Novitates PUBLISHED by the AMERICAN MUSEUM of NATURAL HISTORY CENTRAL PARK WEST at 79TH STREET, NEW YORK, N.Y

Home , Adelobasileus, Dinnetherium, Gobiconodon, Gobiconodontidae, Holotheria, Kuehneotherium

AMERICANt MUSEUM Novitates PUBLISHED BY THE AMERICAN MUSEUM OF NATURAL HISTORY CENTRAL PARK WEST AT 79TH STREET, NEW YORK, N.Y. 10024 Number 3183, 38 pp., 6 figures November 22, 1996

Basicranial Anatomy of Priacodon fruitaensis (Triconodontidae, Mammalia) from the Late Jurassic of Colorado, and a Reappraisal of Mammaliaform Interrelationships

GUILLERMO W. ROUGIER,"2 JOHN R. WIBLE,3 AND JAMES A. HOPSON4

ABSTRACT Left and right petrosals and exoccipitals of the node is Mammalia, formed by the common an- triconodontid Priacodon fruitaensis from the Late cestor of monotremes and therians and all its de- Jurassic Morrison Formation of western Colorado scendants. A petrosal from the Early Cretaceous are described. To elucidate the phylogenetic re- of Mongolia, considered previously to be related lationships of Priacodon, a cladistic analysis of either to triconodonts or Prototribosphenida, is 51 basicranial characters across 18 ingroup and shown here to be allied with the latter. The phy- two outgroup taxa is performed. A monophyletic logenetic relationships of a number of Jurassic Triconodontidae, including Priacodon, Trioraco- taxa traditionally held to be members of Tricon- odonta are weakly resolved, resulting in a para- don, and Triconodon is identified in the six equal- phyletic series: (Adelobasileus (Sinoconodon ly most parsimonious trees obtained; a fourth tri- ((Morganucodon + Dinnetherium) (Megazostro- conodont from the Early Cretaceous Cloverly For- don (Haldanodon + Mammalia))))). Basicranial mation is a sister taxon to this grouping. In three characters employed in previous cladistic analyses of the six most parsimonious trees, the four tri- are discussed, and the implications of our phylo- conodonts are more closely related to therians genetic proposal for taxa known primarily from than are multituberculates; the reverse is true in dentitions, such as Kuehneotherium, are exam- the remaining three trees. The next more inclusive ined.

' Frick Research Fellow, Department of Vertebrate Paleontology, American Museum of Natural History. 2Museo Argentino de Ciencias Naturales, CONICET, Argentina. I Research Associate, Department of Mammalogy, American Museum of Natural History; Department of Anatom- ical Sciences and Neurobiology, School of Medicine, University of Louisville, Louisville, KY 40292. 4Department of Organismal Biology and Anatomy, University of Chicago, Chicago, IL 60637.

INTRODUCTION Traditionally the order Triconodonta in- some with associated postcranial remains, cluded a diverse assemblage of Mesozoic from the Early Cretaceous Cloverly Forma- mammals usually with three anteroposteri- tion of Montana. Basicrania of these speci- orly aligned, main cusps in the postcanine mens have been figured and partially de- teeth (Hopson and Crompton, 1969; Jenkins scribed by Crompton and co-workers and Crompton, 1979). Among the taxa in- (Crompton and Jenkins, 1979; Crompton and cluded in this grouping were the Early Ju- Sun, 1985; Crompton and Luo, 1993). Re- rassic "morganucodonts," Sinoconodon, garding gobiconodontids, Jenkins and Schaff Dinnetherium, and four Late Jurassic/Creta- (1988) described two partial skeletons found ceous families-Triconodontidae, Gobicon- in association with upper and lower jaws of odontidae, Austrotriconodontidae, and Am- Gobiconodon ostromi also from the Cloverly philestidae (Simpson, 1925, 1928, 1929; Formation of Montana. A fragmentary squa- Hopson, 1970; Jenkins and Crompton, 1979; mosal and frontal were found with one of the Bonaparte, 1986, 1992; Kielan-Jaworowska, skeletons. This material included postcranial 1992). However, Amphilestidae has alterna- remains originally reported by Jenkins and tively been suggested to have therian affini- Crompton (1979) to be from an unnamed ties because of similar occlusal patterns amphilestid. (Simpson, 1961; Patterson and Olson, 1961; A new triconodontid species, Priacodon Mills, 1971). Cladistic analyses of mammal- fruitaensis, was named by Rasmussen and iamorph relationships have shown that these Callison (1981) based on a left lower jaw taxa do not form a monophyletic assemblage, with teeth from the Late Jurassic Morrison with the Early Jurassic forms occupying a Formation of western Colorado. Subsequent basal position among mammaliamorphs work on the original block revealed several (Rowe, 1988, 1993; Wible and Hopson, additional bones in association with the left 1993; Wible et al., 1995). The relationships lower jaw. The partial right mandible, both of the four Late Jurassic/Cretaceous families maxillae, and some postcranial elements are have not been addressed within a cladistic described elsewhere (Engelmann and Calli- framework, although Rowe (1993) reported son, in prep.). We report here on the right that Triconodontidae and Gobiconodontidae and left petrosals and exoccipitals of the ho- were identified as sister taxa in some of the lotype of P. fruitaensis. In addition to de- trees yielded by his analysis. scribing these elements, a cladistic analysis The bulk of the material attributed to the of petrosal and exoccipital anatomy address- four Late Jurassic/Cretaceous families, which es the relationships of triconodonts within we informally call triconodonts here, com- Mammaliamorpha. prises upper and lower jaws. However, additional cranial and postcranial remains of MATERIALS AND METHODS triconodontids and gobiconodontids have Rasmussen and Callison (1981) designated been reported. Regarding triconodontids, the holotype of Priacodon fruitaensis as Simpson (1928) noted fragmentary cranial LACM 120451, Natural History Museum of bones, axis, and possible radius associated Los Angeles County, Los Angeles, CA. It with dentitions of Triconodon mordax from was recovered near Fruita, Colorado, in the the Late Jurassic Purbeck Formation of En- uppermost part of the Salt Wash Member, gland. Included among the cranial elements Morrison Formation, Upper Jurassic. In the were both petrosals, a basisphenoid, and part original description, only the left lower jaw of an alisphenoid. These were redescribed by was studied. Subsequently, however, Engel- Kermack (1963) along with two petrosals at- mann and Callison (in prep.) described ad- tributed to the triconodontid Trioracodon ditional material of the holotype, the right ferox, also from the Purbeck. Jenkins and lower jaw, right and left maxillae, and sev- Crompton (1979) reported 20 dental speci- eral postcranial elements. The right and left mens of unnamed triconodontids, ranging petrosals and exoccipitals are described here. from fragmentary jaws to complete skulls, The definition of taxonomic units employed 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 3 here follows Gauthier (1986) and Rowe pendices 1 and 2). An unrooted cladistic (1987). We use the following terms in the sens- analysis was performed using the branch and es of the cited authors: Mammaliamorpha, bound algorithm of PAUP (Swofford, 1993). Mammaliaformes, Mammalia, Theriiformes, All multistate characters were treated as and Theria (Rowe, 1988), Holotheria (Hopson, unordered, and the multiple states occurring 1994), Trechnotheria and Tribosphenida (Mc- in some taxa were interpreted as uncertain- Kenna, 1975), Triconodontidae (Jenkins and ties, because in most instances the multiple Crompton, 1979), Symmetrodonta (Prothero, states resulted from incomplete specimens 1981), and Prototribosphenida (Rougier, 1993). and not from true polymorphism. The un- However, we note that Rowe's (1988) defini- rooted analysis confirmed the monophyly of tion of Mammalia-the clade including the the ingroup and the resulting cladograms most recent common ancestor of living mam- were rooted between the outgroup and in- mals plus all its descendants-is not univer- group. sally accepted. We accept Jenkins and Cromp- The basicranial anatomy of most taxa in- ton's (1979) composition of the family Tricon- cluded here has been described elsewhere, odontidae and define this taxon as the last and scores for these taxa were taken from the common ancestor of Alticonodon, Astrocono- original literature, supplemented by our own don, Priacodon, Triconodon, and Trioracodon direct observations. The literature sources plus all its descendants. The basicrania from consulted were: Simpson (1928), Kuhne the Cloverly Formation (MCZ 19969, 19973) (1956), Crompton (1958, 1964), Kermack figured and partially described by Crompton (1963), Hopson (1964), Kuhn (1971), and co-workers (Crompton and Jenkins, 1979; Maclntyre (1972), Kermack et al. (1981), Crompton and Sun, 1985; Crompton and Luo, Sun (1984), Crompton and Sun (1985), Gow 1993) have been attributed variously to tricon- (1986a, 1986b), Kielan-Jaworowska et al. odontids, triconodontines, and triconodonts. (1986), Novacek (1986), Sues (1986), Hahn Because the associated dentitions have not (1988), Rowe (1988), Zeller (1989), Wible been described, we refer to these basicrania (1990, 1991), Lillegraven and Krusat (1991), here as indeterminate Cloverly triconodonts. Rougier et al. (1992), Lillegraven and Hahn Our use of the term triconodonts for the Tri- (1993), Crompton and Luo (1993), Wible conodontidae, Gobiconodontidae, Austrotri- and Hopson (1993, 1995), Lucas and Luo conodontidae, Amphilestidae, and the indeter- (1993), Hopson and Rougier (1993), Rougier minate Cloverly triconodonts (which may ul- (1993), Luo (1994), Luo and Crompton timately be members of one of the previous (1994), Luo et al. (1995), and Wible et al. groups) is an informal one and does not imply (1995). monophyly of this assemblage, which has yet The basicrania of three taxa included here to be tested. have been only partially described or figured, To evaluate the affinities of Priacodon and scores for these were taken from the fruitaensis and other triconodonts known available published reports: the tritheledontid from basicranial remains, a cladistic analysis Pachygenelus (Wible, 1991; Allin and Hop- of petrosal and exoccipital anatomy was per- son, 1992; Hopson and Rougier, 1993; Wible formed using the method of simultaneous and Hopson, 1993; Lucas and Luo, 1993; outgroup analysis (Nixon and Carpenter, Luo, 1994; Luo et al., 1995); Dinnetherium 1993). Following recent cladistic analyses by (Crompton and Luo, 1993; Luo, 1994); and Luo (1994) and Wible et al. (1995), the in- the unnamed Cloverly triconodonts (Cromp- group was selected to include the common ton and Jenkins, 1979; Crompton and Sun, ancestor of Adelobasileus, Sinoconodon, and 1985; Crompton and Luo, 1993; Luo, 1994). Mammaliaformes plus all its descendants, In addition, observations on the endocranial and Tritheledontidae and Tritylodontidae surface of the petrosal in Sinoconodon and were chosen for the outgroups. Fifty-one the shape of the promontorium in Megazos- transformation series were identified largely trodon were taken from Luo and Crompton from the literature, including characters to re- (personal common). Finally, observations solve the outgroups, and scored across 20 from three specimens collected in the Gobi taxa (18 ingroup plus two outgroup) (see, ap- Desert by the Joint Paleontological Expedi- 4 AMERICAN MUSEUM NOVITATES NO. 3183 tions of the Mongolian Academy of Sciences served. Isolated petrosals are frequently re- and the American Museum of Natural His- covered in the fossil record, because they are tory and cataloged in the Geological Insti- composed of dense, compact bone. This is a tute, Ulaan Baatar are included here. Two of fortunate accident because, other than the these specimens are isolated petrosals from dentition, the petrosal is the only cranial el- the Early Cretaceous locality of Khoobur: ement that allows for assignment of isolated PSS-MAE-104 (referred to as Khoobur pe- specimens to particular higher-level taxa trosal 1), which was described by Wible et (Maclntyre, 1972; Wible, 1990; Wible et al., al. (1995), and PSS-MAE- 1 9 (referred to as 1995). Additionally, there is a growing da- Khoobur petrosal 2), which is to be reported tabase on petrosal morphology for systematic on elsewhere. The third specimen, PSS- studies of early mammals (Crompton and MAE- 113, is a well-preserved skull of a tae- Luo, 1993; Wible and Hopson, 1993; Luo, niolabidoid multituberculate, Kryptobaatar 1994). In contrast, the exoccipital in early dashzevegi, from the Late Cretaceous local- mammals is poorly known and, consequent- ity of Tugrugeen Shireh. ly, has played virtually no role in phyloge- The terminology for anatomical terms netic analyses to date. used here follows Wible (1987, 1990), Rou- gier et al. (1992), and Wible et al. (1995). PETROSAL Abbreviations In Recent mammals, the petrosal is the ossified otic capsule housing the organs of AMNH-DM American Museum of Natural hearing and balance. For descriptive purpos- History, Department of Mam- es, the petrosal can be divided into two com- malogy ponents: the more anterior pars cochlearis, AMNH-DVP American Museum of Natural enclosing the cochlear duct and saccule, and History, Department of Verte- the more posterior pars canalicularis, enclos- brate Paleontology ing the utricle and semicircular canals. We FMNH Field Museum of Natural His- base our description of the petrosal of Pria- tory, Department of Geology codon on the more complete left one, sup- LACM Natural History Museum of Los plementing it with details from the right. The Angeles County petrosal is described in three views: ventral, dorsal, and lateral, with the orientation based MACN Museo Argentino de Ciencias on the inferred position of the petrosal in the Naturales "Bernardino Rivada- skull in relation to the exoccipital. via" In the left petrosal, most of both the pars MAE Collections of the joint Mongo- cochlearis and pars canalicularis are present. lian Academy of Sciences-Amer- Missing are the anterior tip of the promon- ican Museum of Natural History torium, the paroccipital process, and part of Paleontological Expeditions the anterior lamina and lateral flange. The PSS Paleontological and Stratigraph- right petrosal is represented only by the pro- ic Section of the Geological In- montorium and anterior half of the pars can- stitute, Mongolian Academy of alicularis. Sciences ZPAL Institute of Paleobiology, Polish VENTRAL VIEW (fig. IA, D) Academy of Sciences The most distinctive feature in this view is the promontorium or cochlear housing. It is DESCRIPTIONS fingerlike, suggesting that it contained a In addition to the cranial and postcranial straight cochlear duct. On the posterior as- elements of Priacodon fruitaensis described pect of the promontorium are two openings by Rasmussen and Callison (1981) and En- separated by a narrow bridge of bone, the gelmann and Callison (in prep.), right and crista interfenestralis. The more medial open- left petrosals and exoccipitals and an inde- ing is the perilymphatic foramen, which terminate laminar component are also pre- faces posteroventrally. Preserved on the ven- 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 5 tral and dorsal margins of the perilymphatic It is a high, horizontal ridge continuous with foramen are thin, sharp crests (fig. 2). The the ventral surface of the promontorium that gap between these crests at the posterior mar- separates the space behind the perilymphatic gin of the foramen is a preservation artefact, foramen from the main tympanic cavity lat- interrupting what would have been a contin- eral to the promontorium. Consequently, a uous rim. Anteriorly, both the preserved ven- post-promontorial tympanic sinus (Wible, tral and dorsal portions of the continuous 1990) is lacking. crest abruptly turn medially toward the jug- Along the medial margin of the promon- ular foramen, marking the posterior limit of torium anterodorsal to the perilymphatic fo- a deep sulcus, the perilymphatic groove. As ramen is an opening that leads into the pe- interpreted here, the continuous crest so de- trosal (inferior petrosal sinus in fig. ID). scribed marks the boundary between the ca- There is a second opening at the anterior tip vum tympani and the perilymphatic system, of the promontorium (inferior petrosal sinus and attached on the edge of the crest in life in fig. IE). In the less complete right petro- was the secondary tympanic membrane. This sal, the medial promontorial surface is bro- membrane extended over the perilymphatic ken, exposing a canal that in the left petrosal foramen, along the posterior aspect of the would be continuous between the two open- perilymphatic groove, and finally entered the ings. In connecting these two openings, the cranial cavity via the jugular foramen. On the canal would have to turn laterally at the an- right petrosal, the perilymphatic groove was terior end of the promontorium. We interpret nearly closed to form a canal by the crest for this canal as housing the inferior petrosal si- the attachment of the secondary tympanic nus (see Discussion). membrane; unfortunately, additional prepa- Medial to the perilymphatic foramen is the ration damaged the crest and the enclosure jugular notch, which, when completed by the now is not as complete as before. Medial to exoccipital, enclosed a nearly square jugular the crista interfenestralis and posterior to the foramen subequal in size to the perilymphat- perilymphatic foramen is a deep recess de- ic foramen. The perilymphatic and jugular limited anterolaterally by the crest for the at- foramina are housed in a common depres- tachment of the secondary tympanic mem- sion, the jugular fossa. Posterior to the jug- brane. This recess is reminiscent of a de- ular notch is a medially directed, barrel- pression in the platypus called the recessus shaped process on the petrosal that ends me- scalae tympani by Zeller (1989, 1991), but dially in an articular surface for the exoccip- differs in its relationship to the secondary ital. Posterolateral to the perilymphatic tympanic membrane. In extant mammals, the foramen is a distinct, cuplike depression recessus scalae tympani is on the labyrin- (pocket in fig. ID). It is bordered anterolat- thine side of the membrane (Goodrich, 1930; erally by the crista interfenestralis, postero- Kuhn, 1971; Zeller, 1989), whereas the de- laterally by the base of the paroccipital pro- pression in Priacodon is on the tympanic cess, posteromedially by a faint crest, and an- side. teromedially by a low, sharp process. We are The opening lateral to the crista interfe- uncertain of the function of this depression, nestralis is the fenestra vestibuli. It is larger though E. F Allin, a reviewer on an earlier than the perilymphatic foramen, faces ven- version of this manuscript, suggested it might trolaterally, and is elongate anteroposteriorly, house the levator hyoidei. This muscle is with an average stapedial ratio (Segall, 1970) found in extant monotremes (Edgeworth, of 1.45. Anterior to the fenestra vestibuli is 1914), but does not have a distinct depres- a tiny opening with a narrow sulcus running sion of origin on the petrosal. forward (foramen in promontorium in fig. The paroccipital process is broken in both IC, F); the opening corresponds to the fo- petrosals, but is more complete in the right ramen for small nerves and blood vessels one. It was a massive, subrectangular struc- identified in Trioracodon ferox by Kermack ture that projected ventrally. Although the (1963: fig. 3). The crista interfenestralis ex- full extent of the ventral projection cannot be tends posteriorly from the promontorium to established, the paroccipital process descend- the base of the broken paroccipital process. ed at least to the level of the ventral pro- 6 AMERICAN MUSEUM NOVITATES NO. 3183 A

B 'N

)~. /

.1/; C. K . &y>§-. V -Cr

1. i .-m.;

1. r ;1' .k .1.'t 4,! 9,.' k '>V ._ ':.i-.

11 - i

"_..V'A

lmm

Fig. 1. Three views of left petrosal of Priacodon fruitaensis, LACM 120451. A, D, Ventral view. B, E, Dorsal view. C, F, Lateral view. Parallel lines represent damaged surfaces. 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 7

D lateral f lange tensor tympani fossa foramen in anterior lamina foramen in canal promontor i um ,prootic o,ventral ascending groove facial foramen I fossa incudis

promontorium - broken tympanohyal ROSTRAL f enestra vesti buli crista parotica inferior petrosal sinus stapedius fossa MEDIAL perilymphatic foramen jugular notch jugular fossa lateral head vein suIcus exoccipital f acet crista into:erf enestra l is

anterior lamina semi lunar recess prootic canal _ prootic sinus ROSTRAL sulcus subarcuate fossa LATERAL anterior semicircular canal

F foramen in anterior lamina

dorsal ascending groove

squamosal facet

ventral ascending groove

.... posttemporal groove squamosal facet fossa incudis DORSAL broken paroccipital cristacristaparotica process 00ncTDAIKUZi KALL Fig. 1. Continued. 8 AMERICAN MUSEUM NOVITATES NO. 3183

promontorium 5mm

.;, --i .' ". Oei perilymphatic ".'e i foramen

perilymphatic exoccipital groove facet

crest

e\'

Fig. 2. Detail of area about the penlymphatic foramen of the right petrosal of Priacodonfruitaensis, LACM 120451. The irregular pattern indicates matrix, the dashed line the probable posterior margin of the perilymphatic foramen, and the parallel-line pattern the missing posterior portion of the crest for the attachment of the secondary tympanic membrane. montorial surface (fig. IC, F). A sharp crest, for the squamosal extends rostrally on the the crista parotica, whose anteroventral sur- lateral surface of the petrosal, almost to the face is damaged, runs anterodorsally from level of the anterior border of the fossa in- the base of the paroccipital process. Pre- cudis (fig. IC, F); in life, the squamosal may served on the ventromedial edge of the left have contributed to the lateral margin of the petrosal, near the midpoint of the crista par- fossa. otica is a small bump with a notch posterior Z. Luo, another reviewer of an earlier ver- to it; we interpret these features as the broken sion of this manuscript, expressed concerns base of the tympanohyal (fig. ID, F) and the that our reconstructed tympanohyal is posi- stylomastoid notch, respectively. A second tioned too far anteriorly and, given the lo- crest extends dorsally from the base of the cation of the fossa incudis, would have in- paroccipital process in an anterolateral direc- terfered with the incudostapedial articulation. tion. Between this crest and the crista paro- However, a similar positional relationship be- tica is a deeply concave, subtriangular sur- tween the base of the tympanohyal and the face, the fossa incudis, for the reception of fenestra vestibuli occurs in Trioracodon fer- the crus breve of the incus. Whereas the fos- ox (fig. 5E; Wible and Hopson, 1993) and sa incudis is delimited by well-developed also in some extant forms (e.g., Echinosorex, crests medially (the crista parotica) and pos- MacPhee et al., 1988: fig. 4D), yet the tym- teriorly (the paroccipital process), its antero- panohyal does not interfere with the incus lateral and lateral margins lack crests alto- and stapes because it is directed posterome- gether and its anteromedial margin has only dially. Additionally, we know of no expla- a weakly developed ridge. The contact facet nation for a notch in the crista parotica other 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 9 than for the facial nerve, and the tympano- bulk of the floor of the cavum epiptericum. hyal is invariably placed anterior to that A large depression in the anteromedial part notch in extant mammals (Goodrich, 1930). of the preserved lateral trough, delimited me- Anterior to the fossa incudis and fenestra dially by the steep lateral wall of the pro- vestibuli are two openings, the larger, more montorium, is the fossa muscularis major for lateral one being the ventral opening of the the tensor tympani muscle. It is not delimited prootic canal and the other the primary facial by a well-marked posterior crest as occurs, foramen. Running posteriorly from the proot- for example, in Trioracodon ferox (fig. 5E; ic canal, between the crista parotica and fe- Wible and Hopson, 1993). The lateral margin nestra vestibuli, is the wide sulcus for the of the lateral trough is formed by a ventrally lateral head vein (Wible and Hopson, 1995). projecting ridge, the lateral flange, which is It ends posteriorly in a shallow depression, broken anteriorly. The shallowest part of the the stapedius fossa, between the cristae in- preserved lateral flange is at its posterior lim- terfenestralis and parotica. In life, the lateral it where it forms part of a broad notch; the head vein presumably ran medially from the deepest part is at the preserved anterior limit. posterior end of its sulcus, ventral to the sta- The ventral surface of the lateral flange (to pedius fossa, to join the internal jugular vein the extent preserved) shows no facet for ar- ventral to the jugular foramen, as occurs in ticulation with the quadrate ramus of the al- didelphid marsupials (Rougier et al., 1992; isphenoid. However, the presence of a well- Wible and Hopson, 1995). The primary fa- developed quadrate ramus underlying the cial foramen opens into the deep recess that missing anterior part of the lateral flange can- in life housed the geniculate ganglion. The not be excluded. In the roof of the lateral bone surrounding the recess is unbroken, in- trough, between the primary facial foramen dicating that there was no bony floor ventral and the lateral flange, is a shallow longitu- to it; consequently, a bone-enclosed cavum dinal sulcus, which we interpret as contain- supracochleare for the geniculate ganglion is ing the ramus inferior of the stapedial artery, lacking. This arrangement is an unique one inasmuch as this vessel lies in the same po- in mammals (Wible, 1990; Wible and Hop- sition in the platypus (Rougier et al., 1992; son, 1993). Two sulci notch the anterior and Wible and Hopson, 1995). posterior edges of the primary facial fora- A horizontal sulcus on the lateral surface men, marking the course of the branches of of the petrosal (ventral ascending groove in the facial nerve. The anterior sulcus for the fig. ID, F) has an anterior limit that is con- palatine ramus is the deeper of the two, but tinuous with the broad notch in the rear of it can only be traced a short distance forward the lateral flange. Within the sulcus and pass- from the foramen. After a short course, the ing through the posterior part of the notch, posterior sulcus for the hyomandibular ramus we restore the ramus superior of the stapedial merges with the sulcus for the lateral head artery (see below). Given this reconstruction, vein lateral to the fenestra vestibuli. In life, the notch is homologous with the medial wall after sharing a short common course with the of the pterygoparoccipital foramen (the fo- lateral head vein, the hyomandibular ramus ramen for the ramus superior), which is al- presumably diverged laterally, running ven- most universally present in Cynodontia (Wi- tral to the vein, and left the middle-ear cavity ble and Hopson, 1993). through the stylomastoid notch behind the tympanohyal. DORSAL VIEW (fig. lB, E) Anterior to the facial foramen, contributing to the floor of the cavum epiptericum and Four major features are visible in this the roof of the anterior part of the middle- view: medially, the internal acoustic meatus ear cavity, is a broken laminar projection and subarcuate fossa; and laterally, the semi- from the promontorium, which forms the lat- lunar recess and the endocranial aperture and eral trough. In light of broken edges on the sulcus of the prootic canal. preserved lateral trough, the promontorium The internal acoustic meatus for the facial anterior to it, and the lateral flange (see be- and vestibulocochlear nerves is delimited lat- low), the petrosal probably constituted the erally and posteriorly by a raised surface. 10 AMERICAN MUSEUM NOVITATES NO. 3183

However, anteriorly and medially, it is con- the subarcuate fossa. It leads into the endo- tinuous with the dorsal surface of the pars cranial aperture of the prootic canal. Because cochlearis and, therefore, is not a distinct de- the endocranial aperture is essentially dorsal pression. Of the three foramina within the to the ventral aperture, the prootic canal is meatus, the largest and most medial is for the nearly vertical in orientation. It is also very cochlear nerve. Posterolateral to it is the short and anteriorly positioned, at the level opening for the vestibular nerve and antero- of the foramen for the vestibular nerve. lateral to it is the opening for the facial Anterior to the endocranial aperture of the nerve. The last is the smallest and lies in a prootic canal is the recessed area that housed common depression with the vestibular nerve the semilunar ganglion. The walls of the opening. The right petrosal is broken, which semilunar recess are formed medially by the exposes the course and orientation of the fa- crista petrosa (and prefacial commissure), cial canal. It is a long, narrow passageway laterally by the anterior lamina, and posteri- that extends to the primary facial foramen on orly by a distinct, narrow crest immediately the ventral surface. There is no enlarged in front of the endocranial aperture of the space along the canal that might have housed prootic canal. The floor of the semilunar re- the geniculate ganglion, and, as stated above, cess is incompletely preserved and so the full that ganglion occupied a depression ventral extent of the closure of the cavum epipteri- to the primary facial foramen. Given that the cum cannot be ascertained. The existence of facial canal is long, the prefacial commis- a pila antotica is also uncertain because sure-the segment of the crista petrosa dor- breakage in the anterior pole of the pars sal to the facial canal, in this specimen cochlearis. placed anterolateral to the internal acoustic meatus-is very deep. LATERAL VIEW (fig. IC, F) Separated from the internal acoustic meatus by a stout, low crest is the deeply exca- The lateral view is dominated by the an- vated subarcuate fossa, housing the parafloc- terior lamina in front and the articular surface culus of the cerebellum. A small groove and for the squamosal behind. The latter includes foramen for the endolymphatic duct (vestib- two well-developed sulci, one running hori- ular aqueduct) are located on the medial mar- zontally, the other vertically. Also visible in gin of the subarcuate fossa. The bone dorsal lateral view are the promontorium, the lateral to the subarcuate fossa is damaged, exposing flange, the crista parotica, the fossa incudis, the anterior semicircular canal. It is in this and the damaged base of the paroccipital pro- broken area that the sulcus for the sigmoid cess. sinus would be expected. The sigmoid sinus The horizontal sulcus mentioned above is ubiquitous in extant mammals (Gelderen, extends from the narrow mastoid exposure of 1924) and exits the braincase either from the the petrosal to the broad notch anterior to the foramen magnum or jugular foramen (Wible, fossa incudis. Running anterodorsally from 1990). This vessel was probably present in the dorsal margin of the horizontal sulcus is Priacodon as well. We believe that it exited a vertical sulcus, the dorsal ascending via the foramen magnum because (1) the jug- groove, which divides the horizontal sulcus ular foramen is too small to have transmitted into anterior and posterior parts. Given the a sigmoid sinus in addition to the glosso- location of the squamosal facet on the pos- pharyngeal, vagus, and accessory nerves and terior third of the lateral petrosal surface, the (2) no sulcus is found between the preserved preserved portion of the dorsal ascending medial margin of the subarcuate fossa and groove and the posterior part of the horizon- the jugular foramen. There is a faint, very tal sulcus (posttemporal groove in fig. IF) narrow sulcus extending dorsally from the were enclosed in canals walled laterally by posterior margin of the jugular foramen, but the squamosal. The posterior part of the hor- it does not reach the medial margin of the izontal sulcus was thus enclosed in life to subarcuate fossa. form the posttemporal canal, opening on the A deep sulcus for the prootic sinus runs occiput at the posttemporal foramen; the dor- parallel to the dorsal and lateral margins of sal ascending groove was enclosed as the 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS I1I dorsal ascending canal. The anterior part of is roughly triangular, with the posterior face the horizontal sulcus was open, forming the formed by the condyle, the medial face by ventral ascending groove. Following Rougier the basioccipital surface, and the lateral face et al. (1992) and Wible and Hopson (1995), by the petrosal surface. The medial face is we reconstruct arteries as the major occu- complete, exposing an extensive articular pants of these channels: the arteria diploetica facet for the basioccipital. In life, the exoc- magna in the posttemporal canal, the ramus cipital-basioccipital sutures diverged anteri- superior of the stapedial artery in the ventral orly. Between the medial border of the con- ascending groove, and the ramus supraorbi- dyle and the articular facet for the basioccip- talis of the stapedial artery in the dorsal as- ital is a notch lacking an articular surface; cending canal. The area around the posttem- this indicates that the left and right condyles poral foramen on the mastoid exposure were not continuous ventrally and were sep- shows no trace of a sulcus for the occipital arated by an odontoid notch. Most of the ar- artery, the source for the arteria diploetica ticular surface with the petrosal is lost. Pre- magna in, for example, monotremes and ar- served is a stout anterolateral process in front madillos (Wible, 1987). of the condyle that in life faced the barrel- Dorsal to the preserved part of the lateral shaped process of the petrosal described flange and the ventral ascending groove is a above (exoccipital facet in fig. ID, E). An- broken bony lamina forming the anterior teromedial to this process a single large hy- lamina of the lateral braincase wall. Two fo- poglossal foramen is preserved. ramina are preserved in the anterior lamina, Endocranially, in the part of the exoccip- both opening into the temporal fossa. A ital that floors the cranial cavity, are two fo- smaller posterodorsal one lies at the level of ramina of which the more posterior one is the posterior extent of the semilunar recess slightly the larger (fig. 3C). Both foramina (foramen in anterior lamina in fig. IF), cor- lead to canals that converge on the single ap- responding in position to foramen 3 de- erture present on the ventral surface and pre- scribed in Trioracodon ferox by Kermack sumably transmitted branches of the hypo- (1963: fig. 3). Given that this foramen lacks glossal nerve. an endocranial opening, it is likely a nutrient foramen. A large anteroventral foramen is PAUP ANALYSIS represented only by a notch; it is the foramen pseudovale of the same author and is pre- As stated above, a cladistic analysis of 51 sumed to be the exit for the mandibular di- petrosal and exoccipital characters across 20 vision of the trigeminal nerve. taxa was performed to evaluate the affinities of Priacodon fruitaensis and other triconodonts to one another and to other groups of EXOCCIPITAL (fig. 3) mammaliamorphs. Scores for several char- Both exoccipitals are preserved, but are in- acters included here differ from those in our complete. Illustrated here is the more com- prior published analyses (Wible and Hopson, plete left one. Its most conspicuous feature 1993; Rougier, 1993; Wible et al., 1995). is the occipital condyle (fig. 3A, B), which Character reevaluation on the basis of addi- is long, subcylindrical, and dorsoventrally tional specimens has uncovered some incon- compressed. The condyle is not sharply de- gruences and mistakes that are amended marcated from the surrounding bone and here. Character numbers follow the numera- both dorsal and ventral condylar fossae are tion in Appendices 1, 2, and 3, as well as in absent. The condyle is restricted to the ven- figures 4 and 5. tral part of the foramen magnum, contributing to less than half of the height of the fo- CHARACTER DIscusSION ramen (fig. 3B). Extending dorsally from the PETROSAL PROMONTORIUM (1) condyle and contributing to the margin of the foramen magnum is an incomplete occipital We consider a promontorium to be a pe- plate. It is laminar and tilted anteriorly. trosal eminence on the tympanic side of the In ventral view (fig. 3A), the exoccipital cochlear housing. Under such a definition, a 12 AMERICAN MUSEUM NOVITATES NO. 3183 A basioccipital facet rostral

I ty, Xll .r. .. / lateral

f odontoid ,, notch-

occipital plate condyle B dorsal

occipital plate < 9., medial 1. foramenf 1mm I /f _0!1 magnum odontoid 'K notch C condyle Nb

Xii basioccipital basioccipital X facet facet

/;i5\odontoid notch

rostral

- occipital plate condyle medial

Fig. 3. Three views of left exoccipital of Priacodon fruitaensis, LACM 120451. A, Ventral view. B, Posterior view. C, Dorsal view. Parallel lines represent damaged surfaces. 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 13 promontorium is present in all the ingroup reviewer of an earlier version of this manu- taxa included here with the exception of script, suggested that this crest may represent tachyglossids. In these forms (fig. 5B), a the lateral limit of the longus capitis muscle prominent bulge on the cochlear housing is insertion. lacking, although a raised area is present on the posteroventral aspect of the petrosal re- SECONDARY SPIRAL LAMINA (7) lated to the abutment of the hyoid arch Recent papers by Meng and Fox (1993, (Kuhn, 1971). Consequently, we code tachy- 1995a, 1995b) have focused attention on the glossids as lacking a promontorium. phylogenetic utility of inner ear structures, including the secondary spiral lamina. A PARA/BASISPHENOID WING (2) character concerning the presence/absence of Lucas and Luo (1993) reported that a wing the secondary spiral lamina was used in the of the basisphenoid covers the anterior pole phylogenetic analyses by Rougier (1993) and of the promontorium in Adelobasileus. They Wible et al. (1995). In the taxon-character interpreted this condition as intermediate be- matrix of the latter paper, the condition in tween that in non-mammalian cynodonts in Sinoconodon was scored as unknown. How- which the wing reaches the fenestra vestibuli ever, cross sections through the cochlear ca- (fig. 4B) and in mammaliaforms in which the nal of Sinoconodon published by Luo et al. wing is absent. As the parasphenoid is (1995) reveal that a secondary spiral lamina thought to be fused to the basisphenoid in is absent. The presence of a secondary spiral non-mammalian cynodonts and early mam- lamina in Vincelestes is documented by a sul- maliaforms (Miao, 1988), we here call this cus on the endocast of the cochlea (Rougier, the para/basisphenoid wing. 1993: fig. 34). COCHLEAR HOUSING SHAPE (3) PERILYMPHATIC DUCT (11, 12) Fitting the cochlear housing of Ornitho- In the primitive condition in amniotes, the rhynchus (fig. SA) into a shape category perilymphatic duct leaves the cranial cavity posed problems, and we ultimately coded it via the jugular foramen and reaches the inner as fingerlike in outline; this is the shape of ear through the perilymphatic foramen in the the pars cochlearis clearly evident from the rear of the otic capsule (Gauthier et al., endocranial aspect, but it is masked in ven- 1988). This condition is repeated in Orni- tral view by additional bone growth along thorhynchus (Zeller, 1991, 1993), except that the medial border. Additionally, we coded the perilymphatic foramen also provides par- the cochlear housing shape as indistinctive tial attachment for the secondary tympanic for tachyglossids as well as the outgroup membrane (Presley, 1980). This membrane, taxa. which is said to be absent primitively in amniotes (Gauthier et al., 1988), lies at the lat- VENTRAL CREST ON COCHLEAR HOUSING (4) eral limit of an extracranial space behind the perilymphatic foramen, the recessus scalae Among the taxa considered here, a distinct tympani, where the perilymphatic duct con- longitudinal crest on the ventral surface of tacts the cavum tympani (Zeller, 1991, 1993). the cochlear housing (fig. 4G, H) is found in In therians, in contrast to the condition in Sinoconodon (Crompton and Luo, 1993), Ornithorhynchus, a chondrocranial process, Megazostrodon (Gow, 1986b), and Haldan- the processus recessus, encloses the perilym- odon (Lillegraven and Krusat, 1991). This phatic duct within a canal, the cochlear aq- crest is reported to be absent in Morganu- ueduct; this process ossifies with the petrosal codon (Crompton and Luo, 1993), though we in the adult and delimits another aperture in note that a raised area is present in some iso- the rear of the otic capsule, a true round win- lated petrosals of M. watsoni (Kermack et al., dow or fenestra cochleae (fenestra rotunda), 1981: figs. 73C, 80D, F). Until we restudy which is closed by the secondary tympanic M. watsoni, we code Morganucodon follow- membrane (Goodrich, 1930; Zeller, 1985). ing Crompton and Luo (1993). E. F Allin, a Tachyglossus presents a problem in deter- 14 AMERICAN MUSEUM NOVITATES NO. 3183

A B 35(0) 29(0),

1 0(0) -2(0) 11(O) 14(0) 15(0) C D 0 l rn\ 34(1) .-,, o69tu) 37(0)

_ 30(1) 3(1) - 27(1 )_ - 1(1) 31(1)<

16(1) 11(1) . 5 1 (1 ) E F 36(1) 36(1) 33(1) 27(1)

27(1) 24(1), 24(1) - 5 1 (0) 1 6(1) G H

- 4(1) 4(1) 1(1) 1 6(l)

1, -.I= II

Fig. 4. Right petrosals or ear regions in ventral view; not to scale and some have been reversed from the original. Numbers correspond to characters (and states) in appendices 1-3. A, The tritheledontid Pachygenelus; B, generalized tritylodontid; C, Adelobasileus; D, Sinoconodon; E, Morganucodon oehleri; F, Dinnetherium; G, Megazostrodon; H, Haldanodon. The figures are reprinted and modified with permission or redrawn from the following sources: A, Wible and Hopson (1993); B, E, Luo (1994); C, Luo et al. (1995); D, F, Crompton and Luo (1993); G, Gow (1986b); H, Lillegraven and Krusat (1991). 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 15 mining the homologies of the canal for the exhibiting osseous enclosure of the perilym- perilymphatic duct. As in Ornithorhynchus, phatic duct under the same character state. the perilymphatic duct in Tachyglossus en- In considering the complex transforma- ters the inner ear via the perilymphatic fo- tions related to the enclosure of the perilym- ramen. However, whereas the duct runs to phatic duct and associated structures, we rec- the foramen in an open groove in Ornithor- ognize two multistate characters. These are: hynchus; it is enclosed in a canal in Tachy- (11) degree of enclosure of the perilymphatic glossus (Kuhn, 1971). Our study of macer- duct; and (12) position of the lateral aperture ated skulls of Tachyglossus and Zaglossus of the cochlear aqueduct. reveals that the degree of enclosure of the Among the basal taxa considered here, the perilymphatic duct varies both ontogeneti- position of the perilymphatic duct is not cally and individually. For example, in ju- marked on the bridge of bone separating the venile Tachyglossus the duct lies in a sulcus perilymphatic and jugular foramina (fig. 4A). and is open ventrally (Kuhn, 1971: fig. 40; This condition, present in tritheledontids, tri- Wible and Hopson, 1995: fig. 3). On the oth- tylodontids, Adelobasileus, and Sinocono- er hand, adult specimens may show extensive don, can be considered primitive for mam- closure on one side of the skull but not on maliamorphs. In most of the remaining taxa, the other (e.g., AMNH-DM 42176). In at the position of the duct is marked by an os- least some Zaglossus (AMNH-DM 190863), seous channel that shows differing degrees of the ossification enclosing the perilymphatic enclosure; the exceptions are Megazostro- duct, the "processus recessus" of Kuhn don, Dinnetherium, Triconodon, and the un- (1971), is even more developed than in named Cloverly triconodonts in which the Tachyglossus and circumscribes a larger por- appropriate part of the petrosal is not pre- tion of the duct within a canal. The homol- served, described, or figured. An open sulcus ogy of the "processus recessus" in Tachy- for the perilymphatic duct (fig. 4D, E) is glossus (and by extension also in Zaglossus) found in the Jurassic mammaliaforms Mor- with the processus recessus in therians has ganucodon and Haldanodon, and in ptilo- been questioned by Kuhn (1971), because dontoid multituberculates (e.g., compare Me- that of echidnas encloses the glossopharyn- sodma thompsoni FMNH PM 53904; Ptilo- geal nerve in addition to the perilymphatic dus montanus AMNH-DVP 35490). The sul- duct. We agree that the relationships of the cus is deeper and a ventral lappet of bone tachyglossid "processus recessus" are extends posteriorly and dorsally from the unique among living mammals, which may promontorium, nearly flooring a canal, in suggest the nonhomology of the tachyglossid isolated petrosals of taeniolabidoid multitu- and therian cochlear aqueducts, despite their berculates (e.g., cf. "Catopsalis" joyneri striking osteological similarities. In addition AMNH-DVP 118533, 119445) from the Late to the canal for the perilymphatic duct pres- Cretaceous of North America. The Late Ju- ent in tachyglossids and therians, enclosure rassic paulchoffatiid multituberculates also of the duct is also thought to have occurred have a deep sulcus (Hahn, 1988; Lillegraven independently on at least three other occa- and Hahn, 1993), but the presence of a ven- sions in amniotes: in squamates, in crocodil- tral bony lappet from the promontorium can- ians, and in birds (Gauthier et al., 1988). not be ascertained in the fragmentary mate- Only the presence or absence of the canal rial available. The bony lappets on the edges for the perilymphatic duct can be ascertained of the sulcus are expanded further in Pria- for extinct taxa in our cladistic analysis. With codon and produce a nearly complete canal no soft tissues preserved, we cannot know (fig. 2). A sulcus is also present in Triora- whether the perilymphatic duct was the only codon (Kermack, 1963: fig. 3), but the pres- occupant of the cochlear aqueduct as in ther- ence of bony lappets is uncertain from the ians or if it was accompanied by the glos- published figures. sopharyngeal nerve as in tachyglossids. Con- The Khoobur petrosals and prototribos- sequently, despite the risk of treating non- phenidans exhibit a cochlear aqueduct (fig. homologous canals as homologous in a trans- 5I), but a morphocline exists regarding the formation series, we subsume all taxa degree of enclosure. The Early Cretaceous I--CY) 01) 0 m C) _-E I l 0 O0 C\J (\J

0 C?,) 03) (-) I- ^-r m 00

1- 0 C?) IO (-NJ

1- N- CY) 0 CY) (N IIJ I

1-1 (-JC\J CY) CY)

C\3 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 17

Khoobur petrosal 2 exhibits a more enclosed CAUDAL TYMPANIC PROCESS OF THE state than that in tachyglossids, with a pro- PETROSAL (18) cessus recessus that is thicker and that shields the perilymphatic duct all the way The caudal tympanic process is a crest on from the endocranial surface of the petrosal the petrosal situated medial to, and usually to immediately posteromedial to the perilym- continuous with, the paroccipital process, phatic foramen, just outside the inner ear. forming the posterior wall of the tympanic The processus recessus is even thicker in cavity (fig. SF, H; Wible, 1990). In the cla- Khoobur petrosal 1, enclosing the perilym- distic analyses by Wible and Hopson (1993) phatic duct more deeply within the petrosal. and Wible et al. (1995), Multituberculata is It shields the perilymphatic duct all the way scored as lacking a caudal tympanic process, from the endocranial surface to inside the in- the condition present in (for example) the ner ear and, thereby, contributes to the for- ptilodontoid Ptilodus (AMNH-DVP 35490) mation of a true round window (contra Wible and the taeniolabidoid Nemegtbaatar (ZPAL et al., 1995; see below). The lateral aperture MgM-I/76). However, the recently recovered of the cochlear aqueduct is immediately ros- skull of the taeniolabidoid Kryptobaatar tral to the anteromedial rim of the fenestra dashzevegi (PSS-MAE- 113) from the Late cochleae, within the inner ear. Finally, Vin- Cretaceous of Mongolia shows a distinct celestes, a prototribosphenidan from the Ear- caudal tympanic process, as does Kampto- ly Cretaceous of Argentina, shows an even baatar (ZPAL MgM-1/33). Reexamination of more enclosed perilymphatic duct with a other multituberculate ear regions, including massive processus recessus and a medial ap- the Late Jurassic paulchoffatiids (Hahn, erture for the cochlear aqueduct recessed 1988; Lillegraven and Hahn, 1993), reveals from the round window, well within the in- that the area medial to the paroccipital pro- ner ear. This is also the condition present cess is often subject to considerable damage. primitively in marsupials and placentals. We score Multituberculata as polymorphic In their original description of Khoobur because both states are clearly present. petrosal 1, Wible et al. (1995) observed the absence of a sulcus for the perilymphatic DEEP POCKET MEDIAL TO PAROCCIPITAL duct and considered the opening in the back PROCESS (20) of the promontorium as a perilymphatic fo- The distinct cuplike depression of uncer- ramen because no likely cochlear aqueduct tain function medial to the paroccipital pro- was to be found. Restudy of the material af- cess and crista interfenestralis in Priacodon ter further preparation shows a cochlear aq- (fig. ID) is also found in two other taxa con- ueduct to be present. Its medial aperture lies sidered here: Trioracodon ferox (fig. SE) and in a narrow seam in the jugular notch and the Khoobur petrosal 1 (fig. SF). In Triora- the lateral aperture is a small opening within codon, this depression has been identified as the cochlear cavity just in front of the fenes- a possible pit for the hyoid by Kermack tra cochleae. (1963) and as the fossa for the stapedius muscle by Wible and Hopson (1993). Neither JUGULAR FoSSA (13) interpretation is likely for Trioracodon, because the proximal segment of the hyoid Primitively in non-mammalian cynodonts, arch, the tympanohyal, is already fused to the the jugular and perilymphatic foramina are crista parotica and a stapedius fossa is clearly confluent (Quiroga, 1979). In the taxa in- present lateral to the crista interfenestralis. In cluded in our analysis, these two apertures both regards, Trioracodon closely resembles are separated (Wible, 1991). Additionally, in the condition in Priacodon. In their descrip- tritheledontids, tritylodontids, and most Ju- tion of Khoobur petrosal 1, Wible et al. rassic mammaliaforms, including multituber- (1995) did not recognize this pocket, which culates, these two apertures open into a com- is nevertheless visible in their figure 2 (be- mon recessed area referred to here, following tween the lines indicating the caudal tym- Kielan-Jaworowska et al. (1986), as the jug- panic process of the petrosal and the post- ular fossa (fig. 4F). promontorial tympanic sinus). 18 AMERICAN MUSEUM NOVITATES NO. 3183

The Liassic Morganucodon has a depres- least one ear region (Hahn, 1988: fig. 1), the sion on the crista interfenestralis anterior to squamosal is ventrally prolonged and as re- the paroccipital process that was identified as constructed by us would have contacted the the pit for the levator hyoidei muscle by Ker- crista parotica. mack et al. (1981). However, we follow Crompton and Sun (1985) and others in iden- VASCULAR FORAMEN IN LATERAL FLANGE tifying this feature as for the stapedius mus- (31) cle. In addition to Morganucodon, a similar depression on the crista interfenestralis oc- First used in cladistic analysis by Wible curs in tritylodontids, Sinoconodon, Dinneth- (1991), the absence or presence of a vascular erium, Megazostrodon, and Haldanodon. foramen through the lateral flange (fig. 4C) This condition of the stapedius fossa is dis- has been included in subsequent analyses by tinguished from the deep pocket occuring in Wible and Hopson (1993), Lucas and Luo Priacodon, Trioracodon, and Khoobur petro- (1993), Rougier (1993), Luo (1994), and Wi- sal 1, because the latter occurs medial to the ble et al. (1995). We offer the following crista interfenestralis along with a more lat- amendments to the matrix in Wible et al. erally-placed stapedius fossa. (1995). Haldanodon is changed from foramen present to unknown, because the appropriate part of the petrosal is damaged. Vin- CRISTA PAROTICA/SQUAMOSAL CONTACT (23) celestes is altered from absence of the fora- Presence or absence of contact between men to polymorphic, because as reported by the crista parotica and squamosal was re- Rougier et al. (1992) this aperture is variably cently used as a character in the cladistic present. Finally, the foramen in Multituber- analysis by Wible et al. (1995). The follow- culata, said to be absent by Wible et al. ing scores in their matrix we amended here. (1995), is scored here as present. Following Megazostrodon is altered from contact pres- Rougier (1993), we interpret the supragle- ent to unknown, because this feature is not noid foramen of multituberculates and the described or illustrated in the only descrip- lateral flange vascular foramen of other taxa tive treatment of the skull of this taxon as equivalent, because they occupy essen- (Gow, 1986b). Haldanodon is changed from tially the same location and connect with unknown to contact absent, as illustrated in similar vascular canals. This contrasts with Lillegraven and Krusat (1991: fig. 8). Mar- the discussion in Wible and Hopson (1995), supialia and Placentalia, both of which were who did not consider the multituberculate su- scored as contact absent, are treated as poly- praglenoid foramen to be equivalent to the morphic here, because both conditions are lateral flange vascular canal of other taxa. evident in isolated petrosals from the North American Late Cretaceous (Maclntyre, 1972; FACIAL GANGLION FLOOR (33) Wible, 1990). Finally, Multituberculata, Wible et al. (1995) scored the floor of the scored by Wible et al. (1995) as contact ab- facial ganglion as open ventrally in Tritylo- sent, is amended to contact present. Few dontidae. An exception to this is Oligoky- multituberculates preserve this region, and phus, which has a bony floor (Kuhne, 1956). the only form that clearly shows absence of Given the putative basal position of Oligo- contact between the crista parotica and squa- kyphus within Tritylodontidae (Clark and mosal is the Late Paleocene taeniolabidoid Hopson, 1985), the primitive condition for Lambdopsalis, a specialized fossorial form this group is equivocal. Accordingly, we from China (Miao, 1988). Contact is clearly code Tritylodontidae has having two states present in more generalized taeniolabidoids for this character: open ventrally and a broad from the Mongolian Late Cretaceous, e.g., petrosal bridge. Chulsanbaatar, Nemegtbaatar (Kielan-Ja- worowska et al., 1986), and Kryptobaatar PILA ANTOTICA (39) dashzevegi (PSS-MAE-1 13). In addition, although none of the Late Jurassic paulchof- Wible et al. (1995) scored the presence of fatiids can be scored for this character, in at the ossified base of the pila antotica (fig. 4C) 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 19 as uncertain for Megazostrodon. However, the former bends laterally and opens at the this is altered to presence here, following Lu- apex of the promontorium. Canals resem- cas and Luo (1993). bling that in M. watsoni are found in Hal- danodon (Lillegraven and Krusat, 1991), POSITION OF INFERIOR PETROSAL SINUS (42) paulchoffatiid multituberculates (Lillegraven and Hahn, 1993), the Khoobur petrosal 1 In Recent mammals, the inferior petrosal (fig. SF; Wible et al., 1995), Priacodon, sinus drains blood from the cavernous sinus Trioracodon, and Vincelestes (MACN-N05, to the internal jugular vein via a course along N09). The coding of the inferior petrosal si- the petrosal-basioccipital suture. Its course nus by some of us in a prior analysis (Wible may be largely endocranial, intramural, or et al., 1995) is altered here to reflect our new extracranial (Wible, 1983). A comparable interpretation. We note, however, that Ker- vein does not occur in Recent non-mamma- mack et al. (1981) offered a different resto- lian amniotes (Shindo, 1915), and therefore ration of the inferior petrosal sinus in Mor- the sinus is a neomorph originating within ganucodon, as an endocranial vessel verti- synapsids. However, the only synapsids with cally piercing the petrosal-basioccipital su- any osseous indication for the inferior petro- ture. An appropriate sutural gap is not sal sinus are certain mammaliaforms. illustrated in any of their specimens. Kermack et al. (1981) illustrated a number of petrosals of Morganucodon watsoni in ANTERIOR LAMINA (45) which foramina open into internal canals within the promontorium. Within these fo- In Wible et al. (1995), the contribution of ramina and canals, they reconstructed a cir- the anterior lamina to the cranial wall was cumpromontorial venous plexus. The largest coded in reference to the foramina for the and most consistent of these foramina are at maxillary and mandibular branches of the tri- the anterior and posterior ends of a major geminal nerve. In the case of two nerve ap- internal canal along the medial edge of the ertures in or near the anterior lamina, two promontorium (best seen in their fig. 83B). models of nerve restoration have been pro- The anterior opening lies at the anterior pole posed. First, Simpson (1937) identified two of the promontorium within the cranial cav- foramina in the ventrolateral braincase of the ity, and the posterior opening is extracrani- multituberculate Ptilodus montanus and re- ally placed, rostral to the jugular notch. No constructed branches of the mandibular comparable structure is known in any group nerve within them. This model was based on of living mammals, although in some (e.g., Hill's (1935) observation of two foramina for monotremes) the petrosal houses small ve- the mandibular nerve within the alisphenoid nous sinuses (Rougier et al., 1992). These in some living rodents. Second, Patterson small sinuses are highly anastomotic and do and Olson (1961) identified two major ap- not form a single continuously walled canal ertures in the anterior lamina of Sinoconodon that could account for the structure present as transmitting the maxillary and mandibular in M. watsoni. Given that the only major ve- nerves, respectively. The same restoration nous element in this area of the skull in Re- was proposed for Morganucodon (Kermack, cent mammals is the inferior petrosal sinus, 1963, 1967), and the placement of the max- we interpret this vein as the likely occupant illary nerve in the anteriormost opening was of the canal in M. watsoni. A similar arrange- justified by a shallow, anteriorly directed ment is found, for example, in some living groove extending from the foramen, which carnivorans in which the inferior petrosal si- represented an appropriate course for the nus occupies a deep sulcus in the medial maxillary nerve (Kermack et al., 1981). edge of the promontorium that is transformed However, this restoration was challenged by into the petrobasilar canal by contact with the Kielan-Jaworowska (1971), who considered basioccipital and auditory bulla (Hunt, 1977; both foramina in the anterior lamina of "tri- Evans and Christensen, 1979). A striking dif- conodonts" (i.e., Trioracodon and Morgan- ference between the canal in the petrosal in ucodon) to be for mandibular nerve branch- M. watsoni and the petrobasilar canal is that es. Kermack et al. (1981: 97) acknowledged 20 AMERICAN MUSEUM NOVITATES NO. 3183 that "this may well be a possible alternative ly. An additional character (48), also used by interpretation of the condition in Morganu- Wible et al. (1995), treats the number of open- codon. " ings for the trigeminal nerve posterior to the Although the presence of an anteriorly di- alisphenoid as either none, single, or multiple. rected sulcus seems to be a reasonable cri- terion for identifying a foramen for the max- AFFINITIES OF Priacodon fruitaensis illary nerve, the fact is that such grooves may accommodate branches of the mandib- Running the taxon-character matrix in ap- ular nerve as well, such as deep temporal pendix 2 on the branch and bound algorithm nerves (see Evans and Christensen, 1979). of PAUP produced six equally most parsi- Further doubts concerning the identification monious trees (147 steps.) They differ in the of foramina in the ventrolateral wall of the relationships among the triconodontid taxa braincase are raised by the position of the and in the positions of triconodonts and mul- two foramina in paulchoffatiid multituber- tituberculates with respect to Prototribos- culates. In Cretaceous and Tertiary multitu- phenida + the Khoobur petrosals. The strict berculates, such as Ptilodus, the foramina are consensus of these six trees is shown in fig- directed ventrally and anterolaterally, as is ure 6A. Reweighting the data set, using the the case in rodents. In contrast, in the Early rescaled consistency index on PAUP's re- Jurassic paulchoffatiids, both foramina seem weighting option, yielded three equally most to be directed laterally (see Hahn, 1988)- parsimonious trees, the consensus of which reminiscent of the condition in Morganuco- is shown in figure 6B and diagnosed in ap- don. If nerves are restored in paulchoffatiids pendix 3. These three trees differ only in the as in other multituberculates (Simpson, 1937; relationships among the triconodontids. It is Kielan-Jaworowska, 1971), then both of the noteworthy that the most incomplete taxon paulchoffatiid foramina are for the mandib- included here is Triconodon mordax, which ular nerve. However, if the Morganucodon is scored for only 17 of the 51 characters. model (Kermack et al., 1981) is followed, This low degree of completeness (33.3%), then the two foramina in paulchoffatiids are which undoubtedly accounts for the unresol- for the maxillary and mandibular nerves, re- ved relationships within Triconodontidae, is spectively. Recently discovered crania of well below the 86.7% average completeness Sinoconodon and Adelobasileus pose addi- of the taxa in the matrix. tional problems. Both forms have three large Two other phylogenetic analyses of basi- foramina occupying similar positions in the cranial characters across mammaliamorphs anterior lamina. Whereas Crompton and Luo have been completed recently by some of us (1993) allocated two foramina to the man- (Wible and Hopson, 1993; Wible et al., dibular nerve and one to the maxillary nerve 1995). The current analysis considers more in Sinoconodon, Lucas and Luo (1993) al- ingroup taxa and characters. The additional located one foramen for each of these nerves ingroup taxa result from including a second in Adelobasileus, with the third said to be isolated petrosal from Khoobur and from indeterminate in function. splitting Monotremata into OTU's for the Resolution of these problems awaits careful platypus and echidnas and Triconodontidae comparison of the relevant extinct and extant into OTU's for Priacodon, Triconodon, taxa as well as the identification of criteria Trioracodon, and the unnamed Cloverly tri- allowing an unambiguous recognition of max- conodonts. The additional characters include illary and mandibular nerve exits. In the features unique to triconodonts as well as meantime, reflecting the doubts raised above features not considered previously, such as (and contra Wible et al., 1995), we do not that of the exoccipital (see appendix 1). The characterize the anterior lamina's contribution phylogenetic results of the current report to the lateral braincase wall with reference to most closely resemble those of Wible et al. the exits for the maxillary and mandibular (1995), with the main difference being that nerves. Our reworded character description all taxa, except triconodonts and multituber- considers whether the anterior lamina does or culates, are fully resolved in the strict con- does not contact the orbitosphenoid external- sensus tree of the original analysis here (fig. 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 21

\ o~~~6 o, 6\. e

B #,# p~\

BL C

Fig. 6. A, Strict consensus of the six most parsimonious trees resulting from the PAUP analysis of the matrix in appendix 2. Tree length = 148 steps; consistency index = 0.568; retention index = 0.708; these values on the individual trees are 147, 0.571, and 0.712 respectively. B, Strict consensus of the three most parsimonious trees resulting from the reweighted (rescaled consistency index) PAUP analysis. This tree is diagnosed in appendix 3, and the numbers here correspond to the diagnosed nodes. 22 AMERICAN MUSEUM NOVITATES NO. 3183

6A); the strict consensus tree in Wible et al. tidae is dependent upon their relationship (1995) has three unresolved trichotomies be- with Alticonodon, Astroconodon, and the re- tween Mammaliaformes and Prototribos- maining triconodont families (Amphilesti- phenida. However, using MacClade (Maddi- dae, Austrotriconodontidae, and Gobicono- son and Maddison, 1992) to reproduce one dontidae) for which no petrosals, have yet of the most parsimonious trees recovered in been identified. Wible et al. (1995) requires that four steps Triconodontidae here is supported by three be added to the most parsimonious trees unequivocal synapomorphies and one equiv- here. The most striking difference between ocal synapomorphy (appendix 3: node 5). the current report and Wible et al. (1995) on Among the unequivocal characters are the the one hand and Wible and Hopson (1993) presence of a small foramen on the cochlear on the other is the position of triconodonts housing in front of the fenestra vestibuli and with regard to Mammalia; here and in Wible a very small jugular foramen (fig. SE). et al. (1995) triconodonts are included within Mammalia, whereas they are excluded in Wi- TRICONODONT RELATIONSHIPS ble and Hopson (1993). Yet, the position of triconodonts is very labile in all three anal- Our analysis is inconclusive regarding the yses, and in some trees that are one step lon- phylogenetic position of triconodonts among ger than the most parsimonious ones pre- Mammaliaformes. In three of the six most sented here, triconodonts are excluded from parsimonious trees obtained, triconodonts are Mammalia. Within the framework provided more closely allied with therians than are by the current cladistic analysis, we address multituberculates; the reverse is true in the below triconodont interrelationships and oth- remaining three trees. Reweighting of our er issues of mammaliamorph phylogeny. data set identified the same three trees that allied triconodonts with therians to the ex- TRICONODONT INTERRELATIONSHIPS clusion of multituberculates (fig. 6B). In this instance, the clade of triconodonts + the Four taxa (the indeterminate Cloverly tri- Khoobur petrosals + Prototribosphenida conodontids, Priacodon, Trioracodon, and (Theriimorpha, see below) is supported Triconodon) are considered to be tricono- weakly by a single unequivocal and four donts here, based on one unequivocal and equivocal synapomorphies (appendix 3: node four equivocal synapomorphies (appendix 3: 7). The unequivocal synapomorphy, the cran- node 6). The unequivocal feature is a wide iomandibular joint anterior to the fenestra separation between the lateral flange and the vestibuli, is unknown in Priacodon, Triora- crista parotica into which the ventral ascend- codon, Triconodon, and Khoobur petrosal 2. ing groove opens (fig. SE). In non-tricono- The term Theriimorpha was coined by dont mammals, the lateral flange is in closer Rowe (1993) as a stem-based taxon includ- proximity to the crista parotica (or is absent), ing all mammals more closely related to and the ventral ascending groove opens into Theria than are monotremes. In the strict a notch between the two or is enclosed in a consensus trees obtained in our initial anal- foramen by them. ysis (fig. 6A) as well as in the reweighted As defined above, the family Triconodon- analysis (fig. 6B), Theriimorpha (Rowe, tidae is the group including the last common 1993) is synonymous with Theriiformes ancestor plus descendants of the five genera (Rowe, 1988). In order to avoid redundant included in Triconodontidae by Jenkins and and novel terms, following Wible et al. Crompton (1979): Alticonodon, Astrocono- (1995), we use Theriimorpha here to include don, Priacodon, Triconodon, and Trioraco- the last common ancestor of Triconodontidae don. In our analysis, the Cloverly tricono- and Theria plus all its descendants. donts are identified as the sister taxon of an Although, as shown above, the phyloge- unresolved monophyletic group including netic relationships of triconodonts are uncer- Priacodon, Triconodon, and Trioracodon. tain, our results are congruent with those of Whether or not the Cloverly triconodonts recent cladistic analyses (e.g., Rowe, 1988; should be included ultimately in Triconodon- Wible, 1991; Wible et al., 1995) in denying 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 23 the monophyly of Triconodonta in the tradi- obtained in our study is congruent with ba- tional sense. This deeply rooted concept, sicranial evidence presented previously by which groups several Jurassic taxa (e.g., some of us (Wible et al., 1995) as well as, Morganucodon, Sinoconodon, and Dinneth- for example, the evidence of the shoulder erium) with triconodonts as understood here, girdle. As in Vincelestes (Rougier, 1993) and is still used in formal systematics and rep- therians, the scapula of gobiconodontids resents the archetypal Mesozoic "mammal" (Jenkins and Schaff, 1988) and the scapula for many investigator (e.g., Carroll, 1988; attributed to the Cloverly triconodonts (Jen- Bonaparte, 1992; Kielan-Jaworowska, 1992; kins and Crompton, 1979) has a spine sepa- Stucky and McKenna, 1993). Given that all rating supra- and infraspinous fossae, both of recent cladistic analy'ses that include the per- which extend to the glenoid. In contrast, in tinent taxa (e.g., Rowe, 1988, 1993; Wible, more basal taxa, a supraspinous fossa is 1991; Wible and Hopson, 1993; Wible et al., small and dorsally placed (tritylodontids; 1995) have denied the monophyly of Tricon- Sues, 1983; Sun and Li, 1985), rudimentary odonta in the traditional sense, we recom- (multituberculates; Sereno and McKenna, mend that this use of the term be abandoned. 1995), or lacking altogether (monotremes). Whether or not the four families of tricono- However, as is evident by the strict consen- donts recognized by us (Triconodontidae, sus tree (fig. 6A), the position of triconodonts Amphilestidae, Austrotriconodontidae, and with respect to multituberculates is open to Gobiconodontidae) form a natural group has question. In general, the tree topology recov- not been established. Studies identifying den- ered here is labile with respect to a few tal synapomorphies have yet to be published, changes in the data set, and 123 additional and because amphilestids, austrotriconodon- trees only one step longer than the most par- tids, and gobiconodontids are not yet known simonious ones (148 steps) were identified from petrosals, they cannot be incorporated using the branch and bound algorithm on the into our analysis. Furthermore, because the matrix. The strict consensus of the 129 trees four triconodont families may constitute a (147 + 148 steps) erases much of the reso- paraphyletic assemblage, we shall not for- lution recovered in the six most parsimoni- mally define Triconodonta as the last com- ous trees. It is evident from the low number mon ancestor of Triconodontidae, Amphiles- of unequivocal characters defining some tidae, Austrotriconodontidae, and Gobicono- clades (appendix 3) that the least stable re- dontidae plus all its descendants. lationships are those of the pre-Haldanodon The only recent researcher to postulate Jurassic taxa (see below) and those of trico- characters in support of Triconodonta in the nodonts, multituberculates, and monotremes. traditional sense is Luo (1994). He offered In the 123 trees one step longer (148 steps), one dental and five cranial features allying triconodonts are either grouped paraphyleti- morganucodontids and triconodontids. How- cally with the Khoobur petrosals or are the ever, as he pointed out, Mammalia as we un- sister taxa to Prototribosphenida + the derstand it (i.e., the last common ancestor of Khoobur petrosals (as in fig. 6B), to Theri- monotremes and therians-plus its descen- iformes (Multituberculata plus Prototribo- dants) was excluded from his study. In fact, sphenida + the Khoobur petrosals), or to Luo's five purported cranial synapomorphies Mammalia. are found in Mammalia, in multituberculates and/or living mammals and their extinct rel- PROTOTRIBOSPHENIDA atives (e.g., Vincelestes). Additionally, as Luo noted, the single dental character (lower cusp Prototribosphenida (Rougier, 1993), the a occluding between upper cusps A and B) is grouping of Vincelestes and therians, is one not present in amphilestids. The monophyly of the most robustly supported clades in our of Triconodonta in the traditional sense has analysis, with four unequivocal and five not yet been rigorously tested and is not fea- equivocal characters (appendix 3: node 2). sible without consideration of all relevant All cladistic analyses of cranial morphology mammaliamorph taxa. considering Vincelestes (e.g., Wible, 1991; The phylogenetic position for triconodonts Rowe, 1993; Rougier, 1993; Wible and Hop- 24 AMERICAN MUSEUM NOVITATES NO. 3183

son, 1993) have arrived at the same phylo- ians than are triconodonts. Supporting this genetic conclusion. relationship are one unequivocal (51: state 0) and three equivocal (18, 25, 48: state 1) syn- RELATIONSHIPS OF THE KHOOBUR PETROSALS apomorphies. Recent phylogenetic proposals In their description of a mammalian petro- have revived Broom's (1914), Kielan-Jawo- sal from Khoobur (PSS-MAE-104), Khoobur rowska's (1971), and Kermack and Kielan- petrosal 1 of the current report, Wible et al. Jaworowska's (1971) hypotheses of multitu- (1995) left the affinities of this specimen to berculate-monotreme affinities (Wible and Prototribosphenida and to triconodonts un- Hopson, 1993; Meng and Wyss, 1995). Such resolved. Subsequently, a second petrosal relationships are not sustained by our most was identified from the same locality (PSS- parsimonious trees (fig. 6), although a mono- MAE- 1 9), referred to here as Khoobur pe- phyletic Monotremata/Multituberculata ap- trosal 2. In the current analysis, these two pears in six of the 123 trees one step longer. specimens are weakly linked in a monophyletic group (appendix 3: node 3), which in MONOTREMATA turn is the sister group of Prototribosphenida. Significant differences in the auditory and Supporting affinities with prototribospheni- orbitotemporal regions of the platypus and dans are five unequivocal synapomorphies echidnas have been reported previously and one that is equivocal (appendix 3: node (Eschweiler, 1899; Kuhn, 1971; Kuhn and 4), all of which relate to the enclosure of the perilymphatic duct or to the development of Zeller, 1987; Zeller, 1989, 1991; Wible and the posterior wall of the tympanic cavity. If Hopson, 1995). In fact, some of the differ- the Khoobur specimens and Prototribosphen- ences represent potential synapomorphies re- ida are monophyletic, then given the dental lating either the platypus or the echidnas to taxa known from Khoobur, these petrosals other taxa. Examples include the enclosure are those of of the perilymphatic duct in a canal in the likely symmetrodonts. However, echidnas, which also occurs in Prototribos- among the 123 trees one step longer than the + most parsimonious trees obtained here, vary- phenida the Khoobur petrosals, and the ing monophyletic groupings of the Khoobur common aperture of the prootic canal and ra- petrosals and triconodonts are identified. mus superior, which is found in the platypus Given this, Wible et al.'s (1995) uncertainties (fig. 5A) and multituberculates. In previous regarding the affinities of the Khoobur pe- cladistic analyses (e.g., Rowe, 1988; Wible trosals to either Prototribosphenida or tricon- and Hopson, 1993; Wible, et al., 1995), these odonts cannot be considered resolved. morphological discrepancies were addressed fully by reconstructing an ancestral state for Monotremata. This approach relied on a pre- MULTITUBERCULATE RELATIONSHIPS conceived notion about the direction of evo- Multituberculate ear regions are probably lutionary transformation in monotremes and the most distinctive of those pertaining to the yielded two sources of error: attribution of extinct taxa considered here. Of the numer- potentially false ancestral states to mono- ous multituberculate autapomorphies, the tremes and failure to test the monophyly of most striking are a lateral flange that is bent Monotremata. In order to avoid these poten- medially to contact the promontorium and an tial pitfalls, we have coded Ornithorhynchus extensive epitympanic recess on the ventro- and Tachyglossidae as separate OTU's in our lateral surface of the lateral flange. Theri- matrix (appendix 2). In the most parsimoni- iformes, the clade encompassing the last ous trees of our analysis, a monophyletic common ancestor of multituberculates and Monotremata is recovered and supported by therians plus all its descendants (Rowe, three unequivocal synapomorphies and one 1988), is diagnosed by four unequivocal and equivocal synapomorphy (appendix 3: node four equivocal synapomorphies (appendix 3: 9). However, in three of the 123 trees one node 8). As mentioned above, three of the step longer, a paraphyletic Monotremata is six most parsimonious trees of our analysis identified. The four ear region synapomor- ally multituberculates more closely with ther- phies recovered here are among the few 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 25 known for monotremes that can be directly 1988). Additionally, as stated above, the an- compared with those of more primitive out- teroposteriorly aligned cusps represent the groups and supplement those of the ear os- putative basal mammaliaform pattern (Kemp, sicles reported by Zeller (1993). 1983). Consequently, trechnotherian affini- Recently recovered dental remains of ex- ties of triconodonts and multituberculates tinct monotremes (Archer et al., 1985, 1992, cannot be sustained on dental grounds. As a 1993; Flannery et al., 1995) have been used corollary, the purported monotreme and to advance relationships for Monotremata to trechnotherian molar synapomorphies are a variety of taxa, all of which exhibit a re- also compromised. In fact, Archer et al. versed triangles molar pattern. Archer et al. (1992: 15) in their study of the dentition of (1985) proposed relationships with tribos- the Miocene platypus Obdurodon dicksoni phenidans; Kielan-Jaworowska et al. (1987) concluded that its "molar structure is mark- with advanced peramurids; Bonaparte (1990) edly autapomorphic and does not closely re- and Archer et al. (1992, 1993) with dryoles- semble that of any other group of mam- toids; and Hopson (1994) with symmetro- mals." donts. However, some authors (e.g., Archer et al., 1992; Flannery et al., 1995) have MAMMALIA raised concerns about the affinities of monotremes with trechnotherians, the clade in- Mammalia, the clade encompassing the cluding the last common ancestor of sym- last common ancestor of monotremes and metrodonts and therians plus all descendants therians plus all its descendants (Rowe, (McKenna, 1975). Flannery et al. (1995: 1988), is diagnosed by four unequivocal and 419) wrote that "the basic monotreme molar seven equivocal synapomorphies (appendix structure is not therian-like [holotherian], and 3: node 10). Of the four unequivocal char- that any other similarities between mono- acters, three are related to the lateral flange: tremes and therians [holotherians] may be its shape and relationship to the crista paro- convergent." tica and the quadrate ramus of the alisphen- In Wible et al. (1995), as in the current oid. The remaining one is a fingerlike pro- report, triconodonts along with Multituber- montorium. culata are placed between Prototribosphenida and Monotremata (fig. 6). Accepting holoth- RELATIONSHIPS OF LATE TRIASSIC AND erian affinities for monotremes under these JURASSIC MAMMALIAMORPHS phylogenetic hypotheses forces one to include multituberculates and/or triconodonts The relationships of the six remaining in- within Holotheria as well. Given our current group taxa are resolved as follows in all six understanding of mammaliamorph dental of the most parsimonious trees obtained here evolution, it seems unlikely that either of (fig. 6A): these two groups had a primitive triangular The docodont Haldanodon exspectatus arrangement of the main molar cusps. Tri- from the Late Jurassic of Portugal is identi- conodonts retain the putative primitive mam- fied as the sister taxon of Mammalia, sup- maliaform molar pattern in which the main ported by four unequivocal synapomorphies cusps are anteroposteriorly aligned (Kemp, and one equivocal synapomorphy (appendix 1983); this contrasts with the derived con- 3: node 11). Among the unequivocal char- dition present in tribosphenidans and their acters are an anteriorly positioned pterygo- early relatives. Dentally, multituberculates paroccipital foramen and a paroccipital pro- are the most distinctive of the Mesozoic cess extending ventral to the promontorium. mammals, with the earliest representatives Megazostrodon from the Early Jurassic of already showing the characteristic anteropos- southern Africa is the sister taxon of Mam- teriorly aligned rows of cusps on the molars. malia + Haldanodon. Supporting this unna- The typical multituberculate molar pattern med clade are one unequivocal and an equiv- shares no common ground for comparison ocal synapomorphy (appendix 3: node 12). with that of therians, and, therefore, hypoth- The former, a single trigeminal foramen pos- eses of homology are lacking (Simmons, terior to the alisphenoid, occurs only in Me- 26 AMERICAN MUSEUM NOVITATES NO. 3183 gazostrodon and Haldanodon among the in- pleteness in Wible et al.'s matrix is 94.2% group taxa. versus 86.7% here. The Liassic forms Morganucodon and The current results differ more dramati- Dinnetherium are identified as sister taxa, cally from those of Luo (1994), who found supported by one unequivocal and two little resolution among the Jurassic taxa and, equivocal synapomorphies (appendix 2: node in fact, had Triconodontidae nested within 13). Following Luo (1994), we employ the them. Using MacClade to obtain the arrange- term Morganucodontidae for this clade. In ment of Jurassic taxa congruent with that in turn, Morganucodontidae is the sister group Luo's preferred phylogenetic tree requires of Mammalia + Haldanodon + Megazostro- that 10 steps be added to our most parsi- don. This clade, Mammaliaformes of Rowe monious trees. (1988), is supported by two unequivocal As shown above, our analysis does not (elongate, straight cochlear duct and quadrate strongly justify the interrelationships of the articulation on the petrosal) and two equivocal synapomorphies (appendix 2: node 14). five pre-Haldanodon ingroup taxa, with The Liassic form Sinoconodon is the sister only one or two unequivocal synapomor- taxon of Mammaliaformes, based on one un- phies identified at each node. With the ex- equivocal (para/basisphenoid wing absent) ception of Morganucodon, these taxa exhib- and four equivocal synapomorphies (appen- it some missing data for the basicranial dix 2: node 15). characters considered here (with the high in Finally, Adelobasileus from the Carnian of Dinnetherium and Megazostrodon, which west Texas is the sister taxon of Mammali- are scored for only two-thirds of the char- aformes + Sinoconodon. Supporting this ar- acters). In contrast to the labile relationships rangement are six unequivocal and three of the basal taxa, the position of Haldano- equivocal synapomorphies (appendix 2: node don as the outgroup to Mammalia is more 16), with the former including a single hy- firmly founded in our analysis. This is con- poglossal foramen, a promontorium, and a gruent with the results of Wible et al. (1995) cavum epiptericum partially floored by the and accords with the general view that do- petrosal. codonts are mammaliaforms, related to mor- Only two other cladistic analyses have ganucodontids or more derived (Crompton considered all six of the above mammalia- and Jenkins, 1979; Crompton and Sun, morph taxa: Luo (1994) and Wible et al. 1985; McKenna, 1987). A dramatically dif- (1995). Of these two, the results of the cur- ferent hypothesis was offered by Lillegrav- rent report most closely resemble those of en and Krusat (1991) who placed Haldan- Wible et al. (1995), which also identified odon as the outgroup to Sinoconodon and Sinoconodon and Adelobasileus as the suc- Morganucodon. Their hypothesis, however, cessive outgroups to Mammaliaformes and has the merit of being founded in a data set included Dinnetherium, Megazostrodon, based on the anatomy of the whole skull and and Haldanodon within Mammaliaformes. the drawback of too few taxa considered. However, Wible et al. (1995) differed in their support of a clade of Megazostrodon Using MacClade to obtain Lillegraven and and Haldanodon, which in turn was at a tri- Krusat's arrangement of taxa requires that chotomy with Dinnetherium and Mammalia. eight steps be added to the most parsimo- To obtain this arrangement in MacClade, nious trees obtained here. four steps must be added to the most par- With regard to the outgroups employed simonious trees in the current report. The here, there is current debate as to whether differences between Wible et al. (1995) and tritheledontids or tritylodontids are more the current report result from the inclusion closely related to mammaliaforms (see here of more ingroup taxa (18 versus 15) Rowe, 1993; Crompton and Luo, 1993; Hop- and characters (51 versus 44). Scoring four son, 1994; Luo, 1994). This controversy was separate OTU's for triconodonts in the cur- not addressed in our analysis, which in fact rent matrix has slightly increased the would require the inclusion of more non- amount of missing data; the average com- mammaliamorph cynodonts. 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 27

IMPLICATIONS OF THE PAUP ANALYSIS FOR blances in middle-ear morphology character- TAXA KNOWN ONLY FROM TEETH AND JAWS izing monotremes, multituberculates, and therians (Gaupp, 1913; Miao and Lillegraven, Despite the recent dramatic increase in 1986; Meng and Wyss, 1995) have been con- knowledge of the basicranium of early mam- sidered to strongly support a single origin of mals and their relatives, the overwhelming the definitive triossicular middle ear (Patter- majority of Mesozoic mammaliaform taxa son, 1981; Kemp, 1983; Novacek, 1993; are known exclusively from teeth and frag- Rowe, 1995). The tree topology recovered mentary jaws. This is particularly true for here is fully congruent with the latter alter- trechnotherians, because the only petrosals native, a single origin of ear ossicles suspend- associated with dentitions for a trechnother- ed from the skull base, which would be di- ian prior to the Late Cretaceous are those of agnostic of Mammalia. This suggests that the Early Cretaceous Vincelestes. Thus, our anal- triangular arrangement of molar cusps in ysis has had to omit such traditionally im- Kuehneotherium is convergent on that in ther- portant trechnotherians as symmetrodonts, ians or that it is secondarily lost in multitu- dryolestoids, Peramus, and Amphitherium. berculates and triconodonts, and is no longer Another influential taxon omitted is Kueh- recognizable in monotremes. The latter alter- neotherium praecursoris from the Liassic of native seems unlikely for multituberculates Wales (Kermack et al., 1968). Kuehneother- and triconodonts and is, at present, equivocal ium has its main molar cusps forming a tri- for monotremes. angle, which has been interpreted as a de- Kuehneotherium is not the only Late Tri- rived feature supportive of affinities with assic-Jurassic taxon known exclusively from trechnotherians (Kermack et al., 1968; teeth and jaws to present such problems. Crompton, 1971; Prothero, 1981). In addi- Shuotherium from the late Middle or early tion, the lower jaws attributed to Kuehneoth- Late Jurassic of China has molars resembling erium retain a deep mandibular sulcus and those of Kuehneotherium (Hopson, 1995) medial dentary ridge which, when compared and shows evidence for the retention of post- with contemporary taxa such as Morganu- dentary bones (Chow and Rich, 1982). Con- codon, indicate that the postdentary bones re- sequently, the previous comments regarding tained the primitive attachment to the dentary the conflicting characters of dental and lower (Kermack et al., 1968; Cassiliano and Clem- jaw morphology made for Kuehneotherium ens, 1979). This implies that, like Morgan- also apply for Shuotherium. Woutersia from ucodon, Kuehneotherium lacked the defini- the Rhaetic of France is known only from tive triossicular middle-ear system (Allin and isolated teeth, and, though originally includ- Hopson, 1992). ed in Kuehneotheriidae (Sigogneau-Russell, Integrating Kuehneotherium into recently 1983), has recently been assigned to a family published cladograms of mammaliaform re- of its own, Woutersiidae, within Holotheria lationships (Rowe, 1988, 1993; Wible, 1991; (Sigogneau-Russell and Hahn, 1995). The Wible and Hopson, 1993; Wible et al., 1995), latter authors have also observed that Wou- including the ones advanced here, is problem- tersia shares some derived features of the up- atic because of the conflicting distribution of per molars with docodonts. Although doco- characters of the molars and the lower jaw dont affinities for Woutersia were ultimately (and, by inference, the middle ear). Accepting excluded by Sigogneau-Russell and Hahn the topology of our most parsimonious trees, (1995), the observed similarities between if Kuehneotherium is more closely related to these taxa emphasize that alternative inter- therians than are multituberculates, tricono- pretations for the tooth morphology of kueh- donts, and monotremes, as suggested by the neotheriids are feasible. putative molar synapomorphies, then the de- In contrast to the abovementioned hypoth- tachment of the postdentary bones from the esis on the single origin of the definitive lower jaw as strictly auditory structures would triossicular middle ear, Allin and Hopson have occurred independently in multituber- (1992) proposed a polyphyletic origin of culates, triconodonts, monotremes, and theri- these elements. They accepted the dental evans. On the other hand, the striking resem- idence that allies Kuehneotherium to therians 28 AMERICAN MUSEUM NOVITATES NO. 3183 and proposed the independent separation of erly triconodonts are highly labile here. In the articular and angular (malleus and ecto- the 123 trees one step longer than the six tympanic) from the lower jaw in at least two most parsimonious ones, these taxa are either and as many as four clades of mammals. Part in a paraphyletic grouping with the Khoobur of their argument was based on their inter- petrosals or they are the sister group to Pro- pretation of depressions on the inner surface totribosphenida + the Khoobur petrosals, to of the dentary of the trechnotherians Am- Theriiformes, or to Mammalia. Nevertheless, phitherium, and Peramus as surfaces for par- Triconodonta in the traditional sense (mor- tial attachment of the postdentary bones to ganucodontids + triconodonts) is not sup- the lower jaw. ported here, or in any other recent cladistic Resolution of these conflicting character analyses considering the pertinent taxa. sets awaits the discovery of more material We have recently published two other phy- and more comprehensive cladistic analyses logenetic analyses of the basicranium across that incorporate cranial, dental, and postcra- mammaliamorphs that consider fewer in- nial information from a wider array of taxa. group taxa and characters (Wible and Hop- In the meantime, however, even though only son, 1993; Wible et al., 1995). The phylo- a limited number of Mesozoic taxa are genetic results of the current report most known from petrosals and the relationships closely resemble those of Wible et al. (1995). of some mammaliaform groups are not yet The main difference is that here all taxa are fully resolved, it is promising to note that fully resolved, except the triconodonts and there is some general agreement among trees multituberculates, whereas the strict consen- produced by characters of the (1) basicran- sus tree in Wible et al. (1995) had three un- ium (Wible and Hopson, 1993; Wible et al., resolved trichotomies between Mammali- 1995; this study), (2) cranium (Wible, 1991), aformes and Prototribosphenida. However, and (3) cranium plus postcranium (Rowe, the relationships of these mammaliaform 1988, 1993; Rougier, 1993). taxa are not well resolved in the current report, as evidenced by the number of nodes CONCLUSIONS supported by only one or two unequivocal synapomorphies. Moreover, adding one step Our phylogenetic analysis of basicranial to the most parsimonious trees identified here characters allies the triconodontids Priaco- produces the overall topology of the most donfruitaensis from the Late Jurassic of Col- parsimonious trees of Wible et al. (1995). orado with Trioracodon ferox and Tricono- Adding data that are currently missing from don mordax from the Late Jurassic of En- our analysis might increase tree stability, al- gland. The nearest outgroup to these three though the present taxon-character matrix is taxa is the unnamed triconodonts from the about 87% complete. Alternatively, resolu- Early Cretaceous of Montana. Whether or tion awaits the addition of information from not the unnamed triconodonts belong in Tri- other anatomical systems into phylogenetic conodontidae is dependent on their relation- analyses. The tree topologies resulting from ships to taxa not yet known from basicrania: this study are fully congruent with the single the two remaining triconodontid genera, Al- origin of the definitive triossicular middle ticonodon and Astroconodon, as well as the ear, but opposing hypotheses cannot be elim- three remaining triconodont families, Am- inated until some critical taxa, insufficiently philestidae, Austrotriconodontidae, and Gob- known, such as Kuehneotherium and Shuoth- iconodontidae. The three triconodontids + erium, are represented by more complete ma- the Cloverly triconodonts are identified as terial. mammals, but their relationships within Mammalia are not fully resolved. In three of ACKNOWLEDGMENTS the six most parsimonious trees, they are more closely related to therians than are mul- Our foremost thanks are to George Calli- tituberculates; the reverse is true in the re- son and the Natural History Museum of Los maining three trees. However, the relation- Angeles County for the opportunity to study ships of the three triconodontids + the Clov- the cranial remains of Priacodon fruitaensis. 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 29

For access to additional materials, we are Heck for labeling of some figures, and E. F grateful to M. J. Novacek, M. C. McKenna, Allin, J. F. Bonaparte, Z. Kielan-Jaworowska, and R. D. E. MacPhee, American Museum Z. Luo, and J. Meng for comments and im- of Natural History, New York; J. F. Bona- provements on an earlier version. This re- parte, Museo Argentino de Ciencias Natura- search was supported by NSF Grants BSR les, Buenos Aires; J. J. Flynn, Field Museum 8906619, BSR 9119212, and DEB 9625431. of Natural History, Chicago; A. W. Cromp- We also acknowledge NSF Grant DEB ton and F A. Jenkins, Jr., Museum of Com- 930070 for the three specimens collected in parative Zoology, Harvard University, Cam- the Gobi Desert by the Joint Paleontological bridge; D. Dashzeveg, Geological Institute, Expeditions of the Mongolian Academy of Mongolian Academy of Sciences, Ulaan Sciences and the American Museum of Nat- Baatar; and Z. Kielan-Jaworowska, Institute ural History. Finally, G. W. Rougier's re- of Paleobiology, Polish Academy of Sci- search was supported by a Frick Research ences, Warsaw. We also thank Ms. Claire Fellowship from the AMNH and by CONI- Vanderslice for the exquisite artwork, Ed CET, Argentina.

REFERENCES Allin, E. F, and J. A. Hopson Argentina. Actas IV Congr. Arg. Pa- 1991. Evolution of the auditory system in leontol. Biostratigr. 2: 48-61. Synapsida ("mammal-like reptiles" 1990. New Late Cretaceous mammals from and primitive mammals) as seen in the the Los Alamitos Formation, Northern fossil record. In D. B. Webster, R. R. Patagonia. Nat. Geogr. Res. 6: 63-93. Fay, and A. N. Popper (eds.), The evo- 1992. Una nueva especie de Triconodonta lutionary biology of hearing, pp. 587- (Mammalia), de la Formaci6n Los Al- 614. New York: amitos, provincia de Rio Negro y co- Springer-Verlag. mentarios sobre su fauna de mamiferos. Archer, M., T. F Flannery, A. Ritchie, and R. E. Ameghiniana 29: 99-110. Molnar Broom, R. 1985. First Mesozoic mammals from Austra- 1914. On the structure and affinities of the lia-an Early Cretaceous monotreme. Multituberculata. Bull. Am. Mus. Nat. Nature 318: 363-366. Hist. 33: 115-134. Archer, M., F A. Jenkins, Jr., S. J. Hand, P. Mur- Butler, P. M. ray, and H. Godthelp 1939. The teeth of the Jurassic mammals. 1992. Description of the skull and non-vestig- Proc. Zool. Soc. B109: 329-356. ial dentition of a Miocene platypus Carroll, R. L. (Obdurodon dicksoni n. sp.) from Riv- 1988. Vertebrate paleontology and evolution. ersleigh, Australia, and the problem of New York: W. H. Freeman. Cassiliano, M. L., and W. A. Clemens monotreme origins. In M. Augee (ed.), 1979. Symmetrodonta. In J. A. Lillegraven, Platypus and echidnas, pp. 15-27. Syd- Z. Kielan-Jaworowska, and W. A. ney: Royal Zoological Society of New Clemens (eds.), Mesozoic mammals: South Wales. the first two-thirds of mammalian his- Archer, M., P. Murray, S. Hand, and H. Godthelp tory, pp. 150-161. Berkeley: Univ. 1993. Reconsideration of monotreme rela- California Press. tionships based on the skull and denti- Chow, M., and T. H. Rich tion of the Miocene Obdurodon dick- 1982. Shuotherium dongi, n. gen. and sp., a soni. In F S. Szalay, M. J. Novacek, therian with pseudo-tribosphenic mo- and M. C. McKenna (eds.), Mammal lars from the Jurassic of Sichuan, Chi- phylogeny: Mesozoic differentiation, na. Aust. Mammal. 5: 127-142. Clark, J. M., and J. A. Hopson multituberculates, monotremes, early 1985. Distinctive mammal-like reptile from therians, and marsupials, pp. 75-94. Mexico and its bearing on the phylog- New York: Springer-Verlag. eny of Tritylodontidae. Nature 315: Bonaparte, J. F 398-400. 1986. Sobre Mesungulatum houssayi nuevos Crompton, A. W. mamiferos Cretaicicos de Patagonia, 1958. The cranial morphology of a new genus 30 AMERICAN MUSEUM NOVITATES NO. 3183

and species of ictidosaurian. Proc. tris. Pt. 2. Die vergleichenden Ontoge- Zool. Soc. London 130: 183-216. ne der neurokraniellen Venen der Vogl 1964. On the skull of Oligokyphus. Bull. Br. und Saugetiere. Z. Anat. Entwicke- Mus. Nat. Hist. 9: 69-82. lungsgesch. 74: 432-508. 1971. The origin of the tribosphenic molar. In Goodrich, E. S. D. M. Kermack and K. A. Kermack 1930. Studies on the structure and develop- (eds.), Early mammals. Zool. J. Linn. ment of vertebrates. London: Macmil- Soc. 50, suppl. 1: 65-87. lan. Crompton, A. W., and F A. Jenkins, Jr. Gow, C. E. 1979. Origin of mammals. In J. A. Lillegrav- 1986a. The side wall of the braincase in cy- en, Z. Kielan-Jaworowska, and W. A. nodont therapsids and a note on the ho- Clemens (eds.), Mesozoic mammals: mology of the mammalian promonto- the first two-thirds of mammalian his- rium. S. Afr. J. Zool. 21: 136-148. tory, pp. 59-73. Berkeley: Univ. Cali- 1986b. A new skull of Megazostrodon (Mam- fornia Press. malia, Triconodonta) from the Elliot Crompton, A. W., and Z. Luo Formation (Lower Jurassic) of southern 1993. Relationships of the Liassic mammals Africa. Palaeontol. Afr. 26: 13-23. Sinoconodon, Morganucodon oehleri, Hahn, G. and Dinnetherium. In F S. Szalay, M. 1988. Die Ohr-region der Paulchoffatiidae J. Novacek, and M. C. McKenna (eds.), (Multituberculata, Ober-Jura). Palaeo- Mammal phylogeny: Mesozoic differen- vertebrata, Montpellier 18: 155-185. tiation, multituberculates, monotremes, Hill, J. E. early therians, and marsupials, pp. 30- 1935. The cranial foramina in rodents. J. 44. New York: Springer-Verlag. Mammal. 14: 121-129. Crompton, A. W., and A. L. Sun Hopson, J. A. 1985. Cranial structure and relationships of 1964. The braincase of the advanced mam- the Liassic mammal Sinoconodon. mal-like reptile Bienotherium. Postilla Zool. J. Linn. Soc. 85: 99-119. 87: 1-30. Edgeworth, F H. 1970. The classification of non-therian mam- 1914. On the development and morphology mals. J. Mammal. 51: 1-9. of the mandibular and hyoid muscles of 1994. Synapsid evolution and the radiation of mammals. Q. J. Microsc. Sci. 59: 573- non-eutherian mammals. In R. S. Spen- 654. cer (ed.), Major features of vertebrate Eschweiler, R. evolution. Paleontol. Soc. Short Cours- 1899. Die Fenestra cochleae bei Echidna hys- es in Paleontol. 7: 190-219. trix. Anat. Anz. 16: 584-590. 1995. The Jurassic mammal Shuotherium Evans, H. E., and G. C. Christensen dongi: "pseudo-tribosphenic therian." 1979. Miller's anatomy of the dog. Philadel- J. Vertebr. Paleontol. 15: 36A. phia, PA: W. B. Saunders. Hopson, J. A., and A. W. Crompton 1969. Origin of mammals. Evol. Biol. 3: 15- Flannery, T F, M. Archer, T. H. Rich, and R. 72. Jones Hopson, J. A., and G. W. Rougier 1995. A new family of monotremes from the 1993. Braincase structure in the oldest known Cretaceous of Australia. Nature 377: skull of a therian mammal: implications 418-420. for mammalian systematics and cranial Gaupp, E. evolution. Am. J. Sci. 293A: 1-216. 1913. Die Reichertsche Theorie (Hammer-, Hunt, R. M., Jr. Amboss- und Kieferfrage). Arch. Anat. 1977. Basicranial anatomy of Cynelos Jour- Entwickelungsgesch., suppl. 1912: 1- dan (Mammalia: Carnivora), an Aqui- 416. tanian amphicyonid from the Allier Ba- Gauthier, J. sin, France. J. Paleontol. 51: 826-843. 1986. Saurischian monophyly and the origin Jenkins, F A., Jr., and A. W. Crompton of birds. California Acad. Sci. Mem. 8: 1979. Triconodonta. In J. A. Lillegraven, Z. 1-56. Kielan-Jaworowska, and W. A. Clem- Gauthier, J., A. G. Kluge, and T Rowe ens (eds.), Mesozoic mammals: the first 1988. Amniote phylogeny and the importance two-thirds of mammalian history, 74- of fossils. Cladistics 4: 105-209. 90. Berkeley: Univ. California Press. Gelderen, C. van Jenkins, F A., Jr., and C. R. Schaff 1924. Die Morphologie der Sinus durae ma- 1988. The Early Cretaceous mammal Gobi- 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 31

conodon from the Cloverly Formation and inner ear of Late Jurassic multitu- in Montana. J. Vertebr. Paleontol. 8: 1- berculates. J. Mammal. Evol. 1: 47-74. 24. Lillegraven, J. A., and G. Krusat Kemp, T S. 1991. Cranio-mandibular anatomy of Haldan- 1983. The relationships of mammals. Zool. J. odon exspectatus (Docodonta; Mam- Linn. Soc. 77: 353-384. malia) from the Late Jurassic of Por- Kermack, D. M., K. A. Kermack, and F Mussett tugal and its implications to the evolu- 1968. The Welsh pantothere Kuehneotherium tion of mammalian characters. Contrib. praecursoris. Zool. J. Linn. Soc. 47: Geol. Univ. Wyoming 28: 39-138. 407-423. Lucas, S. G., and Z. Luo Kermack, K. A. 1993. Adelobasileus from the Upper Triassic 1963. The cranial structure of triconodonts. of west Texas: the oldest mammal. J. Philos. Trans. R. Soc. London B246: Vertebr. Paleontol. 13: 309-334. 83-102. Luo, Z. 1967. The interrelations of early mammals. 1994. Sister taxon relationships of mammals Zool. J. Linn. Soc. 47: 241-249. and transformations of diagnostic Kermack, K. A., and Z. Kielan-Jaworowska mammalian characters. In N. C. Fraser 1971. Therian and non-therian mammals. In and H.-D. Sues (eds.), Life in the shad- D. M. Kermack and K. A. Kermack ow of dinosaurs: early Mesozoic tetra- (eds.), Early mammals. Zool. J. Linn. pods, pp. 98-128. Cambridge: Cam- Soc. 50, suppl. 1: 103-115. bridge Univ. Press. Kermack, K. A., F Mussett, and H. W. Rigney Luo, Z., and A. W. Crompton 1981. The skull of Morganucodon. Zool. J. 1994. Transformation of the quadrate (incus) non-mam- Linn. Soc. 71: 1-158. through the transition from malian cynodonts to mammals. J. Ver- Kielan-Jaworowska, Z. tebr. Paleontol. 14: 341-374. 1971. Skull structure and affinities of the Luo, Z., A. W. Crompton, and S. G. Lucas Multituberculata. Palaeontol. Polonica 1995. Evolutionary origins of the mammalian 25: 5-41. promontorium. J. Vertebr. Paleontol. 1992. Interrelationships of Mesozoic mam- 15: 113-121. mals. Hist. Biol. 6: 185-201. Maclntyre, G. T. Kielan-Jaworowska, Z., A. W. Crompton, and F 1972. The trisulcate petrosal pattern of mam- A. Jenkins, Jr. mals. Evol. Biol. 6: 275-303. 1987. The origin of egg-laying mammals. Na- MacPhee, R. D. E., M. J. Novacek, and G. Storch ture 326: 871-873. 1988. Basicranial morphology of early Terti- Kielan-Jaworowska, Z., R. Presley, and C. Poplin ary erinaceomorphs and the origin of 1986. The cranial vascular system in taeniol- primates. Am. Mus. Novitates 2921: 42 abidoid multituberculate mammals. Phi- PP. los. Trans. R. Soc. London B313: 525- Maddison, W. P., and D. R. Maddison 602. 1992. MacClade: analysis of phylogeny and Kuhn, H.-J. character evolution. Version 3.01. Sun- 1971. Die Entwicklung und Morphologie des derland, MA: Sinauer Associates. Schadels von Tachyglossus aculeatus. McKenna, M. C. Abh. Senckenb. Naturforsch. Ges. 528: 1975. Towards a phylogenetic classification 1-224. of the Mammalia. In W. P. Luckett and Kuhn, H.-J., and U. Zeller F S. Szalay (eds.), Phylogeny of the 1987. The cavum epiptericum in monotremes primates: a multidisciplinary approach, and therian mammals. In H.-J. Kuhn pp. 21-46. New York: Plenum Press. and U. Zeller (eds.), Morphogenesis of 1987. Molecular and morphological analysis the mammalian skull. Mammalia De- of higher-level mammalian interrela- picta 13: 50-70. Hamburg: Paul Parey. tionships. In C. Patterson (ed.), Mole- Kuhne, W. G. cules and morphology in evolution: 1956. The Liassic therapsid Oligokyphus, 149 conflict or compromise, 55-93. Cam- pp. London: Trustees of the British Mu- bridge: Cambridge Univ. Press. seum. Meng, J., and R. C. Fox Lillegraven, J. A., and G. Hahn 1993. Inner ear structures from the Late Cre- 1993. Evolutionary analysis of the middle taceous mammals and their systematic 32 AMERICAN MUSEUM NOVITATES NO. 3183

and functional implications. J. Vertebr. therapsids. Mem. Soc. Geol. France N. Paleontol. 15: 122-130. S. 139: 159-162. 1995a. Osseous inner ear structures and hear- Prothero, D. R. ing in early marsupials and placentals. 1981. New Jurassic mammals from Como Zool. J. Linn. Soc. 115: 47-71. Bluff, Wyoming, and the interrelation- 1995b. Evolution of the inner ear from non- ships of Non-Tribosphenic Theria. Bull. therians to therians during the Meso- Am. Mus. Nat. Hist. 167: 277-326. zoic: implications for mammalian phy- Quiroga, J. C. logeny and hearing. In A. Sun and Y 1979. The inner ear of two cynodonts (Rep- Wang (eds.), Sixth symposium on Me- tilia-Therapsida) and some comments sozoic terrestrial ecosystems and biota, on the evolution of the inner ear from pp. 235-242. Beijing: China Ocean pelycosaurs to mammals. Gegenbaurs Press. Morphol. Jahrb. 125: 178-190. Meng, J., and A. R. Wyss Rasmussen, T. E., and G. Callison 1995. Monotreme affinities and low-frequen- 1981. A new species of triconodont mammal cy hearing suggested by multitubercu- from the Upper Jurassic of Colorado. J. late ear. Nature 377: 141-144. Paleontol. 55: 628-634. Miao, D. Rougier, G. W. 1988. Skull morphology of Lambdopsalis 1993. Vincelestes neuquenianus Bonaparte bulla (Mammalia, Multituberculata) un mami- and its implications to mammalian evo- (Mammalia, Theria) primitivo lution. Contrib. Geol. Univ. Wyoming fero del Cretacico Inferior de la Cuenca Spec. Pap. 4: 1-104. Neuquina, 720 pp. Diss., Univ. of Bue- Miao, D., and J. A. Lillegraven nos Aires. 1986. Discovery of three ear ossicles in a Rougier, G. W., J. R. Wible, and J. A. Hopson multituberculate mammal. Nat. Geogr. 1992. Reconstruction of the cranial vessels in Res. 2: 500-507. the Early Cretaceous mammal Vince- Mills, J. R. E. lestes neuquenianus: implications for 1971. The dentition of Morganucodon. In D. the evolution of the mammalian cranial M. Kermack and K. A. Kermack (eds.), vascular system. J. Vertebr. Paleontol. Early mammals. Zool. J. Linn. Soc. 50, 12: 188-216. suppl. 1: 29-63. Rowe, T. Nixon, K. C., and J. M. Carpenter 1987. Definition and diagnosis in the phylo- 1993. On outgroups. Cladistics 9: 413-426. genetic system. Syst. Zool. 36: 208- Novacek, M. J. 211. 1986. The skull of leptictid insectivorans and 1988. Definition, diagnosis and origin of the higher-level classification of euther- Mammalia. J. Vertebr. Paleontol. 8: ian mammals. Bull. Am. Mus. Nat. 241-264. Hist. 183: 1-112. 1993. Phylogenetic systematics and the early 1993. Patterns of diversity in the mammalian history of mammals. In F S. Szalay, skull. In J. Hanken and B. K. Hall M. J. Novacek, and M. C. McKenna (eds.), The skull, Vol. 2, Patterns of (eds.), Mammal phylogeny: Mesozoic structural and systematic diversity, pp. differentiation, multituberculates, mon- 438-545. Chicago, IL: Univ. Chicago otremes, early therians, and marsupials, Press. pp. 129-145. New York: Springer-Ver- Patterson, B., and E. C. Olson lag. 1961. A triconodontid mammal from the Tri- assic of Yunnan. In G. Vandebroek (ed.), 1995. Brain heterochrony and origin of the International colloquium on the evolution mammalian middle ear: new data from of lower and specialized mammals, 129- high resolution Xray CT. J. Vertebr. Pa- 191. Brussels: K. Vlaamse Acad. We- leontol. 15(3) suppl.: 50A. tensch., Lett. Schone Kunsten Belgie. Segall, W. Patterson, C. 1970. Morphological parallelisms of the bulla 1981. Methods of paleobiogeography. In G. and auditory ossicles in some insecti- Nelson and D. E. Rosen (eds.), Vicar- vores and marsupials. Fieldiana Zool. iance biogeography: a critique, 446- 51: 169-205. 489. New York: Columbia Univ. Press. Sereno, P. C., and M. C. McKenna Presley, R. 1995. Cretaceous multituberculate skeleton 1980. The braincase in recent and Mesozoic and the early evolution of the mam- 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 33

malian shoulder girdle. Nature 377: simony, version 3.1.1. Washington, D. 144-147. C.: Smithson. Inst. Shindo, T. Wible, J. R. 1915. Uber die Bedeutung des Sinus caver- 1983. The internal carotid artery in early nosus der Sauger mit vergleichend an- eutherians. Acta Palaeotol. Polonica atomischer Berucksichtigung anderer 28: 281-293. Kopfvenen. Anat. Hefte 52: 319-495. 1987. The eutherian stapedial artery: charac- Sigogneau-Russell, D. ter analysis and implications for super- 1983. A new therian mammal from the ordinal relationships. Zool. J. Linn. Rhaetic locality of Saint-Nicholas-de- Soc. 91: 107-135. Port (France). Zool. J. Linn. Soc. 78: 1990. Petrosals of Late Cretaceous marsupials 175-186. from North America and a cladistic Sigogneau-Russell, D., and R. Hahn analysis of the petrosal in therian mam- 1995. Reassessment of the Late Triassic sym- mals. J. Vertebr. Paleontol. 10: 183- metrodont mammal Woutersia. Acta 205. Palaeontol. Polonica 40: 245-260. 1991. Origin of Mammalia: the craniodental Simmons, N. evidence reexamined. J. Vertebr. Pa- 1988. Patterns of tooth eruption and replace- leontol. 11: 1-28. ment in multituberculate mammals. J. Wible, J. R., and J. A. Hopson Vertebr. Paleontol. 8: 89-96. 1993. Basicranial evidence for early mammal Simpson, G. G. phylogeny. In F. S. Szalay, M. J. No- 1925. Mesozoic Mammalia. I. American tri- vacek, and M. C. McKenna (eds.), conodonts. Am. J. Sci. 10: 145-165. Mammal phylogeny: Mesozoic dif- 1928. A catalogue of the Mesozoic Mamma- ferentiation, multituberculates, mono- lia in the geological department of the tremes, early therians, and marsupials, British Museum. London: Trustees of 45-62. New York: Springer-Verlag. the Br. Mus. 1995. The homologies of the prootic canal in 1929. American Mesozoic Mammalia. Mem. mammals and non-mammalian cyno- Peabody Mus. Yale Univ. 3: 1-235. donts. J. Vertebr. Paleontol. 15: 331- 1937. Skull structure of the Multituberculata. 356. Bull. Am. Mus. Nat. Hist. 73: 727-763. Wible, J. R., G. W. Rougier, M. J. Novacek, M. 1961. Evolution of Mesozoic mammals. In G. C. McKenna, and D. Dashzeveg Vandebroek (ed.), International collo- 1995. A mammalian petrosal from the Early quium on the evolution of lower and Cretaceous of Mongolia: implications specialized mammals, 57-95. Brussels: for the evolution of the ear region and K. Vlaamse Acad. Wetensch., Lett. mammaliamorph interrelationships. Am. Schone Kunsten Belgie. Mus. Novitates 3149: 19 pp. Stucky, R. K., and M. C. McKenna U. 1993. Mammalia. In M. J. Benton (ed.), The Zeller, fossil record, 739-771. London: Chap- 1985. Die Ontogenese und Morphologie der man & Hall. Fenestra rotunda und des Aquaeductus Sues, H.-D. cochleae von Tupaia und anderen 1983. Advanced mammal-like reptiles from Saugern. Gegenbaurs Morphol. Jahrb. the Early Jurassic of Arizona, 302 pp. 131: 179-204. Diss., Harvard Univ. 1989. Die Entwicklung und Morphologie des 1986. The skull and dentition of two tritylo- Schadels von Ornithorhynchus anatinus dontid synapsids from the Lower Juras- (Mammalia: Prototheria: Monotremata). sic of western North America. Bull. Abh. Senckenb. Naturforsch. Ges. 545: Mus. Comp. Zool. 151: 217-268. 1-188. Sun, A. 1991. Foramen perilymphaticum und Reces- 1984. Skull morphology of the tritylodont ge- sus scalae tympani von Ornithorhyn- nus Bienotheroides of Sichuan. Sci. chus anatinus (Monotremata) und an- Sin. B27: 970-984. deren Saugern. Verh. Anat. Ges. 84 Sun, A., and Y. Li (Anat. Anz. Suppl. 168): 441-443. 1985. The postcranial skeleton of the late tri- 1993. Ontogenetic evidence for cranial ho- tylodont Bienotheroides. Vertebr. Pala- mologies in monotremes and therians, siat. 23: 135-151. with special reference to Ornithorhyn- Swofford, D. L. chus. In F S. Szalay, M. J. Novacek, 1993. PAUP: phylogenetic analysis using par- and M. C. McKenna (eds.), Mammal 34 AMERICAN MUSEUM NOVITATES NO. 3183

phylogeny: Mesozoic differentiation, 18. Caudal tympanic process of petrosal-absent multituberculates, monotremes, early (0) or present (1). therians, and marsupials, 95-128. New 19. Post-promontorial tympanic recess absent York: Springer-Verlag. (0) or present (1). 20. Deep pocket medial to paroccipital process- APPENDIX 1 absent (0) or present (1). 21. Tympanohyal-separate (0), attached to pe- BASICRANIAL CHARACTERS trosal, posteromedially directed (1), or at- 1. Petrosal promontorium-absent (0) or present tached to petrosal, medially directed (2). (1). 22. Tympanohyal/promontorial contact-absent 2. Para/basisphenoid wing-present (0) or ab- (0) or present (1). sent (1). 23. Crista parotica/squamosal contact-present 3. Cochlear housing shape-indistinct (0), tri- (0) or absent (1). angular in outline, wider posteriorly with 24. Fossa incudis or quadrate articulation on pe- steep lateral wall (1), triangular in outline, trosal-absent (0) or present (1). wider posteriorly without steep lateral wall 25. Epitympanic recess on petrosal-absent (0) (2), or finger-like in outline (3). or present (1). 4. Ventral crest on cochlear housing-absent (0) 26. Cranial aperture of prootic canal-absent (0), or present (1). in rear of cavum epiptericum (1), between the 5. Small foramen on cochlear housing antero- cavum and anterodorsal margin of subarcuate medial to fenestra vestibuli-absent (0) or fossa (2), or at anterodorsal margin of subar- present (1). cuate fossa (3). 6. Cochlear duct-short (0), elongate and 27. Tympanic aperture of prootic canal-absent straight (1), bent (2), or coiled, 3600 or more (0), separate from (1), or confluent with pter- (3). ygoparoccipital foramen (2). 7. Secondary spiral lamina-absent (0) or pres- 28. Pterygoparoccipital foramen-at same level ent (1). or posterior to fenestra vestibuli (0), anterior 8. Semicircular canals enclosed by petrosal to it (1), or absent (2). only-absent (0) or present (1). 29. Lateral flange-laterally directed shelf (0), 9. Fenestra vestibuli shape-rounded, stapedial with vertical and horizontal components in ratio less than 1.2 (0) or ellipsoidal, stapedial lateral view (1), ventrally directed crest (2), ratio greater than 1.4 (1). contacting cochlear housing (3), or absent (4). 10. Fenestra vestibuli margin-with thickened 30. Lateral flange/crista parotica relationship- osseous ring (0) or without osseous ring (1). narrowly separated (0), widely separated (1), 11. Channel for perilymphatic duct-no indica- continuous (2), or lateral flange greatly re- tion (0), open sulcus (1), partially enclosed by duced (3). bony lappets (2), or fully enclosed to form a 31. Vascular foramen in lateral flange-absent canal, a cochlear aqueduct (3). (0) or present (1). 12. Lateral opening of cochlear aqueduct-absent 32. Post-trigeminal canal-absent (0) or present (0), medial (external) to perilymphatic fora- (1). men (1), or within inner ear (2). 33. Facial ganglion floor-open ventrally (0), 13. Jugular fossa including jugular foramen and broad petrosal bridge, flooring part of hyo- perilymphatic foramen/fenestra cochleae- mandibular and palatine rami (1), or narrow present (0) or absent (1). petrosal bridge, flooring only primary facial 14. Jugular foramen size relative to perilymphatic foramen (2). foramen/fenestra cochleae-equal to or larger 34. Lateral trough of petrosal-absent (0) or (0) or smaller (1). present (1). 15. Crista interfenestralis-tall and horizontal, 35. Cavum epiptericum floor-open ventrally (0), extending to base of paroccipital process (0), partially floored by petrosal (1), floored, pri- vertical, delimiting back of promontorium marily by petrosal (2), or floored, primarily (1), or not distinguishable from surrounding by alisphenoid (3). elements (2). 36. Tensor tympani fossa-indistinct (0), deep re- 16. Fossa for stapedius muscle-absent (0), on cess on lateral trough anterior to hiatus Fal- crista interfenestralis (1), or lateral to crista lopii (1), or deep recess on lateral trough pos- interfenestralis (2). terior to secondary facial foramen (2). 17. Paroccipital process-not extending ventral 37. Petrosal contact with quadrate ramus of alis- to level of cochlear housing (0) or extending phenoid (epipterygoid)-elongate (0) or ventral to level of cochlear housing (1). small or absent (1). 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 35

38. Internal acoustic meatus-without distinct tramural, between petrosal and squamosal (2), medial rim (0) or with medial rim (1). endocranial (3), or absent (4). 39. Pila antotica-ossified base present (0) or ab- 49. Orientation of ventral ascending groove or sent (1). canal-vertical (0), horizontal (1), or without 40. Wall separating cavum supracochleare from groove or canal (2). cavum epiptericum absent (0), present with 50. Dorsal ascending groove or canal-open lat- small fenestra (fenestra semilunaris) (1), or erally dorsal to posttemporal canal (0), intra- complete (2). mural, covered laterally by parietal and/or 41. Sulcus for sigmoid sinus extending to jugular squamosal (1), or endocranial (2). foramen-absent (0) or present (1). 51. Hypoglossal foramina-double or multiple 42. Sulcus for inferior petrosal sinus-enclosed (0), single (1), or confluent with jugular fo- in petrosal, laterally directed at rostral end ramen (2). (0), between petrosal and basioccipital, anteroposteriorly directed (1), or Most of the above characters are taken directly or endocranial are modified from the following references, to (2). which 43. Craniomandibular joint position relative to fe- the reader is referred for additional infor- nestra vestibuli-level with (0) or anterior to mation. Characters used for the first time are 5, (1). 12, 13, 14, 15, 20, 25, 30, 32, 41, and 46. 44. Thick, laterally directed surface of petrosal related to glenoid fossa-absent (0) or pres- Characters: ent (1). 1, 2, 4, 10, 51-Lucas and Luo (1993) 45. Anterior lamina-separated externally from 3, 7, 23, 24, 28, 42, 44-Wible et al. (1995) orbitosphenoid by alisphenoid (0), contacts 6, 21, 22, 39, 43-Rowe (1988) orbitosphenoid (1), or absent (2). 8, 34, 47-Lillegraven and Krusat (1991) 46. Small vascular foramen in anterior lamina- 9, 18, 19-Wible (1990) absent (0) or present (1). 11-Kermack et al. (1981) 47. Number of trigeminal nerve exits posterior to 16, 17, 27, 35, 36, 38, 40, 45, 48, 49, 50-Wible alisphenoid-single (0), multiple (1), or none and Hopson (1993) (2). 29-Luo (1994) 48. Ventral ascending groove or canal-open lat- 31, 37-Wible (1991) erally (0), intramural, within petrosal (1), in- 26, 33-Crompton and Sun (1985) 36 AMERICAN MUSEUM NOVITATES NO. 3183

APPENDIX 2 TAXON-CHARACTER MATRIX

Character* 1axm 5 10 15 20 25 30 35 40 45 50 %carpl. rTritbelentidae OOONO 0?000 00000 00000 00000 00000 00000 00000 0?000 00000 0 96.1 Tritylodontidae OOGNO 0?000 00000 10000 10110 00010 1OAOO 00102 0?000 01001 0 96.1 Adeloabsileus 10100 A??O1 00000 00000 00?00 ?1010 10011 0??0? ??000 01001 1 80.4 Smnocm2cdn 11110 00101 AOOOO 10000 00000 11010 10?11 00100 ??000 01000 1 92.2 Ilbrgazccximr 1110A 10101 10000 10000 00110 31010 10111 10000 10000 11001 1 100 DirnetheriLun 11100 1?111 ??000 1?000 00110 D1010 lOC11 10??? ???OA ?AOO? 0 68.6 Aqgazostc6- 11110 1?101 AOO?O 10000 ?0??? D1010 1?Cll ?0?0? ??000 0000? 1 68.6 11110 1?101 10110 11000 10110 21110 ?OC12 00010 10000 lOOOC 1 92.2 Nhltitubrcullata 11300 10111 COAOO 21A00 10011 32132 11212 21100 10001 01111 0 100 Clov. tricano&nts 11300 1?111 ??100 21000 10010 E1121 ?0212 1???2 ??1OA ?AB1? 1 70.6 Priaccdn 1130A 10111 20110 21001 COll? 21121 0?012 1?0?2 00?OA 1A211 1 84.3 Trioraccd2n 11301 1?111 C0110 21001 10110 11121 00212 110?2 00?OA 1A211 ? 86.3 Tricmcnoib 11301 1?1?1 ??110 ?1??? ?0??? 21??? ???1? ????? 33.3 Kbuoir petrosal 1 11300 10111 32101 21111 20000 21120 00212 111?1 0011? ??211 ? 90.2 Khoobur petrosal 2 11300 10111 31101 21?1? ?00?? 21120 0?212 1?1?1 00?OA 0A211 ? 76.5 Vicelestes 11200 21111 32101 21110 20111 31122 A0112 01111 ?0110 01111 0 98.0 Tchsyglossidae 01NO 20101 31102 00000 11110 31143 01112 01102 02101 00201 2 100 OrnidtorhW2chus 11300 20101 10100 01000 21110 32122 00112 01100 02001 00000 2 100 Marscipialia 11200 31111 32101 21110 20A11 31243 OOC03 01112 01102 02422 0 100 Placetali'a- 11200 31111 32101 21110 2OAll 00143 00103 01112 0C102 02302 A 100 *9 = missing data; N = not applicable; 0-4 = character states (see appendix 1); A = 0 & 1; B = 0 & 2; C = I & 2; D = 2 & 3; E = 1 & 2 & 3; %compl. = % of scoreable characters.

APPENDIX 3 6(3)*-cochlear duct coiled 40(2)*--complete wall separating cavum supra- DIAGNOSES OF NODES cochleare from cavum epiptericum The tree diagnosed here is the strict consensus 45(2)*-anterior lamina absent tree of the three equally most parsimonious trees 48(3)*-ventral ascending groove endocranial obtained using the PAUP reweighting procedure 49(0)*-ventral ascending groove vertical (rescaled consistency index) on the matrix in appendix 2 (fig. 6B). Nodes are diagnosed by listing Node 2. Prototribosphenida the derived condition for characters in appendix 3(2)-cochlear housing triangular, wider posteri- 1. Multistate characters or reversals are depicted orly without steep lateral wall within parentheses following the character num- 7-secondary spiral lamina ber. Asterisks (*) denote equivocal characters. 25-epitympanic recess on petrosal Characters are optimized under an accelerated 33(l)-broad petrosal bridge flooring facial nerve transformation (ACCTRAN) assumption. branches Node 1. Theria 6(2)*-cochlear duct bent 23*-crista parotica/squamosal contact absent 29(4), 30(3)-lateral flange absent 26(3)*-cranial aperture of prootic canal at an- 34(0)-lateral trough absent terodorsal margin of subarcuate fossa 35(3)-cavum epiptericum floored primarily by 36(0)*-tensor tympani fossa indistinct alisphenoid 48(1)*-ventral ascending canal intramural with- 42(1)-sulcus for inferior petrosal sinus between in petrosal petrosal and basioccipital, anteroposteriorly directed Node 3. 47(2)-no foramina for trigeminal nerve posterior to alisphenoid 30(0)-lateral flange narrowly separated from 50(2)-dorsal ascending groove endocranial crista parotica 1996 ROUGIER ET AL.: BASICRANIUM OF PRIACODON FRUITAENSIS 37

20*-deep pocket medial to paroccipital process 36(1)*-deeply recessed tensor tympani fossa on 24(0)*-fossa incudis on petrosal absent lateral trough anterior to hiatus Fallopii Node 4. Node 9. Monotremata 12(2)-lateral opening of cochlear aqueduct with- 6(2)-cochlear duct bent in inner ear 22-tympanohyal contacts promontorium 15(l)-crista interfenestralis vertical, delimiting 42(2)-sulcus for inferior petrosal sinus endo- back of promontorium cranial 18-caudal tympanic process of petrosal 51(2)*-hypoglossal and jugular foramina conflu- 19-post-promontorial tympanic recess ent 21(2)-medially directed tympanohyal attached to petrosal Node 10. Mammalia 11(3)*--cochlear aqueduct 3(3)-cochlear housing shape fingerlike in outline 29(2)-lateral flange ventrally directed Node 5. Triconodontidae 30(2)-lateral flange continuous with crista par- 5-small foramen on cochlear housing antero- otica medial to fenestra vestibuli 37-quadrate ramus of alisphenoid small or ab- 14-jugular foramen smaller than perilymphatic sent foramen/fenestra cochleae 4(0)*-ventral crest on cochlear housing absent 20-deep pocket medial to paroccipital process 16(0)*-stapedius fossa absent 23*--crista parotica/squamosal contact absent 38*-internal acoustic meatus with distinct medial rim Node 6. triconodonts 41(0)*-sulcus for sigmoid sinus extending to jugular foramen absent 30(1)-lateral flange widely separated from crista 45(1)*-anterior lamina contacting orbitosphen- parotica oid 38(0)*-internal acoustic meatus without distinct 48(2)*-ventral ascending groove endocranial medial rim 51(0)*-double or multiple hypoglossal foramina 40(2)*-complete wall separating cavum supracochleare from cavum epiptericum Node 1 1. 46*-small vascular foramen in anterior lamina 51(1)*-single hypoglossal foramen 13-jugular fossa absent 17-paroccipital process extending ventral to Node 7. Theriimorpha cochlear housing 28(l)-pterygoparoccipital foramen anterior to 43-craniomandibular joint anterior to fenestra fenestra vestibuli vestibuli 35(2)-cavum epiptericum floored primarily by 26(2)*-cranial aperture of prootic canal between petrosal cavum epiptericum and anterodorsal margin of 31(0)*-vascular foramen in lateral flange absent subarcuate fossa 39*-pila antotica absent Node 12. 40(1)*-fenestra semilunaris 45(0)*-anterior lamina separated from orbitos- 47(0)-single trigeminal foramen posterior to al- phenoid by alisphenoid isphenoid 21(1*)-posteromedially directed tympanohyal Node 8. Theriiformes attached to petrosal 9-fenestra vestibuli ellipsoidal Node 13. Morganucodontidae 33(2)-narrow petrosal bridge flooring primary 36(1)-deeply recessed tensor tympani fossa on facial foramen lateral trough anterior to hiatus Fallopii 47(1)-multiple foramina for trigeminal nerve 4(0)*-ventral crest on cochlear housing absent posterior to alisphenoid 46*-small vascular foramen in anterior lamina 49(1)-ventral ascending canal horizontal 11(2)*-channel for perilymphatic duct partially connecting to cavum epiptericum enclosed 16(2)*-stapedius fossa lateral to crista interfe- Node 14. Mammaliaformes nestralis 6(1)-cochlear duct elongated and straight 23(0)*-crista parotica/squamosal contact absent 24-quadrate articulation on petrosal 38 AMERICAN MUSEUM NOVITATES NO. 3183

23*-crista parotica/squamosal contact absent 3(1)-cochlear housing triangular in outline, wid- 26(3)*--cranial aperture of prootic canal at an- er posteriorly with steep lateral wall terodorsal margin of subarcuate fossa l-fenestra vestibuli margin without thickened osseous ring Node 15. 27(1)-tympanic aperture of prootic canal separate from pterygoparoccipital foramen 2-para/basisphenoid wing absent 34-lateral trough on petrosal 4*-ventral crest on cochlear housing absent 35(l)-cavum epiptericum partially floored by 11(1)*-open sulcus for perilymphatic duct petrosal 16(1)*-stapedius fossa on crista interfenestralis 51(l)-single hypoglossal foramen 33(1)*-broad petrosal bridge flooring facial 8* semicircular canals enclosed by petrosal only nerve branches 26(1)*-cranial aperture of prootic canal in rear of cavum epiptericum Node 16. 41* sulcus for sigmoid sinus extending to jug- 1-promontorium ular foramen

Recent issues of the Novitates may be purchased from the Museum. Lists of back issues of the Novitates, Bulletin, and Anthropological Papers published during the last five years are available free of charge. Address orders to: American Museum of Natural History Library, Department D, Central Park West at 79th St., New York, N.Y 10024. TEL: (212) 769-5545. FAX: (212) 769- 5009. E-MAIL: [email protected]

0 This paper meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).