate both Andinodelphys and Pucadelpys to the didelphimorphians, Andinodelphys was included in the family Pucadelphydae (Muizon 1998).

Here we report new isolated petrosals that were found associated with fossil remains of Pucadelphys and Andinodelphys. A comparative description of their anatomy is made and reveals the proximity of these two genera. It confirms the inclusion of Andinodelphys in the family Pucadelphydae and rejects its close affinities to the Australian marsupials (Marshall et al. 1990; Woodburne and Case 1996).

Material

Isolated petrosals from Tiupampa

Three isolated left petrosals of Pucadelphys andinus are described (MHNC 8364, 8367, 8368 incomplete). Another petrosal was found associated with remains of Pucadelphys andinus (MHNC 8369), but its referral to this taxon needs confirmation as its general structure and morphology differ significantly from the known petrosal bones of Pucadelphys. Previously, Marshall and Muizon (1995) described another isolated petrosal articulated with the squamosal (YPFB Pal 6470) along with several complete skulls (with preserved auditory regions) (YPFB Pal 6105, 6110). The auditory region of two new remarkably complete skulls of Pucadelphys andinus, MHNC 8266 and MHNC 8366, has also been studied.

These new petrosals are compared with two unpublished petrosals of Andinodelphys cochabambensis (MHNC 8370 associated with skull MHNC 8371). Moreover, two new skulls of Andinodelphys cochabambensis are considered (MHNC 8264, 8308).

Comparative fossil and extant taxa

Comparisons are made with other recently described isolated metatherian petrosals: Type I (MNRJ 6726-V, 6727-V) and Type II (MNRJ 6728-V, 6729-V) petrosals (Ladevèze 2004) from the mid Paleocene (Itaboraían) of Brazil (Flynn and Swisher 1995; Marshall et al. 1997). Isolated metatherian petrosals from the Upper Cretaceous of North America are also considered: a probable Pediomys‘Petrosal A’ (FMNH PM53907, Wible 1990) and the stagodontid Didelphodon vorax (UCMP 53896, Clemens 1966). Another relevant taxon is Deltatheridium pretrituberculare (PSS-MAE 132, 133, Rougier et al. 1998) from the Upper Cretaceous of Mongolia, which is represented by an isolated petrosal and skull and jaw remains. These three last taxa are scored after Rougier et al. (1998), Wible et al. (2001) and Luo et al. (2003). Moreover, the Bolivian taxa Mayulestes ferox (MHNC 1249) is included with other Bolivian borhyaenoids (Notogale mitis, MNHN SAL 271, Patterson and Marshall 1978; Sallacyon sallensis, MNHN SAL 92, Villarroel and Marshall 1982; Hoffstetter and Petter 1983) because they are represented by skulls whose ear region is preserved. Scoring for these taxa is mostly based on Muizon (1998), Rougier et al. (1998) and Wible et al. (2001).

Most of the extant taxa considered here are represented by skulls and isolated petrosals. Among South American are six didelphid species: Caluromys philander (CG AC-1880-1780; CG ZMO 17-F 1986-140, -142–144), Didelphis albiventris (CG ZMO 1955-593, 2000-217; RH 120), Didelphis aurita (CG AC 1949-70, 1984-058), Didelphis marsupialis (CG AC-9855, CG ZMO- 1985-1805), Marmosa murina (CG AC 3283, 3310; CG ZMO 2001-2239; RH 82), and Metachirus nudicaudatus (CG ZMO 1988-68, 2001-2175; RH 16); one Caenolestidae, Caenolestes fuliginosus (CG ZMO 1967-1338, 1981-607, 1982-941; RH 85); and the Microbiotheriidae Dromiciops gliroides (CG ZMO-1979-394; FMNH 22675, 134556; IEEUACH 2167, 2162). Among Australian taxa, less derived species by comparison with those from South American have been chosen: one Peramelidae, Perameles nasuta (CG AC 12417; CG ZMO 1982-767); two Peroryctidae: Echymipera rufescens (CG ZMO-1962-2147) and Echymipera kabulu (CG ZMO 1886-1225); three Dasyuridae: Phascogale tapoatafa (CG ZMO-1846-1288, 1897-1492), Dasycercus byrnei (CG ZMO-1987-1489) and Dasyurus hallucatus (CG ZMO-1854-99); and the Thylacinidae Thylacinus cynocephalus (CG AC-1883-352, 1891-61).

Methods

All measurements are in millimetres and were taken with a Wild MMS 235 digital length-measuring system mounted on a stereomicroscope (see Table 1).

Table 1.   Petrosal dimensions of Pucadelphys and Andinodelphys. Measurements were made on isolated petrosals and petrosals included in skulls. Abbreviations: Lfv, length of fenestra vestibuli; Wfv, width of fenestra vestibuli; Lpe, petrosal length; Wpe, petrosal width; * denotes a skull.
SpeciesMaterialLpeWpeLfvWfvStapedial ratio
P. andinus*MHNC 8266 (right)5·96?0·370·301·23
*MHNC 8266 (left)5·64?0·340·241·42
 MHCN 83644·743·360·330·241·38
 MHNC 83674·964·000·450·371·22
 MHNC 8368?2·85???
 MHNC 83694·933·23???
A. cochabambensis*MHNC 8264 (right)8·14?0·410·251·64
*MHNC 8264 (left)8·10????
*MHNC 8308 (right)8·07????
*MHNC 8308 (left)7·78????
 MHNC 83707·684·67???
 MHNC 83716·685·28???

In the following description we use the terminology of several relevant studies (e.g. MacPhee 1981; Wible 1990, 1991; Rougier et al. 1992; Wible and Hopson 1993, 1995, 2001; Luo et al. 2002). We also reconstruct the major vessels and nerves associated with the petrosal, with the help of previous research on the anatomy of extant mammals (MacPhee 1981; Wible 1984, 1986, 1987, 1990, 2003; Novacek 1986, 1993; Rougier et al. 1992; Wible and Hopson 1995).

The complete anatomical description of the reported petrosals allows a definition of anatomical characters for a cladistic treatment. Fifty-five characters from the basicranium were analysed, complemented by 34 general cranial characters and 83 dental characters, using the parsimony method. The complete list of characters is given in the Appendix and the topologies resulting from the parsimony analyses are presented further in the text.

Insotutional abbreviations.
CG AC

Collections d'Anatomie Comparée, Muséum National d'Histoire Naturelle, Paris

CG ZMO

Collections de Mammifères et Oiseaux, Muséum National d'Histoire Naturelle, Paris

FMNH

Field Museum of Natural History, Chicago

IEEUACH

Universidad Austral de Chile, Instituto de Ecología y Evolución, Valdivia

MACN-N

Neuquén collection in Museo Argentino de Ciencias Naturales ‘Bernardino Rivadavia’, Buenos Aires

MHNC

Museo de Historia Natural de Cochabamba (Bolivia)

MNHN

Muséum National d'Histoire Naturelle, Paris

MNRJ

Museu Nacional e Universidade Federal do Rio de Janeiro

PSS-MAE

Palaeontological and Stratigraphical Section of the Geological Institute, Mongolian Academy of Science, Ulaan Baatar

RH

Robert Hoffstetter Collection housed in the MNHN, Paris

UCMP

University of California, Museum of Palaeontology, Berkeley

YPFB Pal

Yacimientos Petrolíferos Fiscales Bolivianos, Colección de Palaeontología, Santa Cruz.

Anatomical abbreviations.
ac

aqueductus cochleae

acf

anterior carotid foramen

AL

alisphenoid

al

anterior lamina

av

aqueductus vestibuli

BO

basioccipital

BS

basisphenoid

cc

crus commune

cr

crista petrosa

ctpp

caudal tympanic process of petrosal

EO

exoccipital

er

epitympanic recess

fai

foramen acousticum inferius

fas

foramen acousticum superius

fc

fenestra cochleae

fi

fossa incudis

fm

foramen magnum

fn

facial nerve

fo

foramen ovale

fsa

fossa subarcuata

fv

fenestra vestibuli

gf

glenoid fossa

gpn

greater petrosal nerve

gpn

greater petrosal nerve

hF

hiatus Fallopii

hff

hypoglossal foramina

iam

internal auditory meatus

ica

internal carotid artery

ijf

inferior jugular foramen

ips

inferior petrosal sinus

lapc

lateral aperture of the prootic canal

lhv

lateral head vein

lw (tt)

lateral wall of epitympanic recess (tuberculum tympani)

lw

lateral wall of epitympanic recess

me

mastoid exposure

mp

mastoid tympanic process

pcv

prootic canal vein

pfc

prefacial commissure

pgf

postglenoid foramen

plf

posterior lacerate foramen

pr

promontorium

ps

prootic sinus

psc

posterior semicircular canal

pt

pterygoid

sff

secondary facial foramen

sica

sulcus for internal carotid artery

sigmoid

sinus

sips

sulcus for the inferior petrosal sinus

smn

stylomastoid notch

SO

supraoccipital

spev

sphenoparietal emissary vein

sps

sulcus for the prootic sinus

SQ

squamosal

ssf

subsquamosal foramen

sss

sulcus for the sigmoid sinus

ssups

sulcus for the superior petrosal sinus

ssups?

probable sulcus for the superior petrosal sinus

sups

superior petrosal sinus

tcf

canal transverse foramen

th

tympanohyal

ts (lt)

tympanic sinus formed in lateral trough

ts

transverse sinus

ttf

tensor tympani fossa

vs

vascular sulcus

vv

vascular vein.

Anatomical description of pucadelphys andinus and andinodelphys cochabambensis

The petrosal bone of recent mammals is an endochondral ossification of the basicranium. This complex bone contains the inner ear (the ossified otic capsule of the chondocranium) and provides attachment area for the muscles and ligaments of the middle ear ossicles.

By convention, the petrosal is divided into two parts (Wible 1990): the anteroventral pars cochlearis, which surrounds the cochlea, and the posterodorsal pars canalicularis, which houses the vestibule and the semicircular canals.

Because the isolated petrosals of Pucadelphys and Andinodelphys are very similar and because the two genera are regarded as closely related, they are described together in a comparative way. The description is made in four views, dorsal, ventral, lateral and medial, with the orientation based on their presumed position in the skull. We illustrate the petrosals MHNC 8364 of Pucadelphys (Text-fig. 1), MHNC 8370 of Andinodelphys (Text-fig. 2) and the auditory region of Andinodelphys (MHNC 8264, Text-fig. 3). The ambiguous isolated petrosal MHNC 8369 is also illustrated (Text-fig. 4), as it may or may not be assigned to Pucadelphys.

Figure TEXT‐FIG. 1..

 Left petrosal of Pucadelphys andinus (MHNC 8364) in A, ventral, B, dorsal, and C, lateral views.

Figure TEXT‐FIG. 2..

 Left petrosal of Andinodelphys cochabambensis (MHNC 8370) in A, ventral, B, dorsal, and C, lateral views.

Figure TEXT‐FIG. 3..

 Right ear region of Andinodelphys cochabambensis (MHNC 8264), with two sides of the petrosal bone exposed, the tympanic side represented by the promontorium, and the mastoid exposure on the occiput.

Figure TEXT‐FIG. 4..

 Right petrosal of MHNC 8369 in A, ventral, B, dorsal, and C, lateral views.

Dorsal side

The cerebellar surface of the petrosal consists principally of a large ovoid body with two large, subequal openings: the fossa subarcuata and the internal acoustic meatus. These openings are separated by a broad osseous wall in Andinodelphys unlike the condition in Pucadelphys, in which the wall is thinner.

Posterodorsally, the large, deep fossa subarcuata accommodates the paraflocculus of the cerebellum. This fossa is bounded by the three semicircular canals, the anterior and the posterior of which join at the crus commune. The anterior edge of the fossa subarcuata is the crista petrosa, which is broad in Andinodelphys (MHNC 8370, 8371), while it is sharp and narrow in Pucadelphys (YPFB Pal 6470, MHNC 8364, 8369, 8368). The crista petrosa forms the posteromedial rim of a large anterodorsally projecting wing. This structure is also present in Pucadelphys and has been regarded by Marshall and Muizon (1995) as a vestige of the anterior lamina of the petrosal. The anterior lamina of the periotic is well developed in non-tribosphenic mammals such as Vincelestes, Morganucodon and multituberculates. In these mammals the anterior lamina is part of the lateral wall of the skull and articulates anteriorly with the alisphenoid and posteriorly with the squamosal. In tribosphenidans, the anterior lamina is generally absent or vestigial, the periotic is excluded from the lateral wall of the braincase, and the alisphenoid extends posteriorly and articulates with the squamosal. In Andinodelphys (MHNC 8308, 8264) and Pucadelphys (YPFB Pal 6105, 6110, MHCN 8266) the anterior lamina is excluded from the lateral wall of the skull but it is still well developed and covered laterally by the squamosal and the alisphenoid (Marshall and Muizon 1995 for Pucadelphys). Andinodelphys and Pucadelphys show a peculiar structure in which the anterior lamina contributes to the internal wall of the braincase, whereas the external wall is formed by the alisphenoid and squamosal (Marshall and Muizon 1995).

The dorsal side of the anterior lamina of Andinodelphys shows a large fossa (= cavum epiptericum) for the trigeminal ganglion of the trigeminal nerve (V) (at least its posterior part) and possibly part of the temporal lobe of the cerebrum. The condition of Pucadelphys is identical in this respect to that of Andinodelphys.

In Andinodelphys (MHNC 8264, 8308) and Pucadelphys (YPFB Pal 6105, MHNC 8266, 8366) the uncovered portion of the anterior lamina, visible in ventral view, forms the posterior rim of the foramen ovale through which exits the mandibular branch (V3) of the trigeminal nerve (Text-fig. 2). In Andinodelphys (MHNC 8264, 8308), as in Pucadelphys (YPFB Pal 6105, 6110, MHCN 8266), the foramen ovale is a simple perforation of the basicranium, delimited anteriorly by the alisphenoid and posteriorly by the anterior edge of the petrosal. This condition (i.e. foramen ovale between alisphenoid and petrosal, without secondary foramen ovale) is plesiomorphic in metatherians, and is also observed in Mayulestes (MHNC 1249; Muizon 1998). A notch observed on the anterior lamina, especially in MHNC 8308 (skull of Andinodelphys) and in MHNC 8367 (isolated petrosal of Pucadelphys), is likely to reflect the passage of the trigeminal nerve between petrosal and alisphenoid, i.e. the posterior edge of the foramen ovale.

Posterior to the dorsal edge of the fossa subarcuata, a short sulcus for the sigmoid sinus is present on Andinodelphys (MHNC 8371), Pucadelphys (MHNC 8364, 8367) and the petrosal MHNC 8369. Anteromedially, the sulcus extends medially up to the aqueductus vestibuli; posteromedially it does not reach the posterolateral angle of the edge of the fossa subarcuata. This sulcus is not present in the petrosal MHNC 8368 (Pucadelphys), and is poorly marked on MHNC 8370 (petrosal of Andinodelphys), but this condition could be owing to an artefact of fossilization. The sigmoid sinus is a branch of the transverse sinus (see below), which becomes the prootic sinus when it enters the lateral side of the periotic after emitting the superior sinus and the sigmoid sinus. On one petrosal of Pucadelphys (YPFB Pal 6470), a tiny medially opening foramen occurs in the deepest dorsal part of the sigmoid sinus and extends laterally. This foramen apparently connects with the canal from the small foramen in the posterodorsal end of the sulcus for the prootic sinus. No such foramina and canal are distinctly observed in Andinodelphys, but a similar canal transmitting a branch from the prootic sinus to the sigmoid sinus is observed on two petrosals from Itaboraí (Type I, Ladevèze 2004), and on the petrosals of Monodelphis (Wible 2003), Caenolestes and Phascogale. In Didelphis the connection between the prootic sinus and the sigmoid sinus is not enclosed in a canal and the posterior edge of the subarcuate fossa is surrounded by a continuous sulcus from the sulcus for the sigmoid sinus to that for the prootic sinus. Therefore, the sulcus for the prootic sinus runs along the edge of the fossa subarcuata. The absence of foramina on the dorsal extremity of the sulci for the prootic and sigmoid sinuses is consistent with the fact that the sulcus for the prootic sinus intersects abruptly with the anterolateral edge of the fossa subarcuata. The sulcus is, therefore, always anterior to the posterior edge of the fossa subarcuata. A similar condition is observed on the four Pucadelphys petrosals analysed in this study and in Itaboraí Type II of Ladevèze (2004), although intersection is more tangential in these taxa. In recent didelphids and in Itaboraí Type I of Ladevèze (2004) the sulcus for the prootic sinus does not intersect with the edge of the fossa subarcuata but runs posteriorly along the posterior edge of the fossa. Therefore, the junction of the prootic sinus with the sigmoid sinus occurs more anteriorly in Andinodelphys and Pucadelphys (anterior to the posterior edge of the fossa subarcuata) than in the extant didelphids, in which it is posterior to the fossa subarcuata. In peramelids, the sulci are even more posterior than in didelphids.

Marshall and Muizon (1995) described a small foramen just medial to the foramen for the sigmoid sinus vein on the Pucadelphys petrosal YPFB Pal 6470. They interpreted it as the cerebellar opening of the mastoid foramen for the occipital emissary vein. Nonetheless, neither a foramen containing the occipital emissary vein nor a foramen connecting the sigmoid sinus to the prootic sinus is present on the isolated petrosals of Andinodelphys (MHNC 8370, 8371), MHNC 8369 and Pucadelphys (MHNC 8367, 8368), except on MHNC 8364 in which a small foramen occurs on the sulcus for the prootic sinus. However, the latter might also be interpreted as a vascular foramen.

Andinodelphys (MHNC 8370, 8371) exhibits a distinct sigmoid sulcus on the anterior edge of the subarcuate fossa that runs on the crista petrosa (the ridge that separates the subarcuate fossa from the anterior lamina), anteroventrally from the dorsal extremity of the sulcus for the prootic sinus, almost reaching anteriorly the posterior edge of the hiatus Fallopii. It receives dorsally the superior sinus, another ramification from the transverse sinus. Intersection of the two sulci is at a very sharp angle on the crest, which forms the anterolateral angle of the subarcuate fossa. The condition of Andinodelphys is also present in Pucadelphys (MHNC 8364) in which the sulcus for the superior sinus is also very well marked. The sulcus for the superior sinus is absent in Didelphis and very short and weak in Monodelphis (Wible 2003). It is present in most other didelphids, although generally less pronounced than in Andinodelphys and Pucadelphys. It is pronounced in Perameles.

The crista petrosa separates the subarcuate fossa from the anterior lamina and runs anteromedially almost until the apex of the periotic. The tentorium cerebelli is attached to the crista petrosa, and separates the cerebrum and cerebellum. In Andinodelphys the crista petrosa is massive and rounded, but relatively low. It is mainly emphasized by the depth of the fossa for the trigeminal ganglion in the anterior lamina. It is irregular because of the inflated area between the posterolateral angle of the internal acoustic meatus and the anterolateral angle of the subarcuate fossa. The crista petrosa is sharper and straighter in Pucadelphys, where this area is not inflated. In this respect, Andinodelphys differs from the Itaboraían petrosal Types I and II, which have a straight, sharp crista petrosa, but resembles the condition in Didelphis in which a similar inflated area is present. In the other didelphids, the crista petrosa is generally weak but no strongly inflated area is present between the subarcuate fossa and the internal acoustic meatus.

Anteroventral to the inflated area, at the anteromedial angle of the anterior lamina and at the anterior third of the crista petrosa, is a large foramen, the hiatus Fallopii, which transmits the greater petrosal nerve, the palatine branch of the facial nerve. The aperture for the greater petrosal nerve is delimited ventrally and posteriorly by shelves of bone from the lateral trough and crista petrosa, respectively. The floor and roof of the cavum epiptericum are of interest in delineating the position of the hiatus Fallopii (Sánchez-Villagra and Wible 2002). In Andinodelphys, the floor of the cavum epiptericum does extend further to the anterior than the roof, so that the hiatus Fallopii is considered to have a dorsal position. Furthermore, the hiatus Fallopii is located well posterior to the anterior edge of the anterior lamina. Pucadelphys exhibits a similar condition, i.e. dorsal tympanic aperture of the hiatus Fallopii, although the hiatus is slightly more anterior than in Andinodelphys. A dorsal position of the hiatus Fallopii occurs in Types I and II of Itaboraí (Ladevèze 2004). In extant didelphids the anterior lamina has totally disappeared and the hiatus opens on the anteromedial edge of the bone (intermediate position of Sánchez-Villagra and Wible 2002). However, the ventral lip of the hiatus is a thin bony wall, which is distinctly posterior to the thick dorsal edge. It seems therefore that the hiatus opens slightly more on the ventral side of the petrosal than on the dorsal side. A condition similar to that of the didelphids is present on petrosals A and B of Wible (1990) from the Upper Cretaceous of North America. In Perameles and Echymipera, the hiatus Fallopii is located on the ventral side of the petrosal. The condition in Deltatheridium is unknown (Rougier et al. 1998).

Anteromedial to the fossa subarcuata, the internal acoustic meatus forms a shallow depression for the facial and vestibulocochlear nerves (Wible 1990). The internal acoustic meatus is subequal to the fossa subarcuata and exhibits a broad transverse septum in Pucadelphys, whereas it is narrower than the fossa subarcuata in Andinodelphys. The floor of the internal acoustic meatus has a rough dumb-bell-shaped depression. The larger and medial aperture of this depression, the foramen acusticum inferius, shows some tiny perforations that are interpreted as evidence of the spiral cribriform tract or tractus spiralis foraminosus, which probably transmitted the fascicles of the cochlear nerve, as in other therians (Meng and Fox 1995a, b). In the posterior part of the foramen acusticum inferius is a large, shallow pit that may be interpreted as the foramen singulare for passage of vestibular nerve bundles. The smaller lateral aperture, the foramen acusticum superius hidden in dorsal view, has a small anterior opening for passage of the facial nerve and a posterior pit, which is interpreted as the cribriform dorsal vestibular area for the passage of the remaining bundles of the vestibular nerve. The tiny perforations of the cribriform tract are not visible on the two specimens of Andinodelphys because of the coarseness of the matrix, the grains being much larger than the perforations. The prefacial commissure is the anterolateral edge of the internal acoustic meatus. It is located on the anterior part of the crista petrosa and overhangs the fossa for the trigeminal ganglion in the anterior lamina. In Andinodelphys the prefacial commissure is deeply notched. No such condition is present in Pucadelphys, Itaborai petrosals I and II and recent didelphids. However, a similar notch exists in Perameles and, to a lesser extent, in Trichosurus and Phalanger.

The aqueductus vestibule for the passage of the endolymphatic duct is a small slit posteromedial to the fossa subarcuata and near the crus commune. The aqueductus cochleae for the passage of the perilymphatic duct is a small, circular foramen located in a conical depression posteromedial to the internal auditory meatus and opens into the posterior lacerate foramen (or jugular foramen). Ventral to the aqueductus cochleae is a distinct notch on the posteromedial edge of the promontorium and dorsal to the fenestra cochleae. This notch is the anterolateral edge of the posterior lacerate foramen (sensuArcher 1976a) or jugular foramen (see Wible 2003, p. 176). In most extant marsupials, the jugular foramen conducts cranial nerves IX, X and XI (Archer 1976a) and the inferior petrosal sinus and lateral head vein. In both Pucadelphys and Andinodelphys, the opening for the inferior petrosal sinus is separated from the jugular foramen (see Text-fig. 3, but the separation is not obvious as the skull was deformed, and some cranial elements are slightly displaced).

Ventral side

This view shows the two main divisions of the petrosal. The pars cochlearis is represented by the large teardrop-shaped and inflated promontorium. The pars canalicularis is lateral and posterior to the promontorium and includes the canal for the facial nerve.

The promontorium is broadest posteriorly and its ventral surface is nearly smooth. The minor topographic variations on its surface reflect the underlying turns of the cochlear duct. On the anterolateral part of the promontorium of Pucadelphys and Andinodelphys is a pronounced fossa for the tensor tympani muscle. The width and depth of the tensor tympani fossa forms the battered surface of the promontorium of both taxa. On the Pucadelphys petrosals YPFB Pal 6470, MHNC 8367 and especially MHNC 8369 (Text-fig. 4), a broad, shallow depression lies medial to the large fossa for the tensor tympanic muscle and indicates the path of the internal carotid artery towards the entocarotid foramen, whereas it is incipient on the petrosal MHNC 8364.

The sulcus for the internal carotid artery is pronounced on the petrosals of Andinodelphys (MHNC 8370, 19). The fossa for the tensor tympani muscle and the sulcus for the internal carotid artery are separated, when present, by a broad shelf of bone. On the skulls of both Pucadelphys (YPFB Pal 6105, 6110, MHCN 8266) and Andinodelphys (MHNC 8264, 8308), the sulcus for the internal carotid artery lies between the promontorium and the large foramen ovale, runs anterior and parallel to the basioccipital suture, and crosses onto the alisphenoid just lateral to the medial lacerate foramen, to the posterior edge of the entocarotid foramen. The medial expansion of the promontorium forms a flat shelf, the epitympanic wing (MacPhee 1981; Wible et al. 2001), which contacts the basioccipital in both skulls, as in extant marsupials.

Two large apertures open into the promontorium. The posterolateral opening is the fenestra vestibuli, which in life was closed by the footplate of the stapes. Marshall and Muizon (1995) reported a stapedial ratio (sensuSegall 1970) of 1·4 for Pucadelphys. The same value is calculated here for the new specimens of Pucadelphys (MHNC 8364, 8266, 8367; Table 1). However, as regards to some specimens (P. andinus MHNC 8369; A. cochabambensis MHNC 8308, 8370, 8371) the estimation of the stapedial ratio is irrelevant because of the crushing of the promontorium. The posteromedial opening, the fenestra cochleae, is kidney-shaped and anteroposteriorly elongate. This aperture was in life closed by the secondary tympanic membrane. The fenestra cochleae is surmounted by a prominent expansion of the promontorium, which forms an anteromedially directed bump that makes a separation between the fenestra cochleae and the aqueductus cochleae. This feature is also present in the skulls and petrosals of Pucadelphys (MHNC 8266, 8367) and Andinodelphys (MHNC 8264, 8308, 8370, 8371).

Lateral to the fenestra vestibuli is a large depression, the epitympanic recess, which in life was dorsal to the tympanic membrane and housed the articulation between malleus and incus. The epitympanic recess is an oblique, slightly concave depression in the roof of the tympanic cavity. Its posteromedial extremity forms the fossa incudis, which is bordered posterolaterally by the squamosal. The epitympanic recess is bounded anteriorly by an oblique (anterolateral–posteromedial) ridge called the petrosal crest (sensuArcher 1976a; Muizon 1999). In the skulls of Pucadelphys, Andinodelphys, Mayulestes and most borhyaenids (Muizon 1998, 1999), and in extant didelphids, the petrosal crest lies medial to the triple point petrosal-squamosal-alisphenoid and lateral to the small crest flooring the medial opening of the prootic canal. In extant didelphids, the petrosal crest also bounds the posterolateral border of the alisphenoid sinus (absent in Pucadelphys and Andinodelphys).

On the petrosals of Pucadelphys, lateral to the epitympanic recess, is a massive, rounded wall, called here the lateral wall of the epitympanic recess (even if the ‘true’ lateral wall of the epitympanic recess is formed by the squamosal, as seen in the skulls). The isolated petrosals of Andinodelphys exhibit a quite different condition, in which the lateral wall is massive, elongate and comprises the large ‘lateral trough’, which borders the promontorium laterally.

In Andinodelphys, the lateral aspect of the pars canalicularis is a well-developed shelf of bone that reduces in width anteriorly; it can be interpreted as a vestigial lateral trough. The lateral trough is a structure found in most non-therian mammals that is reduced in size in therians and floors the geniculate ganglion (Wible 1990). In Pucadelphys, the vestigial lateral trough is less developed but present.

In the ventral view of the skull MHNC 8264 and petrosal MHNC 8371 of Andinodelphys is a small depression anterior to the petrosal crest and epitympanic recess, and lateral to the promontorium. This may correspond to a petrosal hypotympanic sinus, but it has no connection with the alisphenoid as this bone does not have a hypotympanic sinus. However, the isolated petrosal MHNC 8370 of Andinodelphys shows an expansion, but no petrosal hypotympanic sinus. The petrosals MHNC 8364 and 8367 of Pucadelphys exhibit a distinct hypotympanic sinus; nonetheless, no pronounced petrosal hypotympanic sinus was observed on the skulls YPFB Pal 6105, 6110, MHNC 8266 (Pucadelphys) and 8308 (Andinodelphys).

On the ventral side of the pars canalicularis, a thin groove indicates the location of the suture between the petrosal and the squamosal, which articulates with the most lateral part of this surface. Medial to this groove, a large surface faces the posterior side of the skull. This is the mastoid exposure, and is located between the squamosal, supraoccipital and exoccipital. In the intact skulls, the mastoid exposure appears quite large. Lateral to it is a ventrally directed process, the mastoid tympanic process (sensuArcher 1976a; Wible 1990), which is a large, slanted process in Andinodelphys, but is less developed in Pucadelphys. The mastoid tympanic process forms the posteromedial wall of the stylomastoid notch. The latter is bounded laterally by a tiny crest, the tympanohyal, and medially by an anteriorly directed lip, the caudal tympanic process of the petrosal (sensuMacPhee 1981), which extends medially from the stylomastoid notch to the large posterior lacerate foramen, and lies lateral to the exoccipital, whose imprint is visible on the medial face of the pars canalicularis.

Facial nerve

The facial nerve leaves the cranial cavity by the internal acoustic meatus, runs beneath the prefacial commissure and enters into the cavum supracochleare by the primary facial foramen. The cavum supracocleare, which encloses the geniculate ganglion of the facial nerve, is ventrally closed by an osseous floor.

The main branch or hyomandibular ramus of the facial nerve leaves the posterior aspect of the geniculate ganglion and enters the middle ear space via the secondary facial foramen. The latter is not visible on Pucadelphys petrosals MHNC 8364, 8369 and 8368 as the facial canal is broken. On the other petrosals of Pucadelphys (MHNC 8367, YPFB Pal 6110, 6470), the secondary facial foramen occurs just anterior to the dorsal edge of the fenestra vestibuli and opens posteriorly, as in the petrosals of Andinodelphys. The posteroventral branch of the facial nerve runs on the deep sulcus facialis and exits the skull by the stylomastoid notch, which forms a narrow indentation. On the dorsomedial area of the sulcus facialis, the fossa for the stapedial muscle is deep. The large postpromontorial tympanic sinus (sensuWible 1990) extends medial to the stapedial fossa and posterolateral to the promontorium.

The greater petrosal nerve or palatine ramus of the facial nerve leaves the anterior part of the geniculate ganglion and exits the middle ear by the hiatus Fallopii, which is closed by the alisphenoid bone in the intact skull. The hiatus Fallopii is located dorsally on the anterolateral part of the petrosal and opens between the petrosal and the alisphenoid. A fossa, which probably housed a part of the trigeminal ganglion, is visible anterior to the hiatus Fallopii. This feature indicates that the cavum epiptericum was probably floored by the petrosal and alisphenoid as in some metatherians (Deltatheridium, Mayulestes; Wible et al. 2001).

Arteries and veins

The complete reconstruction of the vascular system in fossil skulls or petrosals depends on the impressions left on the bone by the cranial vessels (such as grooves, canals and foramina), which makes it extremely difficult in some cases. On the petrosals of Pucadelphys and Andinodelphys studied here most of the grooves, sulci and vascular canals are similar in size and position to those observed on the petrosals of extant didelphids. Therefore, the vascular reconstruction will be inferred from didelphid anatomy (see Wible 1990, p. 191, fig. 4; Wible and Hopson 1995, p. 342, fig. 5). In describing the vascular system of the petrosals, we follow Wible's (1990) terminology.

Lateral head vein and prootic sinus.  The prootic canal leads a vein from the prootic sinus to the lateral head vein. The cranial aperture of the prootic canal is posterodorsal to the lateral wall of the epitympanic recess and emerges within the sulcus of the prootic sinus, between the pars canalicularis and the squamosal. The tympanic opening of the prootic canal is anterodorsal to the petrosal crest and lateral to the presumed position of the secondary facial foramen.

The lateral head vein runs posteriorly to the geniculate ganglion of the facial nerve and to the otic capsule. It meets the facial nerve dorsal to the crista parotica, runs posteriorly in a groove lateral to the sulcus facialis, and then exits via the posterior lacerate foramen with the inferior petrosal sinus.

In extant marsupials, the prootic sinus gives rise to the sphenoparietal emissary vein at the level of the lateral aperture of the prootic canal. Subsequent to the development of this vein, the tympanic portion of the lateral head vein tends to reduce in size in all marsupials (Wible 1990). The lateral head vein and the prootic canal are only retained in adult monotremes, the eutherian Prokennalestes, the metatherians Deltatheridium, Didelphodon and Pediomys, isolated metatherian petrosals from Itaboraí (Types I and II of Ladevèze 2004), some extant marsupials (didelphids, caenolestids and some dasyurids; Wible 1990), and Pucadelphys and Andinodelphys.

Transverse sinus.  In extant marsupials, the transverse sinus runs down the posterolateral extremity of the petrosal, where it divides into two branches. The anteroventral branch exits the skull via the postglenoid canal (prootic sinus and sphenoparietal emissary vein). The posterodorsal branch or sigmoid sinus lies lateral and posterior to the fossa subarcuata. In Pucadelphys (YPFB Pal 6470, MHNC 8364, 8367, 8368) and Andinodelphys (MHNC 8370, 8371), the groove for the sigmoid sinus starts posterior to the fossa subarcuata and ends near the aqueductus vestibuli, thus suggesting that the sigmoid sinus exits by the foramen magnum, as in monotremes, Prokennalestes, and all adult metatherians (Archer 1976a; Wible 1990; Wible et al. 2001).

Inferior petrosal sinus.  On the petrosals of Pucadelphys (YPFB Pal 6470, MHNC 8364, 8367, 8368) and Andinodelphys (MHNC 8370, 8371), the inferior petrosal sinus runs in a large, deep groove on the anteromedial region of the petrosal seen in dorsal view, towards the inferior petrosal foramen (= opening for inferior petrosal sinus, sensuWible et al. 2001). The latter forms a tiny opening just anteromedial to the large posterior lacerate foramen (= jugular foramen) (YPFB Pal 6105, 6110) (Text-fig. 3). In extant mammals, the inferior petrosal sinus guides the cavernous sinus blood to the internal jugular vein, along the petrosal-basioccipital suture (Rougier et al. 1996b).

Diploetic vessels.  In the petrosal YPFB Pal 6470 of Pucadelphys a tiny sulcus extends posteriorly from the sulcus for the sphenoparietal emissary vein along the ventral edge of the pars canalicularis, then bends posterodorsally. This sulcus apparently transmitted the diploetic vessels from the area dorsal to the postglenoid foramen to the post-temporal foramen (sensuWible 1990; Wible et al. 1995). On the skulls YPFB Pal 6105 and 6110 of Pucadelphys, a small post-temporal foramen occurs on the lateral surface of the occiput between the squamosal and the mastoid exposure of the petrosal. The occurrence of a post-temporal canal is, therefore, variable among the Pucadelphys specimens. Andinodelphys apparently did not have this diploetic pattern.

Lateral and medial sides

The groove for the prootic sinus is large and deep, and is visible on the lateral side. It starts in the posterolateral angle of the pars mastoidea, anterior to the fossa subarcuata, and borders posterolaterally the anterior lamina of the petrosal, before joining the lateral aperture of the prootic canal. At the anterior edge of the anterior lamina lies the hiatus Fallopii for the passage of the greater petrosal nerve. On the medial side, the articular zone for the exoccipital appears very large and runs along the whole pars canalicularis.

Phylogenetic analysis

Fifty-five characters of the petrosal and basicranium, combined with 34 additional cranial characters and 83 dental characters, were sampled among the Early Paleocene metatherians from Tiupampa, Pucadelphys andinus (with specimen MHNC 8369 considered as a terminal taxon), Andinodelphys cochabambensis and Mayulestes ferox; Notogale mitis and Sallacyon hoffstetteri from the Oligocene of Sallas (Bolivia); the Middle Paleocene metatherians from Itaboraí, Types I and II (Ladevèze 2004); the ten living genera of marsupials mentioned above; and the following fossils: Deltatheridium pretrituberculare, ‘Petrosal A’ (a possible Pediomys; Wible, 1990), and Didelphodon vorax. The outgroup is composed of the Early Cretaceous prototribosphenidan Vincelestes neuquenianus from Argentina (MACN-N01, N04-07, N09, N16-19, N21; Rougier and Bonaparte 1988; Rougier et al. 1992; Hopson and Rougier 1993) and the isolated petrosal from the Lower Cretaceous of Mongolia, referred to the eutherian Prokennalestes trofimovi (PSS-MAE 136; Wible et al. 2001). Moreover, two extant genera of Monotremata were included as outgroup taxa: Ornithorhynchus anaticus (CG ZMO-1962–2146, 1963–2145, 1985-1795, CG AC-7128, 1933-222) and Tachyglossus aculeatus (CG ZMO-1962-2144, 1985-1796, CG AC-12452).

The list of characters is given in the Appendix along with the distribution of the character states among the taxa in a taxon/character matrix. Coding for Didelphodon and Pediomys follows that of Rougier et al. (1998) and Wible et al. (2001) for the cranial characters. In case of different coding, a discussion is provided in the list of characters (Appendix).

The taxon/character states matrix was analysed using heuristic parsimony searches implemented by PAUP* 4.0b10 (Swofford 2002). Each heuristic parsimony search employed 1000 replicates of random taxon addition with TBR branch swapping.

All multistate characters were treated as unordered transformation series, and polymorphic taxa were coded with multiple character state entries. All multistate characters are considered unordered, so as to minimize a priori hypotheses, and because there is no argument in favour of ordering many of the characters studied. The polarization of the character states is based on the outgroup comparison criterion. In the following description of the distribution of character states on the cladograms (Text-figs 5–6), we focus on the unambiguous synapomorphies. Nonetheless, in case of alternative optimization of characters for the same topology, both fast (DELTRAN) and low (ACCTRAN) optimizations are considered and debated if necessary.

Figure TEXT‐FIG. 5..

 Strict consensus tree of the two parsimonious trees (L = 528, CI = 0·44, RI = 0·59). Nodes are named and described in the text. The Bremer index is given on the branches, followed by the number of unambiguous and non-homoplasic synapomorphies in brackets. The Bremer index mean is 2·8 (54/19). Partitioned Bremer values are represented by vertical bars: white, dental dataset; grey, general cranial dataset; black, basicranial dataset.

Figure TEXT‐FIG. 6..

 The two equally parsimonious trees (A, first topology; B, second topology) resulting from our analysis (L = 503, CI = 0·46, RI = 0·62). Certain nodes are named and described in the text. The Bremer index is given on the branches, followed by the number of unambiguous and non-homoplasic synapomorphies in brackets. The Bremer index mean is 2·8 (54/19).

Two equally parsimonious trees were obtained (L = 503, CI = 0·46, RI = 0·62) and the strict consensus tree is given in Text-figure 5 (L = 528, CI = 0·44, RI = 0·59). The robustness of each node is estimated by a Bremer support index (Bremer 1988) and the number of unambiguous and non-homoplasic synapomorphies. A complete list of the distribution of the character states among each node of the strict consensus tree and of the resolved nodes obtained in the two equally parsimonious trees is available in a Supplementary Data file at http://www.palass.org. The relationships mostly agree with current views of mammal phylogeny (Rougier et al. 1998; Wible et al. 2001; Luo et al. 2002, 2003; Horovitz and Sánchez-Villagra 2003). Among Metatheria, Deltatheridium is nested within all other metatherians, and the emergence of Notometatheria (a taxon created by Kirsch et al. 1997 that only includes all South American and Australian metatherians) from Asiatic and North American metatherians is not confirmed in the strict consensus tree as all taxa are in polytomy. The two topologies (Text-fig. 6) indicate two alternative hypotheses concerning the phylogenetic status of Notometatheria and borhyaenoids. The phylogenetic status of the ‘borhyaenoids’Mayulestes, Notogale and Sallacyon is not resolved as they could be either a monophyletic group nested within all other notometatherians studied here or a paraphyletic group diverging from Deltatherium and sister taxa of all other metatherians. The hypothesis that Asiatic and North American taxa are nested with Notometatheria agrees with current phylogenetic data (e.g. Kirsch et al. 1997; Rougier et al. 1998; Wible et al. 2001; Luo et al. 2003); thus, the first topology is preferentially discussed here because it suggests the monophyly of Notometatheria, among which the borhyaenoids appear monophyletic. The clade Pucadelphydae (i.e. Pucadelphys and Andinodelphys) is nested within a clade comprising the isolated petrosal morphotypes MHNC 8369 (which cannot be assigned to Pucadelphys) and Type I from Itaboraí, Didelphidae, Caenolestes and Australidelphia. Among extant metatherians, Caenolestes is the sister taxon of Australidelphia and Dromiciops is nested with dasyurids.

In order to determine the impact of each dataset (dental, cranial, basicranial) on the topology, we conducted a partitioned Bremer support analysis (PBS; Baker and DeSalle 1997) under TreeRot version 2 (Sorenson 1999), the result of which is provided in the Supplementary Data file noted above.

An ILD test (Farris et al. 1994) run under Winclada v.1.00.08 (Nixon 2002) and Nona v.2.0 (Goloboff 1993) concluded that the three datasets were not significantly incongruent (P = 0·2549), whereas the PBS revealed a relative incongruence between dental data and data from the basicranium and/or total skull on 65 per cent of the nodes (11/17 nodes of the strict consensus tree; see Text-fig. 5).

Node A: Metatheria

The Metatheria clade is supported by 25 unambiguous synapomorphies, ten of which are non-homoplasic, and a low Bremer support (value of 2 for a mean of 2·8). The dental formula consists of three premolars (10>2), four molars (30>1), and seven postcanine teeth (50>1). All metatherians share an angular process medially inflected (870>), a facial process of the premaxilla reaching the nasal (910>1), a trough-like glenoid fossa (1070>1), the absence of an ascending canal (1140>1), the caudal tympanic process of petrosal (1400>1), a reduced prootic canal with intramural opening (1480>1), and the absence of a transpromontorial sulcus for the internal carotid artery (1540>1).

Fifteen synapomorphies supporting this clade are homoplasic (see Supplementary Data file). A strongly developed postmetacrista, with enlarged paraconid and reduced metaconid on lower molars, reflects an adaptation to carnivory (Cifelli 1993; Muizon and Lange-Badré 1997) (190>1, reversal in Asiatherium, Pediomys, and node D or node C). The first upper premolar is procumbent and separated by diastema (110>1, reversal in australidephians; 120>1, reversal in Asiatherium, Caenolestes and australidelphians). The staggered lower incisor, as defined by Hershkovitz (1982), is present in all metatherians, and lost in Caenolestes and Dromiciops (460>1). The cristids on the trigonid are bladed and sharp (810>1, reversal in many metatherians, see Supplementary Data file) and the paracristid is longer than the protocristid (800>1, subequal in Mayulestes, pucadelphyds, peramelomorphes; protocristid longer in Caenolestes), reflecting an open shape of the trigonid.

Cranial synapomorphies of Metatheria are: coronoid facet absent (880>1, convergent with monotremes), palatal process of premaxilla reaching to, or nearly to, canine alveolus (900>1, reversal in Dromiciops and peramelomorphes), large minor palatine (postpalatine) foramen (1020>1, reversal in Caenolestes plus Australidelphia, but large in Perameles).

The anterior lamina of the petrosal of therians is greatly reduced to absent (Wible 1990) (1230>2), but remains rudimentary in Pucadelphys where it contributes to the internal wall of the braincase, whereas the external wall is formed by the alisphenoid and squamosal (Marshall and Muizon 1995).

In all metatherians the lateral flange is either reduced or absent (1310>1, convergent with Tachyglossus), although it is important to note that Andinodelphys, Mayulestes, Notogale and Sallacyon show a large expansion of the lateral border of the promontorium, which is interpreted as a relic of the large lateral trough of Mesozoic mammals.

A complete wall separates the cavum supracochleare and cavum epiptericum (1180>1). In non-tribosphenic mammals the posterior part of the cavum epiptericum houses the geniculate ganglion, just behind the trigeminal ganglion (Wible 1990). In the echidna and most extant therians, an osseous wall, probably derived from the prefacial commissure, encloses the geniculate ganglion in the petrosal and creates a cavum supracochleare, separated from the back of the cavum epiptericum (Wible 1990). In the present phylogenetic hypothesis, this feature is acquired in the Metatheria clade and then independently lost in Caenolestes and the Type II from Itaboraí. Among extant metatherians, a hiatus semilunaris occurs independently in caenolestids and some marmosine didelphids (Wible 1990).

The inferior petrosal sinus is located between petrosal, basisphenoid and basioccipital (1370>1) but is intrapetrosal in Didelphodon. The stapedial artery is absent in adults (1550>1, convergent with Tachyglossus). The jugular foramen is separated from the opening for the inferior petrosal sinus (1720>1), but both openings are confluent in Caenolestes, Dromiciops and Metachirus.

The absence of an orbitotemporal canal may be regarded as a synapomorphy of metatherians (1050>1; according to Rougier et al. 1998) or of node B (Metatheria except Deltatheridium).

Four characters from the auditory area may be optimized as metatherian synapomorphies, or synapomorphies of all metatherians except Deltatheridium: mastoid tympanic process small, slanted, node-like, and lies on the posterolateral border of the stylomastoid notch and is continuous with squamosal (1390>1, reversal in Didelphodon and Pediomys, indistinct to absent in node E); presence of an alisphenoid tympanic process that encloses a hypotympanic sinus (1590>1, 1600>1, reversal in Mayulestes and pucadelphyds); presence of a tympanic cavity that is partially or entirely enclosed by bony structure and forms an auditory bulla (1660>1, reversal in Mayulestes and pucadelphyds).

Node B: Metatheria without Deltatheridium

This node is supported by 26 unambiguous synapomorphies, 24 of which are homoplasic, and a strong Bremer support (4 for a mean of 2·8).

This clade is that defined by Rougier et al. (1998) containing Notometatheria plus North American and Asiatic metatherians (i.e. excluding Deltatheroidea). It is supported by the same previously described synapomorphies: hypoconulid lingually displaced, being closer to entoconid (611>2); conules strong with cristae (351>2); procumbent protocone (380>1); labial postcingulid present (700>1). Moreover, this clade is supported by the presence of a postglenoid foramen (1110>1).

All metatherians except Deltatheridium pretrituberculare share a cavum epiptericum floored primarily or exclusively by alisphenoid (1191>2). The composition of the floor of the cavum epiptericum, including the fossa for the trigeminal ganglion, varies among Mesozoic mammaliaforms (Wible and Hopson 1993; Luo 1994). Among the taxa considered by Rougier et al. (1998), the cavum epiptericum is floored by the petrosal only in Vincelestes. Prokennalestes is scored as possessing a floor for the posterior part of the trigeminal fossa composed of both alisphenoid and petrosal. In metatherians, the floor is formed by the alisphenoid, either alone or in concert with the petrosal. As for Mayulestes (scored 1), Andinodelphys has a large expansion of the petrosal that is reminiscent of the lateral trough of Mesozoic mammals, which probably floored the posterior part of trigeminal nerve (state 1, contra Rougier et al. 1998; Wible et al. 2001). However, the isolated petrosals of Pucadelphys and MHNC 8369 do not show any similar structure. Thus, the score for Andinodelphys and Pucadelphys, determined by the conditions observed on both skulls and isolated petrosals, is polymorphic here (1 and 2).

Certain synapomorphies supporting this clade B were previously regarded as notometatherian synapomorphies by Rougier et al. (1998). The metacone lies lingual to the paracone (291>2). Character 28 (i.e. metacone size relative to that of paracone) was reviewed and redefined from Rougier et al. (1998) (see Appendix). Deltatheridium is scored as having a metacone subequal to the paracone (1) and most of the metatherians considered in this study have a metacone that is noticeably larger than the paracone (2). Thus, and contrary to Rougier et al. (1998), a metacone slightly smaller than the paracone is not a synapomorphy of metatherians in our phylogenetic hypothesis. Moreover, a metacone larger than the paracone on the second upper molar (281>2) is a synapomorphy of metatherians except Deltatheridium, and not of Notometatheria.

Character 171 (i.e. jugular foramen size) was quantified relative to the size of the fenestra cochleae, and the consequent coding differs from that of Rougier et al. (1998) for Didelphis (scored 1). A very large jugular foramen (1710>1) appears here as an unambiguous synapomorphy of metatherians except Deltatheridium, even if it could have been regarded as a synapomorphy of Notometatheria (node T of the tree 1) as it is unknown in Asiatherium, the Gurlin Tsav Skull, Didelphodon and Pediomys (consecutively nested with Notometatheria), and apomorphic in borhyaenoids and pucadelphyds (first diverging among Notometatheria). Other tested parsimony softwares (i.e. PAUP version 3 and Nona) propose a slow optimization of the appearance of the apomorphic state in nodes U, O, L and in Didelphis. It is worth mentioning this problem of optimization encountered under PAUP*, which presents a bias with respect to the other common parsimony softwares.

The absence of a naso-frontal suture with the medial process of frontals wedged between nasals (960>1) is here unambiguously optimized as a synapomorphy of metatherians except Deltatheridium, but it can also be considered as a synapomorphy of Notometatheria (see node T description), since it is unknown in Asiatherium, Didelphodon and Pediomys, plesiomorphic in the Gurlin Tsav Skull, and apomorphic in Sallacyon, Mayulestes, Andinodelphys and Marsupialia.

The presence of a deep groove for the internal carotid artery excavated on the anterior pole of the promontorium (1270>1) is unambiguously optimized as a synapomorphy of metatherians except Deltatheridium, and represents an unambiguous synapomorphy of Notometatheria in tree 1 (see node T description).

The other synapomorphies supporting this clade are: metacone and paracone bases separated (310>1); M4 preparacristae longer than that of M3 (340>1); protocone somewhat expanded anteroposteriorly (370>1, synapomorphy of Notometatheria in tree 1); transverse protocristid (740>1, synapomorphy of node R in tree 1); palatal expansion behind last molar (1000>1); internal acoustic meatus and fossa subarcuata subequal and separated by a sharp wall (1251>0, synapomorphy of node S in tree 1); presence of a deep, large fossa for the tensor tympani muscle excavated on the anterolateral aspect of promontorium (1280>1, synapomorphy of node S in tree 1); presence of a petrosal crest (1440>1, synapomorphy of Notometatheria in tree 1); petrosal contribution to the lateral wall of the epitympanic recess (1450>1, synapomorphy of node S in tree 1); presence of a hypotympanic sinus formed by squamosal, petrosal and alisphenoid (1630>1).

The stylar shelf slightly reduced labial to the paracone (180>1) was a synapomorphy of Marsupialia in the analysis of Rougier et al. (1998), and can also be regarded as a synapomorphy of node S in the first topology resulting from our analysis.

A tall protocone (390>1) was regarded by Rougier et al. (1998) as a convergent synapomorphy of North American plus Asiatic metatherians and of Notometatheria except borhyaenoids and Mayulestes.

A talonid subequal to, or wider than, the trigonid on m1–3 (580>2) represented a synapomorphy of North American plus Asiatic metatherians except Iqualadelphys in Rougier et al. (1998), and can also be regarded as a synapomorphy of node S in the first topology resulting from our analysis. A labial cristid obliqua (600>2) was a synapomorphy of Notometatheria except borhyaenids in Rougier et al. (1998), and can also be regarded as a synapomorphy of node R in the first topology of our analysis. The presence of palatal vacuities (990>1) was regarded as a synapomorphy of Andinodelphys plus Marsupialia by Rougier et al. (1998).

Node S: American and Australian metatherians

This clade is obtained in the first topology and is supported by five unambiguous synapomorphies, two of which are non-homoplasic. Three can be optimized at node B and were previously described: stylar shelf slightly reduced labial to paracone (180>1); procumbent protocone (380>1); posterior shelf of masseteric fossa present (850>1).

A first lower premolar oblique to the jaw axis (510>1) was a synapomorphy of Notometatheria in Rougier et al. (1998), although it is a feature of Didelphodon and Pediomys. The presence of a postpalatine torus (1010>1) was regarded as a synapomorphy of Notometatheria except borhyaenids by Rougier et al. (1998), but represents here a synapomorphy of node S as it is also present in Pediomys. The presence of a tympanic sinus in the lateral trough of petrosal (1620>1) may represent a synapomorphy of this clade or of Notometatheria (node T).

Node Q: Didelphodon plus Pediomys

This node is supported by eight unambiguous and homoplasic synapomorphies, three of which are reversals. The two North American taxa share an undulating epitympanic wing of the petrosal (1300>1), a fossa incudis separated from the epitympanic recess (1430>1, convergent with Dasycercus), and an imprint of the transverse sinus bifurcation on the petrosal (1490>1, convergent with didelphids).

Two other synapomorphies of this taxon are highly homoplasic: metacone on last upper molar present but not distinct from metastylar corner of tooth (320>1, convergent with node G or H and node P or Sallacyon); protocone with posterior portion expanded (371>2, convergent with Asiatherium, pucadelphyds, node I or node L and Dromiciops).

The three reversals that occur in this taxon are: M4 preparacrista shorter than or equal to that of M3 (341>0); absence of a deep groove for internal carotid artery excavated on anterior pole of promontorium (1271>0); mastoid tympanic process large and vertical (1391>0).

Node U: Mayulestes, Notogale and Sallacyon

This clade results from the first topology and confirms the inclusion of Mayulestes within borhyaenoids (sensuMuizon 1998; contra Rougier et al. 1998). In fact, the phylogenetic position of Mayulestes remains controversial as it can be either (1) the sister taxon of Notogale and Sallacyon (this hypothesis is favoured here as in the case of the monophyly of Notometatheria), or (2) diverging from Deltatheridium and sister taxa of all other metatherians, i.e. Mayulestes, Notogale and Sallacyon (in this hypothesis, Notometatheria is polypheletic).

The monophyly of borhyaenoids (i.e. including Mayulestes) is supported by six unambiguous and homoplasic synapomorphies, four of which are reversals. Mayulestes, Notogale and Sallacyon share a talonid that is narrower than the trigonid on m1–3 (582>1), and the absence of a prootic canal (1470>1, convergent with australidelphians). The four reversals are: protocone low (391>0, as in Caenolestes and peramelomorphs); m2–4 trigonids longer than wide (550>2, as in dasyurids and didelphids); entoconid smaller than hypoconid and/or hypoconulid (642>1, as in Andinodelphys); palatal vacuities absent or just small foramina (991>0, as in Pucadelphys).

One synapomorphy might be considered as a secondary acquisition (fast optimization) or as a first appearance (slow optimization): postmetacrista strongly developed, with paraconid enlarged and metaconid reduced on lower molars (190>1, or convergent with Deltatheridium, Didelphodon and the Gurlin Tsav skull; or acquired preliminarily in Metatheria and reversal in node R). This character supported independently the Deltatheroidea, ‘stagodontids’ and Notometatheria in the analysis of Rougier et al. (1998) and reflects an adaptation to carnivory.

Some characters from the lower molars may be considered as borhyaenoid synapomorphies. Another feature linked to an adaptation to carnivory is the reduction of the metaconid (782>0, convergent with Deltatheridium and Didelphodon) that is lower than the paraconid in Sallacyon and Notogale, but higher in Mayulestes. These taxa also share a metaconid on m1 that is lower than that of other molars (790>1, convergent with Deltatheridium and Caenolestes), and situated at the extreme lingual margin of the tooth (731>0, reversal as in Asiatherium). The cristid obliqua ends abruptly beneath the carnassial notch (602>0). The hypoconulid of m4 is tall and sharply recurved (620>1, convergent with pucadelphyds).

Two basicranial characters may represent synapomorphies of the group: presence of a medial process of squamosal in tympanic cavity (1700>1, convergent with pucadelphyds); very large jugular foramen (1710>1, convergent with pucadelphyds, Didelphis and peramelomorphs).

Notogale and Sallacyon (node P) share nine unambiguous synapomorphies, eight of which are homoplasic. Four represent synapomorphies of borhyaenoids (node U) in the first topology and are discussed above (i.e. 391>0, 582>1, 782>0, 1470>1). The unique unambiguous synapomorphy of this clade is the absence of the metaconid (771>0), which reflects strong carnivory. The entoconid is bladed and included in a prominent wall lingual to the talonid basin (650>1, convergent with node E); the para- and protocristids of lower molars are rounded (811>0, reversal as in didelphids, peramelomorphs, pucadelphyds); a secondary foramen ovale is present (1580>1, convergent with didelphids and peramelomorphs).

Nodes W and X: paraphyly of ‘Notometatheria’ and nesting of Asiatic and North American metatherians with clade C

The second topology suggests the exclusion of Mayulestes, Notogale and Sallacyon from ‘Notometatheria’ (polyphyletic in this hypothesis) and the paraphyly of ‘borhyaenoids’. The synapomorphies excluding these taxa from ‘Notometatheria’ are the eight unambiguous synapomorphies supporting node W. All of these are the inverse of that supporting the monophyly of ‘borhyaenoids’ in the first topology: protocone with posterior portion expanded (371>2) and tall (390>1); trigonids of m2–4 wider than long (552>0); talonid subequal to, to wider than, trigonid in m1–3 (581>2); hypoconulid of last molar short and erect (621>0); entoconid subequal to, to larger than, hypoconid and/or hypoconulid (641>2); metaconid taller than paraconid (780>2); palatal vacuities present (990>1).

The second topology suggests that the North American taxa Pediomys and Didelphodon are the sister group of clade C (i.e. ‘Notometatheria’ without ‘borhyaenoïds’). This relationship is supported by three unambiguous synapomorphies, one of which is non-homoplasic. Clade X is defined by the presence of a posterior shelf of masseteric fossa (850>1), the presence of a metaconid aligned with the paraconid on lower molars (730>1, convergent with Deltatheridium), and the presence of a postpalatine torus (1010>1, convergent with Mayulestes).

Node T: Notometatheria (South American and Australian metatherians)

Notometatheria was erected by Kirsch et al. (1997) on the basis of the rejection of a dichotomy Ameridelphia/Australidelphia (contra Szalay 1982a, b), and of the acceptance of a close relationship between, on the one hand, polydolopimorphs and Paucituberculata, and on the other hand, sparassodonts and didelphimorphs.

This clade is obtained in the first topology and is supported by four unambiguous and homoplasic synapomorphies. Two synapomorphies were also recognized by Rougier et al. (1998): medial process of squamosal in tympanic cavity (1700>1, fast optimization, reversal in node E), and deep groove for internal carotid artery excavated on anterior pole of promontorium (1270>1, reversal in node H and didelphids). Notometatherians also share a postglenoid-suprameatal vascular system below squamosal crest (1100>1), a tympanic sinus formed in the lateral trough of petrosal (1620>1, reversal in node E), and a petrosal crest on the petrosal (1440>1, reversal in MHNC 8369 and node H).

In case of slow optimization, this clade may also be characterized by the following ambiguous synapomorphies: M2 longer than, or subequal in length to M3 (161>0, reversal); medial internasal process of frontals absent (960>1); postglenoid foramen present (1110>1). In case of fast optimization, it may be supported by several ambiguous synapomorphies: protocone somewhat expanded anteroposteriorly (372>1); hypoconulid of m4 tall and sharply recurved (620>1); para- and protocristids subequal in length (801>0, reversal) and rounded (811>0, reversal); internal acoustic meatus shallow with thin prefacial commissure (1260>1). The presence of four lower incisors (451>0) and of the glenoid process of the alisphenoid (1080>1) were two synapomorphies of Notometatheria except for ‘borhyaenids’ in Rougier et al. (1998).

Node C: nesting of Type II, pucadelphyds, and MHNC 8369 with Marsupialia

Type II from Itaboraí is the first diverging taxon among notometatherians except for borhyaenoids, and pucadelphyds are the sister group of a clade containing MHNC 8369 and its sister taxon Marsupialia. This analysis thus reveals that the isolated petrosal MHNC 8369 from Tiupampa cannot be assigned to either Pucadelphys or Pucadelphydae.

This relationship is supported by three unambiguous and homoplasic synapomorphies: fossa subarcuata spherical (1200>1, reversal in Andinodelphys and Echymipera); internal acoustic meatus shallow with thin prefacial commissure (1260>1); dorsal tympanic aperture of hiatus Fallopii (1350>1, ventral in peramelomorphs, intermediate in didelphids and Pediomys).

Ambiguous synapomorphies that may support this clade are: five upper and four lower incisors (61>0, 451>0); stylar cusp E distal or at same level as D (270>1); lacrimal tubercle present (980>1). The other ambiguous synapomorphies optimized for this clade (Acctran) are discussed under node E because they may be optimized so (Deltran), and Type II does not provide any data on these features (unavailable characters).

Node D: nesting of pucadelphyds and MHNC 8369 with Marsupialia

This clade is supported by two unambiguous and homoplasic synapomorphies: absence of post-temporal sulcus and notch on the squamosal surface of the petrosal (1520>1, 1530>1; both reversal in didelphids). The other supporting synapomorphies can also be optimized for node C (see above) and are those supporting the node including Jaskhadelphys, Pucadelphys, Andinodelphys and Marsupialia in the analysis of Rougier et al. (1998): stylar cusp B large but less than paracone (220>2); meta- and paracones not conical, with labial face flat or convex (300>1), V-shaped centrocrista (330>1). Moreover, these taxa share a strongly developed postmetacrista, with paraconid enlarged and metaconid reduced on lower molars (191>0, reversal); the presence of a frontal-maxillary contact (970>1, that precludes a naso-lacrimal contact); a postpalatine torus (1010>1); a glenoid process of alisphenoid (1080>1); and an interparietal (1150>1).

The presence of a stylar cusp C does not support this clade as the coding for the character is different from that of Rougier et al. (1998), given that the homology of the stylar cusps among metatherians is interpreted differently herein (see list of characters in Appendix).

Node O: Pucadelphydae

Pucadelphys and Andinodelphys appear to be closely related, and are the two genera of the family Pucadelphydae (according to Muizon et al. 1997). This relationship is supported by 11 unambiguous synapomorphies, one of which is non-homoplasic and five are reversals.

A broad shelf of bone surrounding fenestra cochleae and making a separation between it and the aqueductus cochleae (1320>1) is an autapomorphy of pucadelphyds. The presence of a stylar cusp C is a synapomorphy of pucadelphyds (230>1, convergent with Didelphis and Marmosa). However, and contrary to Marshall et al. (1990), the presence of an appressed pair of stylar cuspules in C position was not considered in our analysis because it appears to be a highly variable feature in Andinodelphys, as in some extant metatherians (e.g. didelphids).

The other synapomorphies supporting this clade are: protocone with posterior portion expanded (371>2); hypoconulid of m4 tall and sharply recurved (620>1); protoconid subequal to para- and/or metaconid (760>1); para- and protocristids subequal in length (801>0, reversal); vascular groove medially adjacent to prootic sinus sulcus, on the pars mastoidea (i.e. prootic sinus vein or connection) (1510>1); absence of alisphenoid tympanic process (1591>0, 1601>0, reversals); presence of a tympanic sinus formed by petrosal and alisphenoid (1640>1); and absence of an auditory bulla (1661>0, reversal).

Pucadelphyds, Mayulestes and Type II exhibit a deep subtriangular sinus on the vestigial lateral trough of the petrosal that is interpreted here as unique for these taxa since the formation of a so-called hypotympanic sinus in extant taxa implies primarily the alisphenoid and may be completed by the petrosal (see Appendix). In Mayulestes this sinus is completed by the alisphenoid and squamosal and approximates the condition of extant metatherians.

The ambiguous synapomorphies that are likely to support the monophyly of pucadelphyds can also be optimized as synapomorphies of node B or node C (see Supplementary Data file). A first lower premolar obliquely orientated in the jaw (510>1, convergent with node Q, Mayulestes and Caenolestes) may represent alternatively a synapomorphy of pucadelphyds or of metatherians except for Deltatheridium. It was considered by Rougier et al. (1998) to be a synapomorphy of Notometatheria. A medial process of the squamosal in the tympanic cavity was a notometatherian synapomorphy in Rougier et al. (1998; reversal in Marsupialia), and also in our analysis (in the first topology; see node T).

The anterior lamina of the petrosal contributes to the internal wall of the braincase, the external wall of which is formed by the alisphenoid and squamosal (1232>1). This condition may be optimized as present in Andinodelphys (given the currently available data of other taxa) and thus represent a synapomorphy of this group. In non-tribosphenic mammals, the anterior lamina of the petrosal is large and a major contributor to the lateral braincase wall, while it it is greatly reduced and disappears in some therians (Wible 1990). The alisphenoid of most tribosphenidans articulates with the squamosal and excludes the anterior lamina from the lateral braincase wall (Marshall and Muizon 1995). The peculiar condition of Pucadelphys andinus may be interpreted as intermediate because the anterior lamina contributes to the internal wall of the braincase, whereas the external wall is formed by the alisphenoid and squamosal (Marshall and Muizon 1995). However, it is interpreted here as an autapomorphy of pucadelphyds that derives from a vestigial anterior lamina of the petrosal.

Node E: (MHNC 8369 (didelphids (Type I (Caenolestes (australidelphians))))

The sister taxa relationship of MHNC 8369 and Marsupialia is supported by two unambiguous synapomorphies from the basicranium, one of which is a reversal: absence of a deep and large fossa for the tensor tympani muscle excavated on the anterolateral aspect of promontorium (1281>0, reversal); mastoid tympanic process indistinct to absent (1391>2). This last feature was regarded as a synapomorphy of Marsupialia by Rougier et al. (1998).

Node F: Marsupialia

The clade Marsupialia can be defined as including the last common ancestor of living marsupials, plus all its descendants, or as including any metatherian more closely related to living marsupials than to other metatherians. Our phylogenetic hypothesis shows that Didelphidae is the first diverging taxon, nested with Type I and its sister group Caenolestes plus Australidelphia. The monophyly of Caenolestes (or Caenolestidae, or Paucituberculata) and Australidelphia was highlighted in many recent studies (e.g. Sánchez-Villagra 2001a, b; Sánchez-Villagra and Wible 2002; Amrine-Madsen et al. 2003; Horovitz and Sánchez-Villagra 2003; Luo et al. 2003; Asher et al. 2004; Cardillo et al. 2004).

Marsupialia is supported by two unambiguous and homoplasic petrosal synapomorphies: rostral tympanic process of petrosal (1330>1); petrosal contribution to the lateral wall of the epitympanic recess slender and triangular (1460>1). A hypotympanic sinus formed by the alisphenoid and petrosal or by the alisphenoid only (1650>1) may be considered as an unambiguous and non-homoplasic synapomorphy of this clade as it is unknown for extra-Marsupialia metatherians (except for Pediomys, Asiatheriurm and the Gurlin Tsav Skull in which it is polymorphic).

Three ambiguous synapomorphies have been described previously for Marsupialia (Rougier et al. 1998): absence of a deep groove for internal carotid artery excavated on anterior pole of promontorium (1271>0, reversal but secondary acquisition in Type I); absence of a medial process of squamosal in tympanic cavity (1701>0, reversal); jugular foramen subequal to fenestra cochleae (1711>0, reversal).

The other possible characteristics of Marsupialia are: first lower premolar transversal to jaw (511>0, reversal); entoconid subequal to, to larger than, hypoconid and/or hypoconulid (641>2) and included in a prominent and bladed wall lingual to the talonid basin (650>1); hypoconulid salient labially (672>1); presence of a labial postcingulid (701>0); presence of a large anterior cingulum (720>1); condyle of mandible very high relative to tooth row (860>1); absence of large mental foramina (891>0); absence of flaring of cheeks behind infraorbital foramen (950>1); absence of lacrimal tubercle (980>1); presence of a transverse canal (1680>1).

It is worth noting that an intermediate position of the hiatus Fallopii aperture (1351>2) supports the clade Didelphidae (node M), as stated by Sánchez-Villagra and Wible (2002). A ventral aperture of the canal Fallopii appears to be plesiomorphic, as suggested by observations on Morganucodon, Triconodon and Vincelestes. However, these genera may well not be homologous because of the morphology of their periotic, which is very different from that of metatherians. Among other features, the petrosal of the three genera lacks a cavum supracochleare because the geniculate ganglion is not separated dorsally from the cerebral cavity by a bony roof as in tribosphenidan therians. The ventral position of the hiatus Fallopii of peramelomorphs (Echymipera and Perameles herein) can be interpreted as a secondary acquisition.

Another unambiguous synapomorphy of didelphids is the presence of a tall paroccipital process of the exoccipital (1120>1, convergent with dasyurids). This feature does not occur simultaneously with the presence of a very small to indistinct mastoid tympanic process of the petrosal (1401>2 at node E), suggesting that both structures are not homologous. This hypothesis is not in accord with that of Rougier et al. (1998) because none of their taxa presented both structures (see discussion in Appendix).

Node G: (Type I (Caenolestes, Australidelphians))

Type I from Itaboraí is the sister taxon of Caenolestes plus Australidelphia, as highlighted in previous studies (Ladevèze 2004). Two unambiguous petrosal synapomorphies support this relationship: posterior exposure of the pars mastoidea rounded and bulbous owing to the excavation of the fossa subarcuata (1210>1); presence of foramina on the sigmoid sinus and/or prootic sinus, apparently connecting both vessels (1500>1). Twenty-two ambiguous synapomorphies may support this clade in a fast optimization, but all are dental or cranial (except basicranial) characters. They cannot be considered significant for supporting this clade as Type I does not provide any information about these features.

Node H: Caenolestes nested with Australidelphia

This clade is supported by two unambiguous and homoplasic petrosal synapomorphies: caudal tympanic process of petrosal that floors the postpromontorial sinus (1410>1, reversal in Echymipera); absence of petrosal crest (1441>0, reversal). The other synapomorphies that may support this clade, in a slow optimization, are all dental features except for one on the skull: spatulate upper incisors (70>1, reversal in dasyurids); first upper premolar not procumbent (121>0, reversal); last upper molar smaller than penultimate upper molar (170>1, reversal in Echymipera); stylar cusp B smaller or subequal to stylar cusp D on M2 (251>0); metacone on last upper molar present but not distinct from metastylar corner of tooth (320>1, absent in Dromiciops and Echymipera); absence of fossa for the lower canine (492>0, present in node J), and absence of anterolateral process of the maxilla (500>2, present in dasyurids); two cusps on m4 talonid (560>1); entocristid well developed and sharp (680>1, reversal in dasyurids); posteriormost point of premaxillo-nasal contact posterior to the canine (920>1, reversal in Dromiciops); small minor palatine (postpalatine) foramen (1021>0, reversal, large in Perameles).

According to Horovitz and Sánchez-Villagra (2003), the absence of a post-temporal sulcus is a synapomorphy of this clade, but here this feature supports the nesting of pucadelphyds with Marsupialia (node D), as the description of the new isolated petrosals highlighted the absence of such a structure in Pucadelphys and Andinodelphys.

Node I: Australidelphia

The inclusion of microbiotheriids within Australidelphia has been confirmed and largely argued for, especially by most recent morphological, biochemical and molecular studies (e.g. Szalay 1982a, b, 1994; Kirsch et al. 1991, 1997; Luckett 1994; Springer et al. 1998; Phillips et al. 2001; Amrine-Madsen et al. 2003; Horovitz and Sánchez-Villagra 2003; Luo et al. 2003; Nilsson et al. 2003), even if its position within that group remains controversial. It is supported by five unambiguous cranial and basicranial synapomorphies, two of which are non-homoplasic: ectotympanic moderately broadened (1130>1); rostral tympanic process of petrosal well developed anterolaterally, sometimes contacts the ectotympanic, and extends the whole length of the promontorium (1340>1).

The anterior lamina of petrosal is greatly reduced and without a depression (1240>1, convergent with didelphids and Pediomys), the petrosal contribution to the lateral wall of the epitympanic recess forms a thin lamina (1461>2), and the prootic canal is lost in adults (1470>1, convergent with Sallacyon, Notogale, and Mayulestes).

Three synapomorphies described by Rougier et al. (1998) are optimized for Caenolestes plus Australidelphia in our analysis: first upper premolar not procumbent (121>0); last upper molar smaller than penultimate upper molar (170>1); small minor palatine (postpalatine) foramen (1021>0). The absence of a stylar cusp D cannot be considered as an australidelphian synapomorphy any longer because the hypothesis of stylar cusp homology is modified from that of Rougier et al. (1998) herein (see Appendix). A first upper incisor subequal to, or smaller than, remaining incisors, without diastema (90>1, reversal in dasyurids) is a synapomorphy of Australidelphia (Rougier et al. 1998).

Additional synapomorphies that may be optimized for Australidelphia are: postcingulum on M1–3 present (440>1, reversal in Dromiciops); frontal-squamosal contact on braincase (1160>1, reversal in dasyurids); alisphenoid tympanic process extends posteriorly to reach posterior limit of the alisphenoid hypotympanic sinus in ventral view (1610>1, reversal in Echymipera).

The present definition of Australidelphia is congruent with that of Wroe et al. (2000). However, these authors arbitrarily described one topology among the 64 most parsimonious trees obtained (ibid., fig. 2), whereas the Australidelphia did not represent a monophyletic clade in the strict consensus (ibid., fig. 1).

Node J: Dromiciops nested with dasyurids

Different hypotheses are put forward with respect to the phylogenetic status of the microbiotheriid Dromiciops among australidelphians: (1) Dromiciops is the sister group of all other australidelphians to the exclusion of the Peramelina (Eometatheria hypothesis; Kirsch et al. 1997; Asher et al. 2004); (2) Dromiciops is nested with dasyuromorphs (Szalay and Sargis 2001) or (3) with diprotodontians (Kirsch et al. 1991; Horovitz and Sánchez-Villagra 2003).

Our hypothesis suggests a sister group relationship of Dromiciops and dasyurids, supported by ten unambiguous synapomorphies, four of which are non-homoplasic: expansion of the crista petrosa that forms a thin lamina covering the anterolateral part of the fossa subarcuata (1220>1); presence of a stylomastoid foramen (1360>1); presence of a petrosal plate (1420>1); presence of a tubal foramen just posterior and ventral to the foramen ovale (1670>1). Moreover, Dromiciops and dasyurids share: a first upper premolar without diastema (111>0, reversal); absence of conules on upper molars (352>0); a fossa for the lower canine reduced to a small cupule (490>1, large in Dasycercus); conical entoconid (651>0, reversal); absence of a well-developed anterior cingulum on lower molars (721>0, reversal); a postglenoid-suprameatal vascular system above squamosal crest (1101>2).

Conclusions

Contrary to Marshall et al. (1990), Andinodelphys has no relationship with any Australian taxon. Our phylogenetic hypothesis agrees with the interpretation of Muizon et al. (1997), who, on the basis of more complete material, argued that this Tiupampan taxon was not an australidelphian. Andinodelphys and Pucadelphys are both representative of the Pucadelphydae and are the sister group of Marsupialia (i.e. any metatherian more closely related to living marsupials than to other metatherians). They are very similar but differ in certain features. In particular, Andinodelphys differs from Pucadelphys (and Mayulestes) in having small palatal fenestrae, and a small transverse canal.

The placental-marsupial split seems to have occurred in Asia according to the currently available evidence (Ji et al. 2002; Luo et al. 2003). The basal deltatheroidans of the metatherians are also American and Asian (Rougier et al. 1998; Luo et al. 2003; Kielan-Jaworowska et al. 2004). Within the metatherians, both previous hypotheses (Kirsch et al. 1997; Rougier et al. 1998; Wible et al. 2001; Ladevèze 2004; 2007) and our parsimony analysis based on dental and cranial (including petrosal) characters indicate that a North American group gave rise to a South American-Australian clade at or near the end of the Cretaceous Period. The first divergent lineage of the Notometatheria (in the acceptance of its monophyly) includes the borhyaenoids Mayulestes, Notogale and Sallacyon, nested with Type II from Itaboraí and the pucadelphyds from Tiupampa (one of the oldest South American metatherian assemblages; Marshall and Muizon 1988, 1995; Muizon and Brito 1993).

The inclusion of Mayulestes within borhyaenoids was argued for by Muizon (1998) on the basis of several dental and cranial features. Its molar morphology exhibits some characters related to postvallum-prevallid shear (paracone smaller than metacone, postmetacrista enlarged, metaconid smaller than paraconid) that are interpreted as ‘plesiomorphic features not only for a borhyaenoid but also for a metatherian’ (Muizon 1998, p. 85). Moreover, Muizon (1998) showed that most of the cranial characters shared by Mayulestes and other borhyaenoids were therian or metatherian plesiomorphies (e.g. naso-lacrimal contact, absence of transverse canal, absence of palatal fenestrae, and absence of alisphenoid tympanic process except in Cladosictis, Sipalocyon and Notogale). In Mayulestes, a medial process of the squamosal participates in the formation of a hypotympanic sinus, with the petrosal and a lesser part of the alisphenoid. The presence of a medial tympanic process of the squamosal in the tympanic cavity may represent a synapomorphy of borhyaenoids in our phylogenetic hypothesis. Mayulestes and other borhyaenoids also share a very large jugular foramen, as in pucadelphyds.

Consideration of dental, general cranial and basicranial characters in our parsimony analysis led to a lack of resolution in the strict consensus tree, resulting from the ambiguous phylogenetic status of ‘Notometatheria’ and ‘borhyaenoids’, and each dataset provided conflicting phylogenetic information. A complementary study involving supplementary taxa, such as Cretaceous North American taxa, and Paleocene South American taxa (including Tiupampan metatherians), which are unfortunately mostly known by dental remains, would be interesting. This would generate a large number of missing data, but could provide enough information to resolve these uncertain nodes.

Acknowledgements.  Specimens of Pucadelphys and Andinodelphys were collected in 1996, 1998 and 1999 with the financial support of the Institut Français d'Études Andines (Lima, Peru) and the National Geographic Society (grant 6296/98). Field expeditions have been made possible thanks to the invaluable collaboration of Lic. Ricardo Cespedes (MHNC, Cochabamba) and Dr Ramiro Suarez (Fundacion par la Ciencia, Cochabamba). For permitting access to materials and help, we thank J. Bonaparte and A. Kramatz (MACN, Buenos Aires), G. W. Rougier (University of Louisville, Kentucky), L. Bergqvist (UFRJ, DNPM, Rio de Janeiro), A. Kellner and D. Enriques (MNRJ, Rio de Janeiro), and F. Goin (Museo de La Plata). We are grateful to P. Janvier (MNHN, USM 203-UMR 5143, Paris), R. Debruyne (MNHN, USM 203-UMR 5143, FRE 2696, Université Paris VI, Paris), and D. Germain (FRE 2696, Université Paris VI) for fruitful discussions and critical comments. We thank A. Benz, J. Cuisin and F. Renoult (MNHN, UMR 8570, Paris), who gave access to extant marsupial collections in their care, and D. Serrette and P. Loubry (MNHN, USM 203-UMR 5143, Paris) for the photographs. This work was supported by a MENRT Grant (ED 227–2237) from the French Ministry of Research, and fellowships from the MNHN and the Société des Amis du Muséum, Paris.

Appendix

List of the 83 dental (1–83), 34 general cranial (84–117) and 55 basicranial (118–172) characters

Some character states are revised for Pucadelphys and Andinodelphys, according to the observations of Rougier et al. (1998), Wroe et al. (2000) and Wible et al. (2001). References follow most character definitions.

General dentition

  1. Number of premolars: 5 (0), 4 (1), 3 (2), or less than 3 (3). Rougier et al. (1998; ch. 1), Wible et al. (2001; ch. 1).

  2. Tall, trenchant premolar: in last PM position (0), in penultimate PM position (1), or absent (2) (upper dentition considered when possible). Rougier et al. (1998, ch. 3), Wible et al. (2001, ch. 3), Luo et al. (2003; ch. 38).

  3. Number of molars: 4 (0), 3 (1), or 2 (2). Springer et al. (1997, ch. 4), Rougier et al. (1998, ch. 4), Wible et al. (2001, ch. 4), Horovitz and Sánchez-Villagra (2003, ch. 152).

  4. Size of molars increasing posteriorly: absent (0), moderate posterior increase (1), or marked posterior increase (2) (< jaw: all M, > jaw: all but the last). Marshall and Kielan-Jaworowska (1992), Rougier et al. (1998, ch. 6), Wible et al. (2001, ch. 6). In most taxa showing state 1 in Rougier et al. (1998) and Wible et al. (2001) analyses, the molars increase from M1 to M2, but not from M2 to M3, and most of the time only the length increases.

  5. Number of postcanine tooth family: 8 or more (0), 7 (1), or < 7 (2). Rougier et al. (1998, ch. 7), Wible et al. (2001, ch. 7).

Upper incisors

  6. Number of upper incisors: 5 (0), fewer than 5 (1), or none (2). Springer et al. (1997, ch. 1), Rougier et al. (1998, ch. 8), Wroe et al. (2000, ch. 1), Wible et al. (2001, ch. 8), Horovitz and Sánchez-Villagra (2003, ch. 150).

  7. Shape of upper incisors: peg shaped (0), spatulate (1). Reig et al. (1987, ch. 20), Springer et al. (1997, ch. 29), Wroe et al. (2000, ch. 2), Horovitz and Sánchez-Villagra (2003, ch. 165).

  8. Upper incisor arcade shape: U-shape (0), broad V-shape (1), long, narrow V-shape (2). Springer et al. (1997, ch. 25), Horovitz and Sánchez-Villagra (2003, ch. 161).

  9. First upper incisor enlarged, anteriorly projecting, separated from I2 by small diastema (0), subequal or smaller than remaining incisors, without diastema (1), or lost (2). Muizon et al. (1997), Rougier et al. (1998, ch. 9), Wible et al. (2001, ch. 9). Andinodelphys is scored 0, even if the I1 is only slightly larger than I2. A new specimen of Pucadelphys allowed us to score this taxon. This new skull exhibits the rostrum in its totality, and shows that the I1 is neither larger than other incisors nor procumbent. Thus Pucadelphys is scored 1 for this character.

 10. Size of upper I3 vs. I2: I3 > I2 (0), I3 = I2 (1), I3 < I2 (2). Springer et al. (1997, ch. 30), Horovitz and Sánchez-Villagra (2003, ch. 166).

Upper premolars

 11. First upper premolar with diastema: absent (0) or present (1). Rougier et al. (1998, ch. 11), Wible et al. (2001, ch. 11), Luo et al. (2003, ch. 39). This character was modified from the previous references because we stated that the presence of a diastema and the orientation of a tooth had to be coded separately, as there is no evident link between these two features. In case of procumbency, the tooth is separated from its neighbour by a diastema; however, a non-procumbent tooth may also exhibit a large diastema.

 12. First upper premolar procumbent: absent (0) or present (1). Rougier et al. (1998, ch. 11), Wible et al. (2001, ch. 11), Luo et al. (2003, ch. 39).

 13. Posterolingual cuspule on the last upper premolar: absent (0) or present (1). Wroe et al. (2000, ch. 7).

 14. Shape of the last upper premolar: transversely compressed in occlusal view (0) or bulbous and ovate in occlusal view (1). Reig et al. (1987, ch. 24), Wroe et al. (2000, ch. 6).

Upper molars

 15. Upper molar outline in occlusal view: triangular or subtriangular (0), or rectangular or subsquare (1). Springer et al. (1997, ch. 5), Horovitz and Sánchez-Villagra (2003, ch. 153).

 16. Relative sizes of M2 (antepenultimate molar) and M3 (penultimate molar): M2 longer or subequal in length to M3 (0) or M2 shorter than M3 (1). Marshall (1987, ch. 3), Reig et al. (1987, ch. 3).

 17. Last upper molar width relative to penultimate upper molar: subequal (0) or smaller (1). Rougier et al. (1998, ch. 41), Wible et al. (2001, ch. 41).

 18. Stylar shelf uniform in width (0), slightly reduced labial to paracone (1), strongly reduced labial to paracone (2), or strongly reduced or absent (3) (penultimate molar considered when present). Rougier et al. (1998, ch. 17), Wible et al. (2001, ch. 17).

 19. Postmetacrista weakly developed (0) or strongly developed, with a large metastylar area on upper molars, and with paraconid enlarged and metaconid reduced on lower molars (1). (Cifelli 1993), Rougier et al. (1998, ch. 18, 32), Wible et al. (2001, ch. 18, 32). Characters 18 and 32 of Rougier et al. (1998) and Wible et al. (2001) were considered together as they deal with the same functional complex. A large metastylar area with a strongly developed postmetacrista reflects an adaptation to carnivory and is associated with the reduction of the metaconid and enlargement of paraconid on lower molars.

 20. Deep ectoflexus present on one or more teeth (0), or upper molars without a distinct ectoflexus on any tooth (1). Voss and Jansa (2003, ch. 59). The ectoflexus definition provided by Rougier et al. (1998, ch. 19) is not considered here, as it has been argued that didelphids exhibit a significant variation of the ectoflexus development. Voss and Jansa (2003, ch. 59) showed that only Caluromys and Caluromysiops lack any trace of an ectoflexus on the upper molars, and that a distinct ectoflexus is present on M3, on M2 and M3, or (rarely) on M1–M3 in all other didelphids of their analysis. Rougier et al. (1998) scored the presence or absence of a ‘deep’ ectoflexus on M2 and M3, but a complete range of intermediate conditions of the ectoflexus from shallow to deep among the taxa treated herein prevents us from making such distinctions.

 21. Stylar cusp A distinct but smaller than B (0), subequal to larger than B (1), or very small to indistinct (2) (penultimate molar considered when available). Rougier et al. (1998, ch. 20), Wroe et al. (2000, ch. 16), Wible et al. (2001, ch. 20).

 22. Stylar cusp B small (0), vestigial to absent (1), large but less than paracone (2), larger than paracone (3) (penultimate upper molar considered when available). Rougier et al. (1998, ch. 22), Wroe et al. (2000, ch. 17), Wible et al. (2001, ch. 22). Caenolestes has two enormous cusps on the labial side of the trigon. These are identified as extremely inflated stylar cusps B and D. Consequently, the paracone and metacone are greatly reduced and included in the lingual edge of cusps B and D, respectively.

 23. Stylar cusp C on penultimate upper molar: absent (0) or present (1). Rougier et al. (1998, ch. 23), Wroe et al. (2000, ch. 21), Wible et al. (2001, ch. 23). The different extant taxa studied in our analysis have raised the problem of stylar cusp homology. Didelphids, dasyurids, caenolestids and peramelomorphs exhibit a great disparity of forms and development of these structures. Here, we follow the terminology of Archer (1976b) and score the absence of a stylar cusp C for dayurids and peramelomorphs (and consequently the presence of a stylar cusp D for these taxa). The presence of a stylar cusp C appeared highly variable in Didelphis (Flores and Abdala 2001), but this might be affected by the age of an individual (i.e. a worn tooth has an indistinct stylar cusp C). Thus, Didelphis (and other didelphids) is scored 1 as most of the studied specimens show a stylar cusp C on their M3. Moreover, the presence of a twinned cusp in C position is not considered in our analysis because important variations were observed in Pucadelphys and Andinodelphys, as in certain extant didelphids (even in the same individual). Thus this feature, evaluated as an apomorphy shared by Andinodelphys and Australidelphia by Marshall et al. (1990), is not constant and cannot be included in a parsimony analysis.

 24. Stylar cusp D on penultimate upper molar: present (0) or absent (1). Wroe et al. (2000, ch. 18). Here, we follow the terminology of Archer (1976b) and score the presence of a stylar cusp D for dayurids and peramelomorphs. A molar stylar cusp in the D position (i.e. between stylar cusps C and E on the stylar shelf) is present in most metatherians (Cifelli 1993). A meristic gradient, with stylar cusp D increasing from M1–2 and decreasing from M3–4, is almost universal for carnivore and insectivore metatherians (Wroe 1997). Andinodelphys and Pucadelphys show moderate development of stylar cusp D, with stylar cusp B the largest and highest cusp. In didelphids, stylar cusp D is generally present, but of comparable size or smaller than stylar cusp B, but some derived forms show a marked reduction of stylar cusp D (e.g. Lutreolina). Mayulestes has a stylar cusp D that is markedly smaller than cusp B, as in all borhyaenoids in which stylar cusp D is either reduced compared to cusp B or lost (Marshall 1981; Muizon 1994).

 25. Relative size of stylar cusps B and D on M2: B smaller or subequal to D (0), or B larger than D (1). Rougier et al. (1998, ch. 24), Wroe et al. (2000, ch. 19), Wible et al. (2001, ch. 24).

 26. Stylar cusp E: absent or poorly developed (0) or well developed (1). Rougier et al. (1998, ch. 25), Wible et al. (2001, ch. 25), Luo et al. (2003, ch. 110).

 27. Position of the stylar cusp E relative to cusp D or ‘D-position’: E more lingual to D or ‘D-position’ (0), or E distal to or at same level as D or ‘D-position’ (1). Rougier et al. (1998, ch. 25), Wible et al. (2001, ch. 25), Luo et al. (2003, ch. 110).

 28. Metacone size relative to paracone: noticeably smaller (0), subequal (2), or noticeably larger (3) (second upper molar considered when available). Reig et al. (1987, ch. 5), Springer et al. (1997, ch. 7), Rougier et al. (1998, ch. 27), Asher (1999, ch. 41), Wroe et al. (2000, ch. 8), Wible et al. (2001, ch. 27), Luo et al. (2002, ch. 81), Horovitz and Sánchez-Villagra (2003, ch. 155).

 29. Metacone position relative to paracone: labial (0), approximately at same level (1), or lingual (2). Rougier et al. (1998, ch. 28), Wible et al. (2001, ch. 28).

 30. Metacone and paracone shape: conical, with labial face concave (0), or not conical, with labial face flat or convex (1). Rougier et al. (1998, ch. 29), Wible et al. (2001, ch. 23).

 31. Metacone and paracone bases: adjoined (0) or separated (1). Rougier et al. (1998, ch. 30), Wible et al. (2001, ch. 30).

 32. Metacone on last upper molar: present and distinct from metastylar corner of tooth (0), present but not distinct from metastylar corner of tooth (1), or absent (2). Wroe et al. (2000, ch. 9).

 33. Centrocrista: straight (0) or V-shaped (1). Reig et al. (1987, ch. 1), Springer et al. (1997, ch. 8), Rougier et al. (1998; ch.31), Wroe et al. (2000, ch. 10), Wible et al. (2001, ch. 31), Luo et al. (2002, ch. 82), Horovitz and Sánchez-Villagra (2003, ch. 156).

 34. Relative lengths of M3 and M4 preparacristae (penultimate and ultimate upper molars): M4 preparacrista shorter than or equal to that of M3 (0), or M4 preparacrista longer than that of M3 (1). Wroe et al. (2000, ch. 13).

 35. Conules: absent (0), small, without cristae (1), or strong, labially placed, with wing-like cristae (2). Cifelli (1993), Rougier et al. (1998, ch. 35), Wible et al. (2001, ch. 35). A metaconule is present in Deltatheridium according to Kielan-Jaworowska and Nessov (1990).

 36. Trigon basin on upper molars: absent (0) or present (1). Reig et al. (1987, ch. 10), Rougier et al. (1998, ch. 36), Wible et al. (2001, ch. 36).

 37. Protocone on upper molars: small (0), somewhat expanded anteroposteriorly (1), or with posterior portion expanded (2). Reig et al. (1987, ch. 10), Rougier et al. (1998, ch. 36), Wible et al. (2001, ch. 36). Characters 36 and 37 are derived from character 36 of Rougier et al. (1998) and Wible et al. (2001), which was divided in two distinct characters as it dealt with the presence of a trigon basin and the morphology of the protocone.

 38. Procumbent protocone: absent (0) or present (1). Rougier et al. (1998, ch. 37), Wible et al. (2001, ch. 37).

 39. Protocone height: low (0) or tall, approaching para- and/or metacone height (1). Rougier et al. (1998, ch. 38), Wible et al. (2001, ch. 38).

 40. Paracingulum: absent (0), interrupted between stylar margin and paraconule (1), or continuous (2) (penultimate molar considered when available). Rougier et al. (1998, ch. 26), Wible et al. (2001, ch. 26).

 41. Preprotocrista does not (0) or does (1) extend labially past base of paracone and form a paracingulum (double rank prevallum/postvallid shearing). Cifelli (1993), Rougier et al. (1998, ch. 33), Wible et al. (2001, ch. 33).

 42. Postprotocrista does not (0) or does (1) extend labially past base of metacone and form a metacingulum (double rank prevallum/postvallid shearing). Cifelli (1993), Rougier et al. (1998, ch. 34), Wible et al. (2001, ch. 34).

 43. Precingulum on M1–3: absent (0) or present (1). Rougier et al. (1998, ch. 39), Wroe et al. (2000, ch. 24), Wible et al. (2001, ch. 39).

 44. Postcingulum on M1–3: absent (0) or present (1). Rougier et al. (1998, ch. 39), Wroe et al. (2000, ch. 25), Wible et al. (2001, ch. 39).

Lower incisors

45. Number of lower incisors: 4 (0), less than 4 (1), none (2). Springer et al. (1997, ch. 2), Rougier et al. (1998, ch. 42), Wroe et al. (2000, ch. 26), Wible et al. (2001, ch. 42), Luo et al. (2002, ch. 105), Horovitz and Sánchez-Villagra (2003, ch. 151).

 46. Staggered lower incisor (i3 of Hershkovitz 1982): absent (0) or present (1). Rougier et al. (1998, ch. 43), Wroe et al. (2000, ch. 29), Wible et al. (2001, ch. 43), Horovitz and Sánchez-Villagra (2003, ch. 172).

 47. Bilobed i3: absent (0) or present (1). Wroe et al. (2000, ch. 27).

 48. Lower incisors and/or canines semiprocumbent or procumbent (i.e. forming an angle with the jaw of 45 degrees at least): absent (0) or present (1). Springer et al. (1997, ch. 33), Luo et al. (2002, p. 115), Horovitz and Sánchez-Villagra (2003), ch. (167).

Lower canine

 49. Fossa for the lower canine: absent (0), reduced to a small cupule (1), or large (2). Rougier et al. (1998, ch. 81), Wroe et al. (2000, ch. 46), Wible et al. (2001, ch. 81).

 50. Anterolateral process of the maxilla: wholly forms the border of the fossa for the lower canine (0), partially forms the border of the fossa for the lower canine (i.e. with premaxillae) (1), or absent (2). Rougier et al. (1998, ch. 81), Wroe et al. (2000, ch. 46), Wible et al. (2001, ch. 81).

Lower premolars

 51. First lower premolar orientated in line with jaw axis (0) or oblique (1). Marshall (1987, ch. 16), Rougier et al. (1998, ch. 45), Wible et al. (2001, ch. 45).

 52. Second lower premolar: smaller than (0), subequal in size to (1) or larger than third premolar (2). Marshall (1987, ch. 14–15), Rougier et al. (1998, ch. 46), Wroe et al. (2000, ch. 43), Wible et al. (2001, ch. 46).

 53. Shape of p3 (ultimate premolar): trenchant and narrow transversely (0), somewhat inflated and ovoid in occlusal view (1), enormous, bulbous, and has a distinct crushing function (2). Marshall (1987, ch. 13).

Lower molars

 54. Trigonid configuration on m1–3: open, i.e. the angle formed by protocristid and paracristid approximates 90 degrees (0), more acute, i.e. the angle formed by protocristid and paracristid approximates 45 degrees (1), or anteroposteriorly compressed, i.e. protocristid and paracristid are subparallel (2). Cifelli (1993), Rougier et al. (1998, ch. 48), Wible et al. (2001, ch. 48).

 55. Shape of m2–4 trigonids: wider than long (0), or as wide as long (1), or longer than wide (2). Reig et al. (1987, ch. 13).

 56. Number of cusps on m4 talonid: 3 cusps (0), 2 cusps (1), or 1 cusp (2). Wroe et al. (2000, ch. 44). The case of Vincelestes is peculiar as the lower molars of this taxon have a single talonid cusp. Thus the lower molars have no real talonid, and Vincelestes is scored as inapplicable data.

 57. Relative sizes and shapes of m3 and m4 talonids: talonid of m4 similar in size and shape to that of m3 (0), or talonid of m4 reduced and narrower than that of m3 (1). Reig et al. (1987, ch. 16).

 58. Talonid width relative to trigonid on m1–3: much narrower, subequal in width to base of metaconid, and at the lingual half or two-thirds of the protocristid (0), narrower (1), or subequal to wider (2). Springer et al. (1997, ch. 13), Rougier et al. (1998, ch. 50), Wible et al. (2001, ch. 50), Luo et al. (2002, ch. 63), Horovitz and Sánchez-Villagra (2003, ch. 158).

 59. Trigonid noticeably taller than talonid: absent (0) or present (1)

 60. Anterior point of termination of the cristid obliqua in m3 with respect to carnassial notch formed by postprotocristid and metacristid: beneath carnassial notch (0), lingual to carnassial notch (1), labial to carnassial notch (2). Archer (1976b), Cifelli (1993), Springer et al. (1997, ch. 23), Rougier et al. (1998, ch. 51), Godthelp et al. (1999), Wroe et al. (2000, ch. 40), Wible et al. (2001, ch. 51), Luo et al. (2002, ch. 47), Horovitz and Sánchez-Villagra (2003, ch. 160).

 61. Hypoconulid: absent (0), in median position (1), or lingually displaced, being closer to entoconid (2). Springer et al. (1997, ch. 35), Rougier et al. (1998, ch. 52), Wroe et al. (2000, ch. 28), Wible et al. (2001, ch. 52), Horovitz and Sánchez-Villagra (2003, ch. 168).

 62. Hypoconulid of last molar: short and erect (0) or tall and sharply recurved (1). Rougier et al. (1998, ch. 53), Wible et al. (2001, ch. 53).

 63. Hypoconulid notch: present (0) or absent (1). Wroe et al. (2000, ch. 30).

 64. Entoconid: absent (0), smaller than (1), or subequal to larger than (2) hypoconid and/or hypoconulid. Marshall (1987, ch. 12), Rougier et al. (1998, ch. 54), Wroe et al. (2000, ch. 41), Wible et al. (2001, ch. 54).

 65. Entoconid shape: conical (0) or rather bladed and forming a prominent wall lingual to the talonid basin (1). Springer et al. (1997, ch. 20).

 66. Entoconid and metaconid relative positions: distant (0), close (1), or entoconid at an extreme posterior position (2).

 67. Hypoconid: small (0), large and labially salient (1), large but not salient labially (2).

 68. Entocristid: weakly developed (0), well-developed and sharp (1).

 69. Posthypocristid: transverse (0) or not (1) with respect to long axis of the tooth (penultimate inferior molar considered when available).

 70. Labial postcingulid: absent (0) or present (1). Cifelli (1993), Rougier et al. (1998, ch. 55), Wroe et al. (2000, ch. 36), Wible et al. (2001, ch. 55).

 71. Postcingulid in m4: present (0) or absent (1). Wroe et al. (2000, ch. 37).

 72. Well-developed sulcus formed by anterior cingulid: absent (0) or present (1). Reig et al. (1987, ch. 19), Wroe et al. (2000, ch. 31).

 73. Metaconid and paraconid positions: metaconid at extreme lingual margin (0) or aligned with paraconid (1). Rougier et al. (1998, ch. 56), Wible et al. (2001, ch. 56).

 74. Protocristid orientation to lower jaw axis: oblique (0) or transverse (1). Rougier et al. (1998, ch. 57), Wroe et al. (2000, ch. 35), Wible et al. (2001, ch. 57). Phascogale and Dasycercus are scored 0, contra Rougier et al. (1998) and Wible et al. (2001), since all the studied dasyurid specimens have an oblique protocristid.

 75. Size of paraconid relative to that of protoconid in m1: subequal (0), lower (1), or greatly reduced (2). Rougier et al. (1998, ch. 58), Wroe et al. (2000, ch. 34), Wible et al. (2001, ch. 58). This character is modified from Rougier et al. (1998) and Wible et al. (2001), since what they defined as a paraconide low and confluent with precingulid seems inappropriate. Thus, the paraconid size is here defined by comparison with that of another cusp (i.e. protoconid).

 76. Protoconid height: tallest cusp on trigonid (0) or subequal to para- and/or metaconid (1). Reig et al. (1987, ch. 18), Rougier et al. (1998, ch. 59), Wible et al. (2001, ch. 59), Luo et al. (2003, ch. 57).

 77. Metaconid: absent (0) or present (1). Wroe et al. (2000, ch. 33). The molars of borhyaenoids and thylacinids are well adapted to slicing and crushing. The protocone of the upper molars is reduced, bearing a slightly trenchant posterior edge, and the metacone is very enlarged, creating a long slicing postmetacrista. There is no metaconid in the lower molars, and the slicing edge passes from the paraconid, through the protoconid, to the hypoconid.

 78. Relative height of metaconid and paraconid: metaconid lower than paraconid (0), subequal (1), or metaconid taller than paraconid (2) (molars other than the first considered when available). Reig et al. (1987, ch. 17), Rougier et al. (1998, ch. 60), Wible et al. (2001, ch. 60), Luo et al. (2003, ch. 58).

 79. Size of metaconid on m1 relative to that of the other molars: not reduced (0) or reduced (1). Wroe et al. (2000, ch. 32).

 80. Relative lengths of para- and protocristids: subequal (0), paracristid > protocristid (1), or protocristid > paracristid (2).

 81. Para- and protocristids: rounded (0), or bladed and sharp (1).

 82. Last lower molar size relative to penultimate molar: subequal (0) or smaller or lost (1). Reig et al. (1987, ch. 14), Cifelli (1993), Rougier et al. (1998, ch. 61), Wible et al. (2001, ch. 61).

 83. Space between last lower molar and coronoid process: present (0) or absent (1). Rougier et al. (1998, ch. 63), Wible et al. (2001, ch. 63).

Lower jaw

 84. Masseteric fossa: restricted dorsally by crest reaching condyle (0) or extended ventrally to lower margin of dentary (1). Rougier et al. (1998, ch. 67), Wible et al. (2001, ch. 67).

 85. Posterior shelf of masseteric fossa: absent (0) or present (1). Rougier et al. (1998, ch. 68), Wible et al. (2001, ch. 68).

 86. Condyle position relative to tooth row: above (0), very high (1), or below (2). Rougier et al. (1998, ch. 72), Wible et al. (2001, ch. 72).

 87. Angular process medially inflected: absent (0) or present (1). Springer et al. (1997, ch. 47), Rougier et al. (1998, ch. 73), Wible et al. (2001, ch. 73), Luo et al. (2002, ch. 8), Sánchez-Villagra and Wible (2002, ch. 4).

 88. ‘Coronoid’ facet: present (0) or absent (1). Rougier et al. (1998, ch. 76), Wible et al. (2001, ch. 76), Luo et al. (2003, ch. 11).

 89. Two large mental foramina, one under second and third premolars and the other under first and second molars: absent (0) or present (1). Rougier et al. (1998, ch. 77), Wible et al. (2001, ch. 77).

Skull

 90. Premaxilla, palatal process does not (0) or does reach canine alveolus (1). Rougier et al. (1998, ch. 79), Wible et al. (2001, ch. 79), Horovitz and Sánchez-Villagra (2003, ch. 203), Luo et al. (2003, ch. 370).

 91. Premaxilla, facial process does not (0) or does (1) reach the nasal. Rougier et al. (1998, ch. 80), Wible et al. (2001, ch. 80), Luo et al. (2003, ch. 369).

 92. Posteriormost point of premaxillo-nasal contact: anterior or at the canine (0) or posterior to the canine (1). Springer et al. (1997, ch. 52), Horovitz and Sánchez-Villagra (2003, ch. 181).

 93. Shape of nasals relative to anterior edge of the orbits: posteriorly expanded (0) or not posteriorly expanded (1). Asher (1999, ch. 32), Wroe et al. (2000, ch. 71), Horovitz and Sánchez-Villagra (2003, ch. 200).

 94. Exit(s) of infraorbital canal: multiple (0) or single (1). Rougier et al. (1998, ch. 82), Wible et al. (2001, ch. 82), Luo et al. (342).

 95. Flaring of cheeks behind infraorbital foramen, as seen in ventral view: present (0) or absent (1). Rougier et al. (1998, ch. 83), Wible et al. (2001, ch. 83).

 96. Naso-frontal suture with medial process of frontals wedged between nasals: present (0) or absent (1). Rougier et al. (1998, ch. 84), Wible et al. (2001, ch. 84), Luo et al. (2003, ch. 351).

 97. Frontal-maxillary contact: absent (0) or present (1). Archer (1982, p. 459), Springer et al. (1997, ch. 55), Rougier et al. (1998, ch. 86), Asher (1999, ch. 31), Muizon (1999, p. 501), Wroe et al. (2000, ch. 74), Wible et al. (2001, ch. 86), Horovitz and Sánchez-Villagra (2003, ch. 183). This character is the counterpart of coding a nasal/lacrimal contact.

 98. Lacrimal tubercle: present (0) or absent (1). Novacek (1986, ch. 24), Rougier et al. (1998, ch. 87), Wible et al. (2001, ch. 87), Horovitz and Sánchez-Villagra (2003, ch. 184). The orbicularis oculi muscle is attached to a tubercle on the lacrimal bone at the margin of the rostromedial angle of the orbit.

 99. Palatal vacuities: absent, or just small foramina (0), or present (1). Reig et al. (1987, ch. 27), Rougier et al. (1998, ch. 93), Wroe et al. (2000, ch. 47), Wible et al. (2001, ch. 93), Horovitz and Sánchez-Villagra (2003, ch. 202), Luo et al. (2003, ch. 371).

100. Palatal expansion behind last molar: absent (0) or present (1). Rougier et al. (1998, ch. 94), Wible et al. (2001, ch. 94).

101. Postpalatine torus: absent (0) or present (1). Rougier et al. (1998, ch. 95), Wible et al. (2001, ch. 95).

102. Minor palatine (postpalatine) foramen: small (0) or large (with thin, posterior bony bridge) (1). Archer (1984), Wroe (1997), Rougier et al. (1998, ch. 97), Wroe et al. (2000, ch. 48), Wible et al. (2001, ch. 97), Horovitz and Sánchez-Villagra (2003, ch. 204). The minor palatine foramen (Wible and Rougier 2000) is the posterolateral foramen of Osgood (1921) and the posterolateral palatine foramen of Wroe et al. (2000).

103. Palatine reaches infraorbital canal: present (0) or absent (1). Rougier et al. (1998, ch. 98), Wible et al. (2001, ch. 98).

104. Pterygoids contact on midline: present (0) or absent (1). Rougier et al. (1998, ch. 99), Wible et al. (2001, ch. 99), Luo et al. (2003, ch. 337).

105. Orbitotemporal canal: present (0) or absent (1). Rougier et al. (1998, ch. 103), Wible et al. (2001, ch. 103). The orbitotemporal canal of extant mammals transmits the ramus superior from its union with the arteria diploëtica magna forward to the orbit, where it emerges as the ramus supraorbitalis (Rougier et al. 1992; Wible and Hopson 1995). An orbitotemporal canal is widely present among Mesozoic mammals, but is absent in metatherians (Rougier et al. 1998).

106. Squama of squamosal: absent (0) or present (1). Rougier et al. (1998, ch. 113), Wible et al. (2001, ch. 113), Luo et al. (2003, ch. 254).

107. Glenoid fossa shape: concave, open anteriorly (0) or trough-like (1). Rougier et al. (1998, ch. 115), Wible et al. (2001, ch. 115). This character is the counterpart of coding the shape of the mandible condyle, which is why character 71 of Rougier et al. (1998) and Wible et al. (2001) is not considered here. An ovoid condyle of the jaw articulates with a concave glenoid fossa, while a cylindrical condyle of the jaw articulates with a trough-like glenoid fossa.

108. Glenoid process of alisphenoid: absent (0) or present (1). Rougier et al. (1998, ch. 117), Wible et al. (2001, ch. 117).

109. Postglenoid process: absent (0) or present (1). Springer et al. (1997, ch. 65), Rougier et al. (1998, ch. 118), Wible et al. (2001, ch. 118), Luo et al. (2002, ch. 188), Horovitz and Sánchez-Villagra (2003, ch. 187). We coded the postglenoid process in Vincelestes as absent following Rougier et al. (1998), differing from the coding in Luo et al. (2002).

110. Postglenoid-suprameatal vascular system: absent (0), present, below squamosal crest (1), or present, above squamosal crest (2). Novacek (1986), Rougier et al. (1998, ch. 119), Wible et al. (2001, ch. 119). The postglenoid-suprameatal system corresponds to the suprameatal or subsquamosal foramen. Andinodelphys is scored 1 and is not polymorphic (contra Rougier et al. 1998). Mayulestes is scored 1 (?contra Rougier et al. 1998).

111. Postglenoid foramen: absent (0) or present (1). Wible and Hopson (1995), Springer et al. (1997, ch. 69), Rougier et al. (1998, ch. 120), Wible et al. (2001, ch. 120), Luo et al. (2002, ch. 189), Horovitz and Sánchez-Villagra (2003, ch. 190).

112. Tall paroccipital process of exoccipital: absent (0) or present (1). Rougier et al. (1998, ch. 135), Wible et al. (2001, ch. 135). To avoid confusion between the paroccipital process of the opisthotic (or petrosal or mastoid process of petrosal) and the paroccipital process of the exoccipital, we maintained the term ‘paroccipital process of the exoccipital’ and we employed the term ‘mastoid process’ for the process of the petrosal that bears the sternomastoid and digastric muscles.

113. Ectotympanic: ring-shaped (0) or moderately broadened (1). Springer and Woodburne (1989, ch. 6), Springer et al. (1997, ch. 64), Rougier et al. (1998, ch. 142), Wroe et al. (2000, ch. 58), Wible et al. (2001, ch. 142), Horovitz and Sánchez-Villagra (2003, ch. 186).

114. Ascending canal: present (0) or absent (1). Rougier et al. (1998, ch. 152), Wible et al. (2001, ch. 152). Kielan-Jaworowska et al. (1986) described an ascending canal for the intramural canal in multituberculates within the suture between the anterior lamina and the squamosal, dorsal to the middle ear. Wible (1989), Rougier et al. (1992), and Wible and Hopson (1995) re-examined this structure and offered a reconstruction of its major occupants, which are the ramus superior of the stapedial artery and accompanying veins.

115. Interparietal: absent (0) or present (1). Rougier et al. (1998, ch. 155), Wible et al. (2001, ch. 155).

116. Parietal-alisphenoid or squamosal-frontal contact on braincase: alisphenoid-parietal contact (0) or frontal-squamosal contact (1). Flannery and Archer (1987), Springer et al. (1997, ch. 44), Muizon (1998), Wroe et al. (2000, ch. 64), Horovitz and Sánchez-Villagra (2003, ch. 176).

117. Hypoglossal foramina: confluent with jugular foramen (0), one (1), two or more (2). Horovitz and Sánchez-Villagra (2003, ch. 205).

Petrosal

118. Complete wall separating cavum supracochleare from cavum epiptericum: absent (0) or present (1). Wible (1990, ch. 2), Wible and Hopson (1993, ch. 6), Rougier et al. (1996a, b, ch. 40; 1998, ch. 128), Wible et al. (2001, ch. 128), Luo et al. (2002, ch. 203), Sánchez-Villagra and Wible (2002, ch. 11).

119. Cavum epiptericum floored by: petrosal (0), petrosal and alisphenoid (1), or primarily or exclusively by alisphenoid (2). Wible and Hopson (1993, ch. 4), Rougier et al. (1996a, b, ch. 35; 1998, ch. 109), Wible et al. (2001, ch. 109). In Andinodelphys, Pucadelphys and Mayulestes, the trigeminal ganglion is posteriorly floored by the anterior lamina of the petrosal. The contribution of the petrosal to the floor of the cavum epiptericum being substantial, these taxa are coded 1. Petrosal Types I and II from Itaboraí exhibit such a condition, albeit in a minor way (i.e. the anterior lamina is less expanded) and are thus coded 2. Monotremes exhibit a special configuration. The platypus has a cavum epiptericum that is partially floored by the lamina obturans and appositional bone from the petrosal (Wible and Hopson 1993). The echidna has a complete floor formed by the lamina obturans, appositional bone from the petrosal, ectopterygoid, and palatine (Kuhn and Zeller 1987). Though these configurations do not correspond exactly to any of the scores, both genera of monotremes were scored 0 because the petrosal is part of the floor of the cavum epiptericum.

120. Fossa subarcuata: smaller than its aperture (i.e. conical shape) (0) or larger (i.e. spherical shape) (1). Ladevèze (2004, ch. 2).

121. Posterior exposure of the pars mastoidea: dorsoventrally elongated and approximately flat (0) or rounded and bulbous owing to the excavation of the fossa subarcuata (1). Ladevèze (2004, ch. 3).

122. Expansion of the crista petrosa that forms a thin lamina covering the anterolateral part of the fossa subarcuata: absent (0) or present (1).

123. Anterior lamina of petrosal exposure on the lateral wall of the braincase: present and large (0), or rudimentary (1), or absent (2). Wible (1990, ch. 14), Wible and Hopson (1993, ch. 1, 2), Rougier et al. (1998, ch. 108), Wible et al. (2001, ch. 108).

124. Anterior lamina of petrosal: shows a large depression that may have received a part of temporal lobe of the brain and the posterior part of the trigeminal ganglion (0) or shows no depression (1). Ladevèze (2004, ch. 6).

125. Internal acoustic meatus and fossa subarcuata: subequal and separated by a sharp wall (0), internal acoustic meatus narrower than the opening of the fossa subarcuata and separated from the latter by a thick shelf of bone (1). Tachyglossus is scored ? as the configuration is peculiar: the fossa subarcuata forms a shallow but large fossa.

126. Internal acoustic meatus: deep with thick prefacial commissure (0) or shallow with thin prefacial commissure (1). Rougier et al. (1998, ch. 153), Wible et al. (2001, ch. 153).

127. Deep groove for internal carotid artery excavated on anterior pole of promontorium: absent (0) or present (1). Muizon et al. (1997), Rougier et al. (1998, ch. 148), Wible et al. (2001, ch. 148), Sánchez-Villagra and Wible (2002, ch. 16), Horovitz and Sánchez-Villagra (2003, ch. 220).

128. Deep, large fossa for the tensor tympani muscle excavated on the anterolateral aspect of promontorium, creating a battered ventral surface of the promontorium: absent (0) or present (1). Wible (1990, ch. 12), Luo et al. (2002, ch. 221).

129. Epitympanic wing of petrosal: absent (0) or present (1). Rougier et al. (1998, ch. 122); Wible et al. (2001, ch. 122). This term is used for outgrowths of any basicranial bones that contribute to the tympanic roof (MacPhee 1981). An epitympanic wing extends medially from the promontorium in many therians (Rougier et al. 1996a, 1998).

130. Epitympanic wing of petrosal: flat (0), undulating (1), or confluent with bulla (2). Rougier et al. (1998, ch. 122); Wible et al. (2001, ch. 122). Notogale is scored 0 (i.e. epitympanic wing flat) contra Rougier et al. (1998) and Wible et al. (2001) who coded an epitympanic wing undulated for borhyaenids.

131. Lateral flange: large and lateral to promontorium (0) or greatly reduced or absent (1). Wible (1990, ch. 3), Rougier et al. (1996a, b, ch. 29; 1998, ch. 126), Wible et al. (2001, ch. 126). In non-therian mammals, a large shelf of bone extends lateral to and on the whole length of the promontorium. This is the lateral trough, the lateral edge of which is downturned to form the lateral flange (Wible et al. 1995). In therians, the lateral flange is either greatly reduced or absent (Wible et al. 1995), while a lateral trough persists in Andinodelphys, Pucadelphys and Mayulestes (pers. obs.).

132. Broad shelf of bone surrounding fenestra cochleae and making a separation between it and aqueductus cochleae: absent (0) or present (1). Such a structure was observed on the petrosals of Pucadelphys and Andinodelphys. In monotremes, the fenestra cochleae is not separated from the jugular foramen.

133. Rostral tympanic process of petrosal: absent (0) or present as a distinct crest or erect process (1). Wible (1990, ch. 9), Rougier et al. (1998, ch. 130), Wible et al. (2001, ch. 130). The presence of a low, tiny ridge or tubercle, anterolateral to the fenestra vestibuli is not regarded as equivalent to the development of a tympanic process, since it is not a process, i.e. a raised shelf of bone. Pediomys and Didelphodon are thus coded 0. The monotreme Tachyglossus exhibits a large, broad process on the ventral surface of the promontorium. Nevertheless, the ontogenies of the promontorium processes in monotremes and metatherians are not entirely comparable. The process observed in monotremes results from a co-ossification of Reichert's cartilage and the cochlear capsule (Kuhn 1971; Presley 1980), whereas that of metatherians results from a periosteal outgrowth from the ossifying cochlear capsule (Maier 1987). Thus, Tachyglossus is scored here as not possessing a rostral tympanic process of petrosal.

134. Rostral tympanic process of petrosal: forms an anterolaterally directed wing, sometimes contacting the ectotympanic, that does not (0), or does (1) extend over the whole length of the promontorium. Sánchez-Villagra and Wible (2002, ch. 7), Horovitz and Sánchez-Villagra (2003, ch. 215).

135. Tympanic aperture of hiatus Fallopii: ventral (0), dorsal (1), or intermediate (2). Sánchez-Villagra and Wible (2002, ch. 12), Horovitz and Sánchez-Villagra (2003, ch. 218).

136. Stylomastoid foramen: absent (0) or present (1). Archer (1976a), Wroe et al. (2000, ch. 54). The monotreme Tachyglossus has an open facial sulcus that runs into a stylomastoid foramen. This configuration is different from that defined by Archer (1976a) and Wroe et al. (2000), i.e. a facial canal that opens onto a stylomastoid foramen within the petrosal.

137. Inferior petrosal sinus: intrapetrosal (0) or between petrosal, basisphenoid and basioccipital (1) or endocranial (2). Rougier et al. (1996a, b, ch. 42; 1998, ch. 151), Wible et al. (2001, ch. 151).

138. Mastoid exposure: large (0), narrow (1), or reduced pars mastoidea, internal to the braincase and wedged between the squamosal and exoccipital (2). Ladevèze (2004, ch. 18).

139. Mastoid tympanic process: large and vertical (0), small, slanted and node-like, on the posterolateral border of the stylomastoid notch and continuous with squamosal (1), or indistinct to absent (2). Wible (1990, ch. 7), Rougier et al. (1998, ch. 131), Wroe et al. (2000, ch. 67), Wible et al. (2001, ch. 131), Ladevèze (2004, ch. 19). Two scores differ from Wible et al. (2001): Pucadelphys and Andinodelphys are scored as having a slanted mastoid tympanic process, quite well-developed and node-like, that forms a posterolateral wall for the stylomastoid notch and follows the squamosal. Prokennalestes was scored unknown, because that part of the petrosal is broken, and even if Wible et al. (2001) inferred that the mastoid tympanic process was probably well developed and vertical. Deltatheridium was scored unknown as we could not observe this structure.

140. Caudal tympanic process of petrosal: absent (0) or present (1). Wible (1990, ch. 8), Wible and Hopson (1993, ch. 26), Rougier et al. (1996a, b, ch. 18; 1998, ch. 132), Wible et al. (2001, ch. 132), Luo et al. (2002, ch. 200). According to MacPhee (1981), the term ‘caudal tympanic process of petrosal’ is applied to the process that arises from the tympanic surface of the pars canalicularis in the area posterior to the cochlear fossula and medial to the stylomastoid opening. As for the character of the rostral tympanic process of the petrosal (ch. 133), the presence of a low, tiny ridge or tubercle in the area of the caudal tympanic process is not regarded as a tympanic process.

141. Caudal tympanic process of petrosal: forms a small crest that does not wholly floor the postpromontorial sinus (0) or forms an expanded lamina that does floor it (1). Ladevèze (2004, ch. 20).

142. Petrosal plate: absent (0) or present (1). Wible (1990, ch. 8), Sánchez-Villagra and Wible (2002, ch. 8), Horovitz and Sánchez-Villagra (2003, ch. 216), Ladevèze (2004, ch. 21). The petrosal plate (sensuVan der Klaauw 1931) is formed by the joined caudal and rostral tympanic processes of the petrosal. This feature is linked to a pneumatization of the petrosal (see Dromiciops and dasyurids) and is also found in the didelphid Caluromys and in the fossil microbiotheriid Microbiotherium (Marshall 1982).

143. Fossa incudis and epitympanic recess: continuous (0) or separated by a distinct ridge (1). Rougier et al. (1998, ch. 137), Wible et al. (2001, ch. 137). An epitympanic recess is found in multituberculates, Vincelestes and therians, whereas a fossa incudis is more widely distributed among mammaliaforms (Rougier et al. 1996a). Some metatherians have these two fossa separated by a distinct ridge (i.e. Didelphodon, Pediomys, borhyaenids, and some dasyurids; Rougier et al. 1998).

144. Petrosal crest: absent (0) or present (1). Archer (1976a), Muizon (1999, ch. 15), Ladevèze (2004, ch. 23). Archer (1976a) defined the petrosal crest as the bony edge that separates the anterior part of the epitympanic recess from the posterior part of the hypotympanic sinus of the alisphenoid.

145. Petrosal contribution to the lateral wall of the epitympanic recess: absent (0) or present (1). Ladevèze (2004, ch. 24). In extant metatherians, the epitympanic recess is laterally bordered by the squamosal, which supports the external acoustic meatus. On isolated petrosals, a shelf of bone of the petrosal bounds laterally the epitympanic recess, but is covered by the squamosal in entire skulls. In Andinodelphys, Pucadelphys and Mayulestes, the epitympanic recess is divided and lies in both the squamosal and the petrosal.

146. Petrosal contribution to the lateral wall of the epitympanic recess: massive, large shelf of bone, sometimes rounded (0), slender and triangular (1), or forming a thin lamina (2). Ladevèze (2004, ch. 24).

147. Prootic canal: present (0) or absent (1). Wible (1990, ch. 17), Wible and Hopson (1993, ch. 18), Wible and Hopson (1995), Wroe et al. (2000, ch. 77), Sánchez-Villagra and Wible (2002, ch. 9, 10).

148. Prootic canal: large with endocranial opening (0) or reduced with intramural opening (1). Wible (1990, ch. 18), Wible and Hopson (1993, ch. 19), Rougier et al. (1998, ch. 124), Wible et al. (2001, ch. 124).

149. Imprint of the transverse sinus bifurcation on the petrosal: absent (0) or present (1). Ladevèze (2004, ch. 27). This character is only available on isolated petrosals.

150. Foramina on the sigmoid sinus and/or prootic sinus, apparently connecting both vessels (i.e. sigmoid sinus vein): absent (0) or present (1)

151. Vascular groove medially adjacent to prootic sinus sulcus, on the pars mastoidea (i.e. prootic sinus vein or connection): absent (0) or present (1).

152. Posttemporal sulcus on the squamosal surface of the petrosal: present (0) or absent (1). Wible and Hopson (1995), Rougier et al. (1998, ch. 144), Wible et al. (2001, ch. 144), Sánchez-Villagra and Wible (2002, ch. 13), Horovitz and Sánchez-Villagra (2003, ch. 219).

153. Posttemporal notch/foramen: present (0) or absent (1). Wible (1990, ch. 24), Rougier et al. (1998, ch. 144), Wible et al. (2001, ch. 144), Luo et al. (2002, ch. 251), Sánchez-Villagra and Wible (2002, ch. 14).

154. Transpromontorial sulcus: present (0) or absent (1). Wible (1986), Rougier et al. (1998, ch. 146), Wible et al. (2001, ch. 146), Luo et al. (2002, ch. 220).

155. Sulcus for stapedial artery: present (0) or absent (1). Wible (1987, 1990, ch. 16), Rougier et al. (1998, ch. 147), Wible et al. (2001, ch. 147), Luo et al. (2002, ch. 219), Horovitz and Sánchez-Villagra (2003, ch. 221).

156. Cochlear coiling: absent or less than 300 degrees (0), or fully coiled (more than 360 degrees) (1). Rougier et al. (1998, ch. 129), Wible et al. (2001, ch. 129), Luo et al. (2002, ch. 194), Horovitz and Sánchez-Villagra (2003, ch. 222).

Additional basicranial characters

157. Primary foramen ovale bounded by: lamina obturans/petrosal (0), alisphenoid and petrosal (1), alisphenoid or alisphenoid and squamosal (2). Gaudin et al. (1996), Wroe (1997), Rougier et al. (1998, ch. 111), Muizon (1999, ch. 7), Wroe et al. (2000, ch. 50), Wible et al. (2001, ch. 111), Horovitz and Sánchez-Villagra (2003, ch. 194).

158. Secondary foramen ovale: absent (0) or present (1). Archer (1976a), Gaudin et al. (1996), Wroe (1997), Wroe et al. (2000, ch. 51–53).

159. Alisphenoid tympanic process: absent (0) or present (1). Rougier et al. (1998, ch. 121), Wible et al. (2001, ch. 121).

160. Alisphenoid tympanic process well developed and encloses a hypotympanic sinus: absent (0) or present (1). Springer et al. (1997, ch. 62), Wroe et al. (2000, ch. 57), Luo et al. (2002, ch. 222), Horovitz and Sánchez-Villagra (2003, ch. 185).

161. Posterior expansion of the alisphenoid tympanic process that completely floors the (hypo)-tympanic sinus in ventral view (i.e. covers the petrosal and squamosal): absent (0) or present (1). Springer et al. (1997, ch. 62), Wroe et al. (2000, ch. 57), Luo et al. (2002, ch. 222), Horovitz and Sánchez-Villagra (2003, ch. 185).

162. Tympanic sinus formed in the lateral trough (or anterolateral expansion of the pars canalicularis): absent (0) or present (1). A large fossa excavated on the ‘lateral trough’ or on the lateral expansion of the pars canalicularis was observed on the petrosals of the Tiupampan taxa Mayulestes, Pucadelphys, and Andinodelphys, and Itaboraían Type II, and was interpreted as a tympanic sinus. This deep sinus is a large, subtriangular fossa that lies anterolateral to the fossa for the tensor tympani muscle and anterior to the epitympanic recess and petrosal crest. In Mayulestes it corresponds to the posterior half of what Muizon (1998) named ‘alisphenoid hypotympanic sinus’ even if the term ‘hypotympanic’ is not appropriate in this case. In Mayulestes this sinus is completed by the alisphenoid and squamosal and approximates the condition of extant metatherians, while in Pucadelphys and Andinodelphys it is completed by the posterior edge of the squamosal (medial process). In the borhyaenids Notogale and Sallacyon, such a sinus is found on the lateral trough, and is completed by the alisphenoid in a hypotympanic sinus.

163. (Hypo)-tympanic sinus formed by squamosal, petrosal, and alisphenoid: absent (0) or present (1). Muizon (1994, 1998), Rougier et al. (1998, ch. 140), Wroe et al. (2000, ch. 56), Wible et al. (2001, ch. 140), Luo et al. (2002, ch. 223; 2003, ch. 312). According to Rougier et al. (1998) and Wible et al. (2001), the tympanic sinus is formed by the squamosal, petrosal, and alisphenoid in Didelphodon, Mayulestes and borhyaenids, and by the petrosal and alisphenoid in Marsupialia. In addition, several metatherians have an alisphenoid hypotympanic sinus, but the contributing elements are uncertain. Included are Eodelphis, Pediomys, Turgidodon, Asiatherium, and the Gurlin Tsav skull (see (Szalay and Trofimov 1996).

164. (Hypo)-tympanic sinus formed by petrosal and alisphenoid: absent (0) or present (1).

165. (Hypo)-tympanic sinus formed by alisphenoid and petrosal, or by alisphenoid only: absent (0) or present (1). Muizon (1994, 1998), Rougier et al. (1998, ch. 140), Wroe et al. (2000, ch. 56), Wible et al. (2001, ch. 140), Luo et al. (2002, ch. 223; 2003, ch. 312).

166. Tympanic cavity partially or entirely enclosed by bony structure (i.e. auditory bulla): absent (0) or present (1).

167. Tubal foramen just posterior and ventral to the foramen ovale: absent (0) or present (1). Wroe (1997, 1999), Wroe et al. (2000, ch. 70).

168. Transverse canal: absent (0) or present (1). Archer (1976a), Wroe (1999), Rougier et al. (1998, ch. 104), Muizon (1999, ch. 8), Wroe et al. (2000, ch. 63), Wible et al. (2001, ch. 104), Sánchez-Villagra and Wible (2002, ch. 1), Horovitz and Sánchez-Villagra (2003, ch. 196, 197).

169. Squamosal epitympanic sinus: absent (0) or present (1). Archer (1976a), Wroe et al. (1998, 2000, ch. 55), Wroe (1999).

170. Medial process of squamosal in tympanic cavity: absent (0) or present (1). Muizon et al. (1997), Rougier et al. (1998, ch. 141), Muizon (1999, ch. 1), Wroe et al. (2000, ch. 59), Wible et al. (2001, ch. 141), Horovitz and Sánchez-Villagra (2003, ch. 199).

171. Jugular foramen very large, at least three times larger than the fenestra cochleae: absent (0) or present (1). Rougier et al. (1996a, b, ch. 14; 1998, ch. 149), Wible et al. (2001, ch. 149). Monotremes were scored as inapplicable data because the fenestra cochleae is not separated from jugular foramen.

172. Jugular foramen: confluent with opening for inferior petrosal sinus (0) or separated from opening for inferior petrosal sinus (1). Rougier et al. (1998, ch. 150), Wible et al. (2001, ch. 150).

Data matrix showing the distribution of 172 dental and cranial (including petrosal) characters among 26 taxa: ?, non-preservation; -, non-applicable character state; polymorphism: A = 0 + 1, B = 1 + 2, C = 0 + 2.

Vincelestes
3010210001000100100120001101100200000000000010012000102-10010?10--1-00
100010110001110000000000000000000000000000?00?20000?0000000-000-000000
-0000-00000000000000-00000000000   
Prokennalestes
01100???????--000000000010-010000011100211000????010011001111101002100
100110120?000000001????11???0?????01??11???0???01?0010110010000-000???
?00010000?000001?????0000?0???0?   
Ornithorhynchus
3?2?22??????????????????????????????????????2?????????????????????????
?????????????100010?0?11001000000000000000000?0001000011000-000-002000
-0-00-0010010-0010000000000000-0   
Tachyglossus
3?2?22??????????????????????????????????????2?????????????????????????
?????????????20?010?0?11001001000100?00000000?010-0000?1000-100-002020
-0-00-0010010-1011000000000000-0   
Deltatheridium
2002110?1?11000110102001-101100200110002100011012000012?101110?100??00
1?1010101111110011111?01000000010??110120??1???1100020100010100--01??1
00000-0100000111?????0000?0???01   
Didelphodon
22021???????01010110C200110220110021211210001?????10220002122002002001
1011011001100110?11?????1??????????11010???1???1200020000111100-?00001
001???011?000111?????1100?0??0?1   
Pediomys
20011???????0000020000A001022011002121121100??????10110012122002001111
001120120110011A?11????1????1111??1????????1???1200021000111100-201001
0010100110000111?????0????0????1   
Asiatherium
22011?????1000010000100000-11010012120121111??????00010002122002002011
0001101201100101?11????10???1101???11?1??0????????????????12??10?0??11
00???????????1??B?1100???1?10???   
Gurlin Tsav Skull
200211?0??11??010010220010-22010012100121000??????????????????????????
???????????????????1100100001101??111012?????0??B?0?2???????1???????1?
?0??????????????B?1100???1?00???   
Notogale
2?0111????1???0?0110????????201?0?211102100011?1???00120111?21011-2101
10--1000--00??????1??????????????11????1???1?02???????1?1110100-00121?
?001??1-0????1112111111001011111   
Sallacyon
??01??????????0001102101-0-22001012111021000?????????12??11?2??11-2001
?0--?000--0???????????0??100000???1?101?1??1?0?120?02?0??11?100-?012?1
0001??1-?????1111111?1100????111   
Mayulestes
2001100000110000011010001102201001211102100001012110012011102101002001
1101101110100?????1111010100011100?1111110?1002?1?0?2???1110100-?01111
0001101-?????1111000-11000000111   
Pucadelphys
2001100011110000010012101112211011212111100001012010010012122102002101
101111120000011011111001001001110?11111110?11021B10010011110110-101011
000110010A11A1111000-10100000111   
Andinodelphys
2001100000110000010002101112211011212112100001012010010012122101002001
001111120000011011?11001011111110111111110?11021B000?0111110110-101011
00011001001111111000-10100010111   
MHNC 8369
??????????????????????????????????????????????????????????????????????
???????????????????????????????????????????????12100?0011010100-101021
0000100100011111?????0??????????   
Type I
??????????????????????????????????????????????????????????????????????
???????????????????????????????????????????????12110200110101010101021
0001110101111111?????0??????????   
Type II
??????????????????????????????????????????????????????????????????????
???????????????????????????????????????????????0210020011110100-101011
0001110100000111?????1??????????   
Didelphis
2001100101110101010022101112211011011112100001002002112012122002101010
1111101201000111110110011111111111111111110110B12100211100101010201121
00011101100001111111000011010011   
Marmosa
2001100101110001010022101112211011011112100001002002012012122002101010
1111101201000111110110011111111111111111110110212100210100101010201121
00011101100001111A11000011010001   
Metachirus
2001100101110001010022001112211011211112100001002002012012122002101010
1111101201000111110110011111111111111111110111212100211100101010201121
00011101100001111111000011010000   
Caenolestes
20001112021000101301230000-22?110?111102000010010210010112022002121100
0111011212110111110111011111111011111111100110102110200100101010101021
10001101010111111011000011010000   
Dromiciops
220010101100000011002101-0-2211200012112100000001200111112122002021110
1011201201110111110010011111111011111112101111212111211100121011111021
110?121-?1?1111120111000111A0000   
Phascogale
2001110002000000110A220000-221111101111210011101100?01B112102002002001
1010101201100111110111010111111001111112111110212111210100101011111021
1100121-011111111011100011111001   
Dasycercus
20011100020000011201220000-221111101111210011101210?0?2112102002002001
1010101201100111110111010111111001111112111110212?112?0100101011111021
11101?1-????11111011100011111001   
Perameles
2001101111101010110103000102211111112102100111110200010112102012111110
1111101200000111110011111111111101111111101111212100210100101011001021
1000121-000111111111100011011011   
Echymipera
20011111111010A0010013000102211211212100100111110200020112?02012111110
111111120?000111110011111111111001111111101111B?20002?010010101100?021
00001?1-??0111111111000011010011   

Ancillary