Institutional Abbreviations

AMNH: American Museum of Natural History, New York, USA; MPEF: Museo Paleontológico Egidio Feruglio, Trelew, Argentina.

Familiar Definition

Henosferidae is the clade including the most recent common ancestor of Henosferus and Asfaltomylos and all its descendants.

Etymology

From the Greek henos (old) and from the Latin ferus (animal).

Holotype and Only Known Species

Henosferus molus.

Diagnosis

This new taxon is diagnosed by the combination of the following characters (autapomorphic traits are indicated with an asterisk): lower dental formula consisting of i4, c1, p5, m3; variously developed diastemata between most premolars; molars with an obtuse to right-angled trigonid and a basined talonid with two well-developed cusps and a ridgelike structure in the position of entoconid/entocristid; procumbent paraconid in a more labial position than the metaconid; talonid slightly wider and much lower than trigonid; talonid wider than long; blunt prominent hypoconid cusp not fully differentiated from broad, bulbous hypoconulid connected by a broad, low crest (hypocristid); talonid lingually closed by a strong rounded entocristid, well-developed lingually*; lack of talonid wear; slender lower jaw having a Meckelian groove, a prominent medial flange associated with a lateral ridge of the dentary, and a deep dentary trough (possibly indicating presence, even though reduced, of postdentary bones still attached to dentary); and prominent, transversely wide, spoonlike angular process with a medial crest occupying a position homologous to pterygoid crest of other mammals*.

Etymology

From the Latin mola, meaning millstone in reference to the well-developed talonid of the lower molars.

Diagnosis

As for the genus, for monotypic attribution.

Holotype

MPEF 2353: Right lower jaw with the dentary well preserved, bearing the distal half of p1 and p2, and almost complete m1 (fig. 2). The posterior portion of the dentary shows a deep trough and a medial flange; it also bears a deep and small coronoid facet, a remarkable Meckelian groove, and a transversely wide angular process. This specimen is chosen as the holotype because it preserves an almost complete molar that bears numerous diagnostic features, making it possible to compare with most other Mesozoic mammals.

Figure 2

Henosferus molus holotype MEFP 2353. Stereophotograph of the right lower jaw in lingual view and accompanying line drawing. Gray pattern indicates broken bone and matrix. Scale bar is 5 mm. Abbreviations: an  =  angular process; con  =  condyle; cop  =  coronoid process; cor  =  scar for the paradentary coronoid bone; dt  =  dentary trough; i1r  =  root of the lower first incisor; m1  =  lower first molar; mck  =  Meckelian groove; mf  =  mandibular foramen; mfl  =  medial flange; p1  =  lower first premolar; p2  =  lower second premolar; sym  =  mandibular symphysis. The p1 and p2 are broken, preserving only the distal halves of their crowns; the premolars appear thus to be like single-rooted but they are in fact double-rooted.

Hypodigm

MPEF 2354: Left lower jaw bearing the roots of the first and second incisors, the broken crown of the third and fourth incisors, a complete canine, five premolars, and three damaged molars (fig. 3). Behind the level of the coronoid process, the dentary is partially broken and preserved mostly as a natural cast of the medial aspect. MPEF 2357: Left lower jaw exposed in labial view preserving the canine, four premolars, and a molar with the trigonid mostly damaged (fig. 4). Floating in the matrix near the jaw are two teeth here identified as p1 and m2 (not shown in figures). The p1 is complete but the m2 is missing the protoconid and a small portion of the talonid. The posterior part of the dentary is complete, adding information on the lateral view of the jaw not accessible in the other specimens.

Figure 3

Henosferus molus referred specimen MEFP 2354. Stereophotograph of the left lower jaw in labial view and accompanying line drawing. Gray pattern indicates broken bone and matrix. Scale bar is 5 mm. Abbreviations: an  =  angular process; c  =  lower canine; cop  =  coronoid process; menf  =  mental foramina; i1 =  lower first incisor; i3  =  lower third incisor; i4  =  lower fourth incisor; p1  =  lower first premolar; p2  =  lower second premolar; p3  =  lower third premolar; p4  =  lower fourth premolar; p5  =  lower fifth premolar; m1  =  lower first molar; m2  =  lower second molar; m3  =  lower third molar.

Figure 4

Henosferus molus referred specimen MEFP 2357. Stereophotograph of the left lower jaw in labial view and accompanying line drawing. The isolated premolar in the stereophotograph corresponds to p1 of the same specimen. Gray pattern indicates broken bone and matrix. Scale bar is 5 mm. Abbreviations: an  =  angular process; c  =  lower canine; con  =  condyle; cop  =  coronoid process; p2  =  lower second premolar; p3  =  lower third premolar; p4  =  lower fourth premolar; p5  =  lower fifth premolar; m1  =  lower first molar.

Tentatively Referred Specimen

MPEF 2355: isolated upper premolar enclosed in a small block, which also includes indeterminate fragments of bone (not figured).

Locality and Horizon

All specimens come from the Queso Rallado locality (43°24′33.55″S/69°13′50.1″W), about 3 mi west-northwest of the village of Cerro Condor, Chubut Province, Argentina (fig. 1); Puesto Almada Member (Silva Nieto et al., 2003), Cañadón Asfalto Formation, Callovian-Oxfordian (Stipanicic et al., 1968; Tasch and Volkheimer, 1970).

Lower Jaw

The holotype and the referred specimens have a well-preserved dentary morphology, providing several features relating to an understanding of the evolution of the mammalian middle ear (see Discussion). The default specimen used to describe the jaw is the holotype (figs. 2, 5); discordant or accessory information provided by the hypodigm is noted.

Figure 5

Henosferus molus holotype MEFP 2353. Stereophotograph of the posterior portion of the right lower jaw in lingual view and accompanying line drawing. Scale bar represents 5 mm. Abbreviations: an  =  angular process; anc  =  concave surface of the angular process; con  =  condyle; cop  =  coronoid process; cor  =  scar for the paradentary coronoid bone; cr  =  medial crest of the angular process; dt  =  dentary trough; mck  =  Meckelian groove; mf  =  mandibular foramen; mfl  =  medial flange; s  =  step may be homologous to the “diagonal ridge” of Morganucodon (Kermack et al., 1973).

The horizontal ramus of the dentary is low and elongated with the alveolar and ventral edges almost parallel. In cross section, the labial surface of the dentary is convex while the lingual is almost flat (in specimen MPEF 2357, the labial surface is somewhat concave, especially below the molar area, but this feature appears to be related to deformation). The minimum depth of the lower jaw is located at the level of the posterior diastema of the canine, clearly shown in MPEF 2354. The incisor alveolar border is facing anterodorsally, intersecting obliquely the ventral edge in the most anterior point of the dentary.

In labial view of MPEF 2354, the wall of the i1 alveolus is partially broken, exposing the root of this incisor. An interdental process between the incisors is clearly observable. Both the holotype and MPEF 2357 show four somewhat damaged incisor alveoli. The alveolar borders of the cheek teeth of the holotype and MPEF 2354 are well preserved. The p1, p3, and p4 have sharp interadicular processes (still covered by rock in the p2) that are taller than the interdental processes. In all of the specimens of Henosferus, the labial alveolar edge is set below the level of the lingual alveolar edge. There are four mental foramina in MPEF 2354 and 2357, with little variation in their positions (figs. 3, 4). The anteriormost foramen is located below the i3–i4 diastema, near the ventral edge; the second one is located below the mesial border of the canine or slightly in front of it; the third one is found under the diastema between the canine and the p1. All three foramina are small, with some variation in their relative sizes. The fourth, located below the anterior root of the p2, is the largest. All of the anterior mental foramina face anteriorly, but the fourth foramen seems to face laterodorsally; we believe its ventral edge, which is formed by a thick ventral rim, to be natural, but it could also be caused by postmortem deformation. The labial view of the posterior portion of the jaw is available only in MPEF 2357. The coronoid process is well developed, twice as high as the horizontal ramus. The anterior border is straight, with a thick coronoid crest. The coronoid process, as determined by its anterior edge, is inclined posteriorly 115° with respect to a horizontal alveolar edge. The dorsal border was probably straight (as suggested by the medial view in the holotype), although slight damage in MPEF 2357 gives it the appearance of being dorsally concave. The posterior border of the coronoid process parallels the anterior border to the level of the pedicle of the condyle. The most conspicuous feature in labial view is the development of the lateral ridge of the dentary (fig. 4); this ridge extends horizontally from the level of the condyle to the anterior base of the coronoid process, forming the ventral edge of a distinct masseteric fossa (although the masseter muscle probably extended beyond this crest into the flatter area ventral to the ridge). The lateral ridge is a blunt and relatively robust crest, equally developed along the entire preserved portion. A small section of the crest is missing immediately in front of the condyle; however, we believe it is clear that the lateral ridge was confluent with the pedicle of the condyle, as seen, for example, in Morganucodon (Kermack et al., 1973), and the lack of connection is artificial. The anterior extension of the ridge is distorted by a low, round depression, probably an artifact. The crest seems to become weaker anteriorly and probably reached the anterior margin of the masseteric fossa, showing once more a condition not dissimilar to that of Morganucodon (Kermack et al., 1973, 1981; Crompton and Luo, 1993) and other basal mammaliaforms (Crompton, 1974; Jenkins et al., 1983; Gow, 1986; Lillegraven and Krusat, 1991). The bulk of the masseteric fossa is dorsal to the lateral ridge (i.e., it is restricted to the dorsal area of the coronoid process, about the level of the alveolar margin) and could correspond with the attachment of the portion of the masseter mucle identified by Turnbull (1970) as M. zigomaticomandibularis. The dorsal portion of the masseteric fossa is relatively deep, roughly rectangular in shape, with shallow horizontal scars for muscular attachment. This fossa, probably associated with portions of the temporalis muscle group (Turnbull, 1970), is deeper in the area immediately in contact with the lateral ridge and does not extend ventrally to it. The surface below the lateral ridge, ventrally bordered by the massseteric crest (see below), which becomes more distinct posteroventrally, probably served for the attachment of a small portion of the masseter muscle. No masseteric foramen is recognized in Henosferus.

The general features of the angular process are described in detail under the medial view (figs. 2, 5); particular features of the labial aspect include the masseteric crest and the concave surface between the angular process and the bottom of the condylar pedicle. The masseteric crest is developed on the ventral border of the angular process; it is blunt but becomes more distinct posteriorly and protrudes laterally more pronouncedly. Coupled with the larger development of the masseteric crest is the deepening of the portion of the dentary immediately below the lateral ridge; these combined factors result in a progressively more concave surface for the dentary in the vicinity of the angular notch. The development of a concave surface between the angular process and the lateral ridge is also found in Morganucodon (Kermack et al., 1973), Megazostrodon (Gow, 1986), Haldanodon (Krusat, 1980; Lillegraven and Krusat, 1991), Castorocauda (Ji et al., 2006), and most other mammaliaforms with a pronounced angular notch and lateral ridge (Allin, 1975; Clack and Allin, 2004).

The medial view of the dentary is well preserved in the holotype (figs. 2, 5). The mandibular symphysis is unfused, smooth, and roughly oval, anteroposteriorly elongated and extending posteriorly to the level of p1. Posterior to the symphysis, the ventral edge of the dentary is fairly straight back to the level of m2, where it curves dorsally, resulting in a concave outline that extends to the angular process.

Subtle scars near the dorsal border of the extensive coronoid process indicate the attachment area of deep temporal muscle fibers. The condyle is eroded in the holotype and the shape is not clearly defined, but specimen MPEF 2357 suggests that the condyle is mostly spherical. The articular surface of the condyle is incompletely preserved and faces posterodorsally, lacking a well-defined constriction (neck) at the base of the condyle. The main axis of the condylar process in the holotype forms an angle of approximately 160° with the alveolar edge and results in a condyle located above the level of the tooth row. The holotype has suffered a fracture that artificially masks this feature.

The angular process is prominent, protrudes posteroventrally, and it is located below the level of the tooth row. The holotype has sustained a sharp break just in front of the anterior root of the m1, resulting in a lower and forward position for the anterior end of the jaw and thus giving the exaggerated impression that the angle is far below the tooth row. The angular process is located posteriorly, relatively close to the condyle (fig. 5), and therefore the angular notch is proportionately narrow and small compared to Morganucodon. The anteriormost extent of the notch is at the level of the posterior border of the coronoid process. The angular process is transversely wide, more expanded lingually than labially; its medial border is determined by a round crest that decreases in size anteriorly. This crest, medial to the angular process, runs along the ventromedial edge of the dentary and occupies a position homologous to the pterygoid crest of other mammals. This crest, however, is well developed only in the vicinity of the angular notch, helping to delimit the ventral edge of the dentary trough, and does not continue anteriorly much further. The posterodorsal surface of the angular process is concave, almost spoonlike, and with a rounded posterior edge. A similar morphology was found in basal mammaliaforms for accommodating the reflected lamina of the angular (ectotympanic) (Crompton and Luo, 1993). The posterior orientation of the concavity of Henosferus resembles more that of docodontans (Lillegraven and Krusat, 1991; Ji et al., 2006) than that of Morganucodon and Sinoconodon, which is medially exposed (Kermack et al., 1973; Crompton and Luo, 1993). Both the medial crest of the angular process and the medial flange of the dentary project lingually to the same degree.

There is a relatively small, concave surface for the coronoid bone (not preserved), located at the level of the alveolar edge, above the mandibular foramen (fig. 5). The anterior and dorsal edges of this facet are round and clearly defined, while the posterior and posteroventral borders are not clearly delineated. The coronoid facet faces posteromedially, having its deepest point adjacent to the anterodorsal border.

The mandibular foramen is relatively distinct and located posteroventral to the coronoid facet (figs. 2, 5). This elliptical foramen faces posteriorly and is continuous with the dentary trough. A sharp crest forms the dorsal edge of the foramen and decreases in size posteriorly before becoming continuous with the protruding medial flange. The dorsal rim of the mandibular foramen is very close to the medial flange of the dentary in Henosferus (fig. 5). In contrast, in other basal mammaliaformes such as Morganucodon, these two structures seem to be separated by a wider gap (Kermack et al., 1973; Crompton and Luo, 1993). The gap between the dorsal rim of the mandibular foramen and the medial flange would permit the passage of the inferior alveolar nerve and artery from the infratemporal fossa into the dentary.

The medial flange, medial trough, and Meckelian groove are the most remarkable features in the medial view of the lower jaw of Henosferus (fig. 5). The medial flange is sharp and prominent, forming a protruding shelf extending posteriorly from the mandibular foramen and continuing to the ventral surface of the condyle. The flange forms an angle of approximately 170° with the alveolar border. The medial-most projection of the medial flange occurs at the level of the angular process, and as mentioned above, has an extension similar to that of the medial crest of the angular process. There is limited evidence of articulation of the medial flange with other elements, as most of its surface shows parallel lineation of the periosteum similar to that in other portions of the dentary, but along its mid-length there is a smooth, elongated surface that could be interpreted as a facet, probably indicating the presence of postdentary bones (see Discussion).

The dentary trough is deep and develops below the medial flange and posterior to the mandibular foramen (fig. 5). In the dentary trough, two different surfaces can be observed separated by a blunt and not very distinct ridge of bone likely homologous to the diagonal ridge of Morganucodon (Kermack et al., 1973). Accepting the homology of this feature between Morganucodon and Henosferus, we use the same terminology, although the morphology is somewhat different. The surface dorsal to the diagonal ridge is placed immediately posterior to the mandibular foramen and below the notch of the medial flange, interpreted here as allowing transmission of the inferior alveolar nerve and accompanying vessels. We, therefore, believe that this surface dorsal to the diagonal ridge is occupied mostly by the vascular and nervous structures just mentioned. The second surface, below the diagonal ridge, is mainly related to the Meckelian groove and passes below the mandibular foramen to the posterior portion of the horizontal ramus of the dentary. The Meckelian groove is broader and deeper porteriorly between the medial flange and the angular process.

Behind the mandibular foramen, the dentary trough connects the Meckelian groove with the concave posterodorsal surface of the angular process (fig. 5). The trough becomes wider posteriorly, its direction suggested by a different bony texture indicating that the groove is directed dorsally toward the ventral surface of the medial flange. The dentary trough is delimited ventrally by the forward continuation of the pterygoid crest. In the area of the angular notch, the dentary is mediolaterally expanded, forming a sharply concave surface. This surface is wider along the dorsal edge of the angular process, resulting in a well-delimited space between the angular process, the edge of the angular notch, and the ventral surface of the medial crest, including its continuation as the condylar process.

The surface ventral to the diagonal ridge narrows strongly in front of the mandibular foramen and continues anteriorly as a continuous Meckelian sulcus, which reaches the posterodorsal surface of the symphysis without losing much width along the way, although it becomes progressively shallower (figs. 2, 5). The Meckelian sulcus curves ventrally from the area around the mandibular foramen to reach the ventral margin of the jaw, at the level of the posterior premolars, and then curves upward to reach the symphysis.

Dentition

For descriptive purposes, we followed the dental nomenclature of Crompton (1971), Crompton and Kielan-Jaworowska (1978), and Bown and Kraus (1979) (fig. 6A); however, an alternative nomenclatural hypothesis is also suggested (fig. 6B; see further explanation in the Discussion).

Figure 6

Schematic drawing of the m1 of Henosferus molus (MEFP 2353) in occlusal view depicting the tribosphenic nomenclature (A) and an alternative interpretation of the australophenidan molar homologies (B). Abbreviations: acin  =  anterior cingulid; cin  =  cingulid; co  =  cristid obliqua; cr  =  crest; d  =  cusp d; enc  =  entocristid; end  =  entoconid; f  =  cingular cuspule f; hf  =  hypoflexid; hyd  =  hypoconid; hyld  =  hypoconulid; lc  =  lingual cingulid; med  =  metaconid; pad  =  paraconid; pcin  =  posterior cingulid; prd  =  protoconid.

With the exception of the m1 (fig. 7), the description of the dentition is based on the referred specimen (MPEF 2354; fig. 3). The lower dental formula of Henosferus is interpreted as i4, c1, p5, m3; tooth morphology has been the sole deciding criterion for distinguishing premolars from molars, because tooth replacement evidence is lacking. The incisors are placed close to one another and are separated by small diastemata that increase slightly in size among the posterior elements. The i1 and i2 are only represented by their roots; these teeth are circular in cross section and strongly procumbent. More posterior incisors are less procumbent, with a mesiodorsal orientation of the crowns. The alveolus of i1 is parallel to the ventral edge of the dentary, as clearly shown by the exposed root of i1 in the holotype, reaching back at least to the middle of the symphysis, but likely extending back to the level of the canine. According to the size of the alveoli and the cross section of the teeth, all incisors seem to have been of similar size, although the i1 could be slightly larger. In the holotype, all incisors are broken off (fig. 2), but some of the alveoli are clearly visible and agree with the morphology of MPEF 2354. The tip of the crowns of the i3 and i4 of MPEF 2354 are damaged, preserving only their subcircular bases, where no individual cusps can be identified.

Figure 7

Henosferus molus holotype MEFP 2353. Stereophotograph of the m1 in lingual and occlusal views, with accompanying drawings. Scale bar represents 1 mm.

The canine is complete in MPEF 2354 and MPEF 2357 and represented only by the alveolus in the holotype. It is the tallest tooth of the entire dentition, single rooted, and approximately cylindrical with a blunt apex that develops abruptly. The canine is posterodorsally projected in both specimens preserving it, although its inclination is more pronounced in MPEF 2354 than in MPEF 2357. We believe the canine of MPEF 2357 approaches the normal condition (fig. 4), while that in MPEF 2354 is a preservational artifact. The mesial edge is rounded and convex near the tip of the crown, while the distal edge is straight and slightly sharper than the mesial one. The canine is supported by a robust single root that makes a shallow bulge on the labial surface of the dentary. The canine is separated from the incisors by a smaller diastema than that separating the canine from the premolars (but see below). The canine is absent in the holotype, and although they are difficult to identify, it seems that the alveolus and diastemata are smaller than in MPEF 2354.

We interpret that Henosferus has five premolars (fig. 3); however, in the holotype, in addition to the five premolars, there are unambiguous indications of an extra alveolus partially plugged by bone between the canine and the alveolus of the mesial root of p1. Reabsorption of alveoli occurs widely among Mesozoic mammals, when a tooth is shed and not replaced or when the replacement tooth erupts behind the tooth being replaced (Luckett, 1993), so that both generations of a single tooth family are present at the same time in an individual. We believe this alveolus represents a somewhat more persistent deciduous p1, which was shed and replaced by a p1 occupying a more distal position in the jaw; plugging of alveoli is common among basal mammaliaforms (Crompton and Luo, 1993; Rougier et al., 2001). The dentaries MPEF 2354 and MPEF 2357 do not show unambiguous evidence of alveoli in the corresponding portion of the lower jaw. The premolars occupy about 45% of the total length of the horizontal ramus of the dentary. The premolars are separated from each other by proportionately large and subequal diastemata, resulting in a very open premolar series with teeth positioned at regular intervals (figs. 3, 4). The premolars are transversely narrow, almost four times longer than wide. All premolars have two cylindrical roots subequal in size. The main cusp is centrally located in the crown with the roots supporting the anterior and posterior halves of the teeth respectively in the p4–p5. In the anterior premolars (p1–p3), the main cusp is located mesially (i.e., over the anterior root) or close to the midline. The roots diverge slightly ventrally and are continuous with the crown, without a distinct neck. The limit between the roots and crown is determined by the presence of a distinct enamel–dentine juncture.

The crowns of the p1 and p2 are very similar; both teeth are dominated by a prominent main cusp (protoconid) that determines a trenchant, triangular outline for the teeth. The p1 and p2 lack accessory cusps and cingula, at least on their labial surface (fig. 3). The tip is acute, located slightly mesially on the crown, resulting in a nearly symmetrical tooth, with the mesial edge slightly convex and the distal edge slightly concave. The p1 seems to be slightly more robust than the p2.

The p3 is taller than the preceding premolars and slightly shorter than the p4. The apex of the main cusp is mesially displaced; the anterior edge is fairly straight and the distal one is slightly concave, resulting in a more asymmetrical tooth than the preceding premolars. This feature is accentuated by the presence of a small posterior accessory (cingular) cuspule at the base of the crown (fig. 3).

The p4 is slightly taller than the p3 and almost symmetrical in lateral view. The tip of p4 is acute, centrally placed in the crown, and both mesial and distal edges are fairly straight (fig. 3). There is a distinct posterior accessory cuspule that has its tip truncated by wear. The posterior accessory cuspule of the p4 is more prominent, closer to the alveolar margin, and more lingually located than that of the p3. The p4 has an incipient mesiolingual cingulid at the base of the crown, which is higher than the posterior accessory cusp, although this condition might be exaggerated by the distal inclination of the tooth.

The p5 seems to be subequal in height (once the missing tip is considered) but longer than the p4 (fig. 3). The jaw MPEF 2354, however, is fractured through the p5, the main cusp being almost bisected by this crack, which is slightly open, increasing the distance between the two fragments of p5 and giving the impression that the tooth is much longer than it actually is. This tooth is better preserved on the MPEF 2357 (fig. 4). On the lateral aspect, the crown is almost symmetrical, bearing anterior and posterior cuspules located at the same level near the base of the crown. The mesial border of the protoconid is straighter than the distal border, which is slightly concave. The lateral surface of the protoconid is slightly convex, while the medial surface is flat. The anterior accessory cusp is tiny, placed slightly lingually at the base of the crown and separated from the main cusp by a transverse groove. The posterior accessory cusp is located at the distolingual corner of the main cusp, and a small flake of enamel has been displaced anteriorly. The posterior half of the tooth is broader buccolingually than the anterior half, especially in the immediate vicinity of the posterior accessory cusp, where in specimen MPEF 2357 there is abundant wear. No basined talonid is present on this premolar.

In the holotype, the alveoli for p3–p5 are discernible; these alveoli differ from MPEF 2354 in the size and number of the diastemata. Instead of the regular, subequal diastemata of the referred specimen, the holotype lacks diastemata between the p1–p2 and p4–p5; furthermore, only very small diastemata are present between p2–p3 and p3–p4. This difference in diastema size cannot be explained as an age difference between the two specimens, because we would expect to have larger diastemata in older individuals. However, based on relative wear of the molars, the holotype is older than MPEF 2354. The specimen MPEF 2357 agrees more closely with MPEF 2354 than with the holotype, although there is a wide diastema between p3 and p4. We interpret these differences as reflecting individual variations and preservational differences.

There are three double-rooted molars implanted close to one another (i.e., there are no intermolar diastemata) in Henosferus. The molars decrease in size posteriorly, the last one being considerably smaller than the others. The last molar is located at the base of the anterior edge of the coronoid process, without leaving any substantial space between the tooth and the coronoid process.

The holotype preserves a fairly complete m1, providing good information on talonid morphology (fig. 7). The molars of MPEF 2354 and MPEF 2357 are damaged, making their interpretation challenging. The m1of the holotype shows a pattern with a trigonid and a fully basined talonid to which tribosphenic terminology can be readily applied. The trigonid is open with an angle of approximately 110° (fig. 7). The talonid is slightly wider and much lower than the trigonid. The labial surface of the trigonid is somewhat worn, even though a small cuspule (f) is present on the mesiolabial edge at the base of the trigonid. This cuspule seems to be a mesial elaboration of an anterior cingulid and, at least as preserved, lacks a conspicuous apex. The cingulid is also poorly individualized, and it is mostly expressed as a thickening of the crown base immediately above the root.

In the holotype, the tips of the three main cusps of the trigonid are truncated by wear, resulting in a single amalgamated wear surface that is horizontally oriented; in addition, the metaconid is also worn on its disto-labial surface (fig. 7). The protoconid is the most prominent cusp, followed by the paraconid and then by a small metaconid. On the mesiolingual and lingual base of the trigonid, there is a slightly crenulated cingulid that ends just anterior to the mesial edge of the metaconid. There is some damage on the mesial surface of the paraconid, which partially obscures the area, but enough is preserved to show that the lingual cingulid extends mesially to form part of the interlocking system between the p5 and m1. As preserved, the m1 of the holotype does not show a distinct cusp e. The mesiolingual corner of the trigonid, formed by the sharp base of the paraconid and the mesial extension of the cingulid, extends further mesially than the small cusp f; therefore, the interlock between p5 and m1 was oblique, with a greater lingual than buccal overlap. The trigonid is broad but not basined (fig. 7); most of it is occupied by the prominent base of the protoconid and only a small surface is present between the slopes of the three main cusps and the lingual cingulid that marks the medial edge of the trigonid. The paraconid is slightly procumbent and certainly occupies a more labial position than the metaconid, which constitutes the most lingual feature on the molar.

A heavily worn distal metacristid occurs in the holotype; the relevant area of the tooth on MPEF 2354 is truncated by a broad wear facet. The m1 of MPEF 2357, however, has a well-preserved talonid and distal face of the metaconid showing unequivocal evidence of a distal metacristid. The crest is broad and descends from the metaconid apex distolabially toward the weak cristid obliqua and prominent hypoconid.

The talonid is fully basined, rectangular in shape, and wider than long (fig. 7). The hypoconid is located on the distolabial corner of the talonid. Much of the hypoconid has been worn off in the holotype, but in MPEF 2354 and MPEF 2357, it is a blunt prominent cusp that is not fully differentiated from the hypoconulid; both cusps are connected by a broad, low hypocristid (Martin and Rauhut, 2005) crest. The labial face of the hypoconid bulges strongly labially beyond the base of the supporting root. From the mesiolingual face of the hypoconid arises the cristid obliqua, which determines the labial border of the basin of the talonid; this crest meets the trigonid slightly labial to the base of the metaconid. A wide hypoflexid is defined between the labial edge of the cristid obliqua and the hypoconid. The deep hypoflexid narrows the talonid basin to only two-thirds of the total width of the talonid. As in other Mesozoic mammals, the area of the hypoflexid is oblique and not fully vertical as seen in many later therians; therefore, there is not a great difference in height between the basin of the talonid and the portions of the talonid immediately labial to the cristid obliqua.

In the holotype, the hypoconulid is broad and bulbous and is the most prominent cusp of the talonid. In posterior view, the hypoconulid is partially obscured by some remnant sediment, but in the rear slope of the talonid a distinct vertical grove separates the hypoconid and hypoconulid. The hypoconulid would project posteriorly to participate in the interlocking mechanism as a broad crest descending posteriorly from the apex of the cusp and anchoring the posterior portion of the tooth between cusp f and the mesial extension of the paraconid on the succeeding tooth. The entoconid is hard to recognize due to wear; however, the talonid is lingually closed by a strong rounded entocristid. This crest is well developed lingually. The entoconid would be very close to the hypoconulid; therefore, the postcristid would also be very short. The best evidence for the position of the entoconid is produced on the lingual surface and expressed as a rounded eminence thought to represent the base of the entoconid (fig. 7). The entocristid is separated from the metaconid by a small but deep entoflexid (talonid notch); the entocristid is not directed mesially toward the base of the metaconid, but instead extends lingually, forming the posterior edge of the talonid notch.

In the referred specimen MPEF 2354, the trigonid and talonid of the m1 are crushed, and the paraconid, the labial basal part of the protoconid, the distal face of the metaconid, and the distolingual border of the talonid are missing. Despite breakage, the height of the protoconid is probably complete, showing remnants of a mesiobuccal wear facet on its tip. In posterior view, the hypoconid and the hypoconulid are not separated by a vertical groove, as happens in the holotype. MPEF 2357 adds further variability to the hypoconid–hypoconulid relationship and size. The talonid of this latter specimen seems to be complete and well preserved; while the hypoconid is of identical position and relative size as the holotype and MPEF 2354, the hypoconulid is very small and apparently closely apressed against the entoconid, which is quite distinct. MPEF 2357 resembles Ausktribosphenos and Bishops in these features (Rich et al., 1999a, 2001a). We are unsure about the meaning of the differences in the observed variability of the talonid cusps and relationships.

The remaining molars (m2–m3) are preserved in the referred specimen MEPF 2354 (fig. 3), but not in the holotype, and are the sole elements on which this description is based; the m2 of MPEF 2357 awaits preparation. The m2 is the best preserved molar of MPEF 2354. Missing or broken are the mesiolingual border of the base of the paraconid, the tip of the protoconid, most of the metaconid, and the lingual rim of the talonid. The main cusps of the trigonid form an angle of 85°. This angle is approximately 25° more acute than that in the m1 of the holotype (the m1 is too damaged in MPEF 2354 to estimate an angle). The protoconid is the tallest preserved cusp and likely the tallest in the crown. The protoconid has a bulbous base and a somewhat flattened mesial face. The lingual slope is flat, almost vertical, and occupies most of the surface of the trigonid; this feature, added to the separation between the bases of the paraconid and metaconid and the proximity of the lingual cingulid, results in the almost complete absence of a trigonid basin.

The relative heights of the paraconid and metaconid are impossible to estimate because the metaconid is broken at its base. The paraconid seems to be a fairly conical cusp truncated by an oblique wear facet on the labial aspect. Although a fair amount of wear is present, there is no well-developed metacristid; that is, the crest is not a vertical structure aligned with the posterior base of the trigonid. The condition in Henosferus resembles that of more generalized forms in which wear removes a substantial amount of material from the facet of the cusp bases to elaborate continuous wear facets; nevertheless, the condition here does not seem to be as primitive as that seen among Zhangheotheriidae (Rougier et al., 2003b; Tsubamoto et al., 2004). The protocristid also seems to be produced by a mixture of innate morphology and wear. A thin lingual cingulid can be distinguished from the base of the paraconid, extending lingually to the base of the metaconid. The cuspule f of the m2 is not as individualized as that of the m1 of the holotype and extends posteroventrally to the level of the base of the protoconid. The cuspule f delimits labially a notch for the interlocking with the hypoconulid, which strongly projects posteriorly in the m1 preserved only in the holotype. Some damage to the mesial portion of the crown of the m2 of MPEF 2354 may exacerbate the depth of the notch for interlocking. The notch is completed lingually by the procumbent paraconid, which, as in the m1 of the holotype, projects mesially beyond the position of cusp f. The talonid is mesiodistally shorter and probably subequal to or wider than the trigonid. The hypoconid is unworn, but a thin flake of its posterior face has been broken off. The cristid obliqua is low and reaches the base of the metaconid just lingual to the metacristid notch. The hypoconulid is rounded, bulbous, and massive, missing most of the labial surface and separated from the hypoconid by a shallow groove. The entoconid and entocristid are missing in this tooth, resulting in a molar basin that is artificially open lingually.

The last molar only preserves part of the trigonid and a small piece of tooth placed posterior to the metaconid. The labial portions of the trigonid (protoconid) are missing. As in m2, the paraconid bears a wear facet on the labial face of the tip. The lingual cingulid along the margin of the trigonid is more pronounced than in m2 and m1 and extends only between the bases of the paraconid and metaconid without reaching the mesial surface of the paraconid. The cingulid bears two small, blunt cuspules interrupted by a groove, located at the level of the protoconid. These grooves that delimit the small cingular cusps are the lingual continuation of the groove that demarcates the bases of the three main cusps (protoconid, paraconid, and metaconid), converging on the surface of the trigonid basin. Grooves such as these are commonly seen in forms with broad, open molariforms, such as amphilestids and “symmetrodonts” (Kielan-Jaworowska and Dashzeveg, 1998; Cifelli and Gordon, 1999; Cifelli and Madsen, 1999; Kielan-Jaworowska et al., 2004).

All of the molars have two vertically implanted cylindrical or slightly oval roots. The anterior one supports the trigonid and the posterior one the talonid. The posterior roots may be somewhat more compressed buccolingually than the anterior and seem to be longer mesiodistally.

Wear Facets

For nomenclature of the wear facets of lower molars, we follow Crompton (1971) and later amendments by Crompton and Kielan-Jaworowska (1978). In the holotype specimen, the only molar preserved (m1) has the trigonid and talonid strongly worn (attrition and apical wear), making recognition of the wear facets difficult. The lower molars of Henosferus seem to have all major cusps present in tribosphenic mammals; however, the occlusal features do not appear to conform to that expected among members of Tribosphenida.

Facet 1, covering the metacristid and the posterior slope of the trigonid and extending into the hypoflexid, is completely preserved on the m2 of the referred specimen (MPEF 2354) and partialy on the m1 in the same specimen. Facet 1 is the most prominent of all the facets of Henosferus and results from the wear produced by a presumably robust paracone and preparacrista of the upper molar. Facet 2 extends from the protoconid to the paraconid along the paracristid on the anterior surface of the trigonid, and can be best observed on the mesiolabial face of the m2 of the referred specimen hypodigm and is complemented with partial views of the same from the m1 and m3 of the same specimen. This facet determines a relatively deep gully stretching from the labial surface of the paraconid and extending down to the anterior basal cingulid; along the anterior surface of the trigonid, facet 2 is flat and ornamented with striae directed inferiorly and labially. Facet 3 is developed on the mesiolabial face of the hypoconid and extends labially to merge with facet 1 at the deepest point of the hypoflexid. A deep trough is determined jointly by facets 1 and 3. Facet 3 is clearly seen on the m2 of the referred specimen and partially on the m1 of the same specimen and in the holotype. The holotype shows that facet 3 covers the mesial slope of the hypoconid and obliquely truncates the apex of this cusp when wear progresses. Facet 4 would be present on the distolabial face of the hypoconulid, but it cannot be clearly recognized in the holotype of Henosferus; the only possibility for the presence of this facet is on the m2 of MPEF 2354, in which the appropriate area displays a roughened surface that could either be a chipped piece of enamel or the facet. Based on the texture of other, unambiguously identified, facets, we lean toward the first of the two alternatives. Among tribosphenic mammals facet 5 develops on the distal face of the metaconid and extends down into the mesiolingual portion of the talonid, resulting from the likely shearing action of the mesial surface of the protocone along the preprotocrista. As interpreted here, facet 5 is likely not present in Henosferus. In the m2 of MPEF 2354, there could be a very small and limited area of wear on the base of the metaconid, lingual to the distal metacristid; this possible wear may only represent an artificial change in texture produced by the preparation of the specimen. Despite adequate preservation, neither the holotype nor the referred specimen MPEF 2357 show any evidence of a putative facet 5 in the relevant areas of the talonid. Facet 6, developed in mesiolabial surface of the entoconid in tribosphenic mammals (Crompton, 1971), is absent in Henosferus; evidence in favor of the absence of wear in this area of the talonid is furnished more definitively by MPEF 2357, where, despite extensive wear of the metaconid and distal metacristid, the mesial surface of the talonid cusps still preserve rounded bases and no wear.

A reconstruction of the jaw in Henosferus (fig. 8) is presented as a way to summarize the morphology in this early australosphenidan and to facilitate comparisons with other taxa.

Figure 8

Reconstruction of the right lower jaw of Henosferus molus in labial and lingual views.

Tooth Formula

The lower dental count of Henosferus molus is i4, c1, p5, m3; this interpretation agrees with the reformulation of dental homologies proposed for Asfaltomylos by Martin and Rauhut (2005; contra Rauhut et al., 2002). The presence of four lower incisors and five lower premolars is a trait commonly found in the earliest eutherians, such as Eomaia, Prokennalestes, and zhelestids (Kielan-Jaworowska, 1975; Kielan-Jaworowska and Dashzeveg, 1989; Nessov et al., 1998; Ji et al., 2002). Among nontribosphenic mammals, a high number of premolars (five or more) is known in Amphitherium (Simpson, 1928b) and Peramus (McKenna, 1975; Butler and Clemens, 2001; but see Clemens and Mills for a different interpretation). As far as the interpretation permits, the number of premolars in the Early Cretaceous genera of Australia is six in Bishops (Rich et al., 2001a) and five in Ausktribosphenos (Rich et al., 1997, 1999a). In these taxa, the transition between premolars and molars is gradual and, together with the lack of replacement evidence, complicates the interpretation of the count of the postcanine teeth. There is a clear morphological break between the premolar and molar series in Henosferus; this is a purely morphological (and somewhat arbitrary) determination, because we do not have any evidence bearing on tooth replacement. Among early eutherians, the break between permanent premolars and molars is not as evident as that observed in metatherians (Clemens and Lillegraven, 1986; Rougier et al., 1998). The likely reduced, and derived, number of three molars is also present in Ausktribosphenos, Bishops, and derived monotremes (Teinolophos has four or more molars; 151Rich et al., 2005) (Archer et al., 1985, 1992, 1993; Rich et al., 1997, 1999a, 2001a) and in the nontribosphenic cladotheres Peramus (McKenna, 1975; Butler and Clemens, 2001; Kielan-Jaworowska et al., 2004) and Vincelestes (Bonaparte, 1986a), and it is traditionally recognized as synapomorphic at the base of Eutheria. Three molars is one of the most clear plesiomorphic traits for many extant placentals (e.g., Rougier et al., 1998). Plotting tooth count in the cladogram results in the recognition of multiple events of molariform/premolariform gains and losses, suggesting also a probable plastic boundary between premolar and molars (i.e., a premolar position becomes a molar or vice versa). Most notable are the extremes, such as six premolars in Kuehneotherium and Bishops and up to nine molariforms occuring among some dryolestoids (Amblotherium, not included in the analysis but implied by the position of reference taxa). Tooth count is, therefore, a problematic character for phylogenetic reconstruction. The risk of nonhomologous comparison entailed by the pure count of teeth grouped by morphological discontinuities is enormous. Within a limited phylogenetic framework, a minimum of homology can be secured (e.g., among therians) and characters such as these are helpful. However, when comparing the m1 of a triconodont, multituberculate, monotreme, and placental, it is unclear to us that we are actually comparing homologous structures. Wide ranging statements of homology that can bridge widely disparate groups are tempting and help tidy up distinct portions of a cladogram defined by characters with highly localized distributions. Dental count is one such character, but until a better understanding of tooth formula evolution is reached, topologies based on, or supported by, tooth count should be regarded as provisional.

Tribosphenic Molars

The term tribosphenic was first used by Simpson (1936) to indicate the mortar and pestle opposing action of protocone and talonid and a wedgelike, alternating and shearing action of trigon and trigonid. The tribosphenic lower molar consists of an anterior triangle of cusps (the trigonid) and a posterior basin (the talonid) flanked by two or three cusps. The cusps of the trigonid are the labial protoconid, mesiolingual paraconid, and distolingual metaconid. The cusps of the talonid are the lingual entoconid, labial hypoconid, and distomedial hypoconulid (Osborn, 1907; Patterson, 1956; Bown and Kraus, 1979).

The traditional view of molar evolution of tribosphenic mammals (Crompton, 1971; Crompton and Kielan-Jaworowska, 1978) can be summarized as follows: Kuehneotherium and other basal mammaliaforms from the Late Triassic and Jurassic have the three main cusps of the trigonid of later therians forming an obtuse angle with a heel formed by the cusp d (the hypoconulid; Crompton and Jenkins, 1968; Kermack et al., 1968) and without a basin. In later forms of cladotherians from the Early Cretaceous, including Amphitherium, Arguimus, Palaeoxonodon, Arguitherium, and Vincelestes, the talonid heel becomes wider and larger than in basal “symmetrodonts” but a basined talonid and a hypoconid cusp are still absent (e.g., Dashzeveg, 1979, 1994; Freeman, 1976; Rougier, 1993). The talonid becomes larger in Peramus and Kielantherium from the Early Cretaceous; in those taxa a hypoconid and an incipient basin are present (Clemens and Mills, 1971; Dashzeveg, 1975; Butler and Clemens, 2001). Finally and completing the basic talonid morphology, an entoconid cusp is developed in Aegialodon (Kermack et al., 1965; Crompton, 1971), Tribactonodon (Sigogneau-Russell et al., 2001), Pappotherium (Butler, 1978), and therians (e.g., Cifelli, 1999; Kielan-Jaworowska et al., 2004).

The assumption of a double origin of the tribosphenic molar (Luo et al., 2001a, 2002; Sigogneau-Russell et al., 2001; Rauhut et al., 2002; Kielan-Jaworowska et al., 2004; Martin and Rauhut, 2005) accepts this series of homologous transformations (or most of them; see discussion of Kuehneotherium in Rougier et al., 1996a and Godefroit and Sigogneau-Russell, 1999), but argues that an independent tribosphenic molar was achieved among australosphenidans.

Simpson's (1936: 797) definition of the tribosphenic molar (“suggestive of the mortar and pestle, opposing action of protocone and talonid and the wedge-like, alternating and shearing action of trigon and trigonid”) implies that the components involved in the formation of the tribosphenic molar (protocone, entoconid, hypoconid, etc.) must be homologous, preserving a specific and particular hypothesis of homology. Not just any cusp occluding on a basin will make a tooth tribosphenic. The definition of tribosphenic employed by Luo et al. (2001a) is based only on general functional terms (“The tribosphenic lower molar is defined by a basin-like heel (talonid), which grinds (tribein) with the large inner cusp (protocone) on the upper molar– functionally analogous to mortar-to-pestle grinding. This is in addition to a wedge-like trigon (sphen) to shear with the crests of the corresponding upper tooth.” Luo et al., 2001a: 55). Defining a group or structure on a functional basis is a nonhierarchical procedure per se, which obviates ancestor–descendant relationships and phylogenetic internesting of characters. Despite similarity of function, there is no reason to, a priori, give the same name to two structures believed to have different phylogenetic origins; that is to say if the “tribosphenic” molar of asutralosphenidans and that of therians is not homologous, there is no compelling argument to apply to them the same term and, in particular, to be surprised by the independent acquisition of nonhomologous but morphologically indistinguishable functional complexes (Luo et al., 2001a, 2002; Kielan-Jaworowska et al., 2004). The independent acquisition is a corollary of the nonhomology, or lack of phylogenetic continuity, between these character complexes. Chow and Rich (1982) and Wang et al. (1998), when faced with a similar problem of a functional equivalent of the therian tribosphenic molar, opted to called the molar “pseudotribosphenic” and the protocone “pseudoprotocone”, a solution we believe to be a sensible one. In fact, a similar nomenclatorial approach has been followed by Martin and Rauhut (2005) for the australosphenidan Asfaltomylos. The relatively well-established term such as Tribosphenida (McKenna, 1975), referring to a character-defined taxon (i.e., a group of mammals diagnosed by having a protocone, occluding on the talonid) was changed to Boreosphenida, based on the fact that a nonhomologous, similar functional complex was present in the unrelated australosphenidans (Luo et al., 2001a, 2002, 2003; Kielan-Jaworowska et al., 2004). The naturalness of Tribosphenida, supported among other features by the synapomorphic acquisition of a prominent protocone and basined talonid, is not questioned by Luo and coauthors or by any recent study (e.g., Woodburne et al., 2003; Martin and Rauhut, 2005). Only tribosphenic mammals (i.e., members of Tribosphenida) have a protocone and a basined talonid that is homologous among members of the group. Tribosphenida is an unambiguous term that refers to a clearly monophyletic group. Australosphenidans do not have a tribosphenic molar, although they may possibly have a functional equivalent of it, a question explored below.

The functional equivalence of the austalosphenidan molar and the tribosphenic molar seems to be at present tenuously supported. The basal members of the australosphenidan radiation are known only by mandibular elements and lower dentitions; therefore, the presence of an upper protocone or functionally equivalent cusp must be deduced from the lower tooth morphology and wear facets. Martin and Rauhut (2005: 422) wrote: “The talonid wear pattern of Asfaltomylos differs fundamentally from that of Laurasian tribosphenic boreosphenidans, because it shows no wear within the talonid basin itself”; the same applies to Henosferus and Ausktribosphenidae (Hunter, 2004; G.W.R., personal observation, 2004). No wear facets are distinguishable in the talonid of the Australian australosphenids. In the Middle Jurassic (Ambondro) facets 5 and 6 have been identified (Flynn et al., 1999), but Martin and Rauhut (2005) have subsequently challenged this interpretation. We are uncertain about this feature in Ambondro; whatever the case, most of the wear is nonetheless labial to the cristid obliqua in Ambondro.

The information summarized above calls into question the presence of a single major cusp occluding in the basin, a sine qua non trait of tribosphenic molars. Furthermore, the only putative (see Woodburne, 2003 and Woodburne et al., 2003) australosphenidan group known by upper teeth, Monotremata, is almost universally interpreted as missing a protocone cusp, either truly homologous or functionally analogous (Greene, 1936; Luckett and Zeller, 1989; Pascual et al., 1992a,b, 2002; Woodburne, 2003), with teeth functioning in a very different way from those of tribosphenic mammals (Kielan-Jaworowska et al., 1987; Archer et al., 1992; Pascual et al., 1992a,b, 2002). Luo et al. (2002: 26) identified a small cuspule in Monotrematum as the sole remnant of a functional protocone that later would be lost in more derived members of Monotremata. Pascual et al. (2002) argued against such identification. We concur with Pascual et al. (2002) in considering such homology unlikely. There is, then, no direct homology between therian dentitions and the arguably modified monotreme “tribosphenia” (Luo et al., 2001a, 2002, 2003; Kielan-Jaworowska et al., 2004), because of the radical lack of protocone in monotremes and a very lax if at all present functional correspondence in the molars. To the extent of the known materials, the same arguments can be extended to most of the remaining australosphenidans. Therefore, the presence of a true protocone or a functional analog is yet to be documented, and overall occlusion as evidenced by wear facets does not seem to agree closely with those of tribosphenic mammals. The presence of a small protoconal cusp in Vincelestes (Bonaparte and Rougier, 1987) that determines no evident wear on the talonid serves also as a cautionary note against deducing upper molar morphology based on lowers and vice versa. An even more dramatic example is provided by the large “protocone” of Shuotherium (Wang et al., 1988) that determines no wear on the talonid (Chow and Rich, 1982). We believe the presence of a “protocone” cusp (but with limited or no occlusal function in the talonid [Martin and Rauhut, 2005]) in the upper molars of australosphenids is probable.

An Alternative View of Australophenidan Molars

Based on the homologies accepted here and on the resulting cladogram, Australosphenida and Tribosphenida share the major trigonid cusps and the primitive posterior talonid cusp, traditionally viewed as the hypoconulid (Kermack, 1968; Crompton, 1971). The hypoconid and the entoconid would be independently acquired in the austrilosphenid and tribosphenid clades.

Following this interpretation, an alternative view of the austalosphenidan talonid cusps can be offered (fig. 6B). Australosphenidans are bracketed between “symmetrodonts” (Tinodon, Zhangheotherium) and basal cladotheres (Amphitherium, dryolestoids); therefore, it is appropriate to keep in mind symmetrodonts and dryolestoids when evaluating australosphenidans and attempting to understand their morphology. The lingual view of the molars of Ausktribosphenos, Bishops, Ambondro, and to a lesser degree Asfaltomylos and Henosferus, resembles closely a basal “symmetrodont” like Zhangheotherim or Maotherium (Hu et al., 1997; Rougier et al., 2003b), in particular the last premolar of the Ausktribosphenidae. As suggested by serial homology (Van Valen, 1994), the posterior cingular cusp of the last premolar must be the homolog of the large posterolingual cusp of the talonid (see Rich et al., 1999a, 2001a), which in turn is cusp d of therians and other mammals. We view as compelling the close morphological correspondence between the anterior cingular cusp, anterior portion of the lingual cingulid, position of the metaconid, posterior portion of the lingual cingulid, and posterior cingular cusp of the last premolar of ausktribosphenids and Ambondro on one hand and the “wrapping cingulid”, lingual metaconid, and “preentocristid offset and past the base of metaconid” of the molars in the remaning australosphenidans on the other. Under this view, the anterior cingular features used by Luo et al. (2001a, 2002) to diagnose australosphenidans become simply the retention of the anterior half of the lingual cingulid of basal trechnotheres, which, as in those basal forms, extends to the front of the tooth to reach an anterior cingular cusp. The primitive lingual cingulid becomes partially interrupted by the lingual position of the metaconid, and the distal portion of the cingulid is enlarged, reaching a relatively large posterior cingular cusp (cusp d or hypoconulid). This large cusp is the core of the australosphenidan talonid, and, therefore, the basin of the talonid would be formed by a buccal expansion. The “hypoconid” would be a neomorphic or enlarged cingular cusp (Hunter, 2004). The cristid obliqua can be directed to the hypoconulid (Hunter, 2004) or possibly more labially, contacting the lingual slopes of the “hypoconid” as in Henosferus. The nonhomology of the hypoconid is dictated by the phylogenetic position of the australosphenidans and is, in turn, reinforced under the interpretation suggested above. The elongated and crestlike entoconid of australosphenidans would be simply small cuspules of the posterior portion of the lingual cingulid. These cuspules vary in development and position in Ausktribosphenos and Henosferus, suggesting a cusp under no occlusal control. For a summary of homologies see figure 6B. Toothed monotremes would have greatly modified dentitions, with a complete anterior cingulid and anterior cingular cusp, a hypoconulid enormously developed, and a large “hypoconid”.

Following the logical implications of this alternative interpretation affects the scoring of characters 47, 48, 55, 56, 63–65, 67, 70–74, and 99. The searches with these changes, using the same set of ordered characters as in previous runs, result in two trees of 903 steps that agree closely with the one illustrated here in figure 9. The only substantive difference is the inclusion of multituberculates inside Australosphenida as the sister group of Monotremata, a result also obtained during the standard runs under certain conditions. Zhangheotherium and Tinodon are moved basal to dryolestoids. When the matrix was run unordered, 33 MPT trees of 857 steps were obtained. The consensus was poorly resolved and the 50% majority rule identified many poorly supported nodes. The branches in the vicinity of, and inside, Australosphenida either failed to form monophyletic clusters or had very low support. Not even the node Asfaltomylos/Henosferus was recovered in all the trees. This area of the tree is obviously very labile and susceptible to drastic topological collapses, probably due to the lack of cranial, postcranial, and upper dental information for most of the australosphenidans.

Regardless of which of the two main hypotheses of cusp homology are followed, ausktribosphenids are not clustered within eutheria. Addmittedly, our sample of crown-group Theria is poor to the extreme but several alternatives of character weighting, additivity, and reinterpretation have repeatedly failed to dislodge ausktribosphenids from a basal position in Mammalia. A southern (Gondwanan) origin for Eutheria based on australosphenidan taxa is at present unparsimonious.

Meckelian Groove and Cartilage

In medial view, the dentary of Henosferus has a sigmoid groove near the ventral border of the horizontal ramus that, from the level of the symphysis, extends backward to the area of the mandibular foramen (figs. 2, 5). The small groove descends toward the ventral border of the jaw and at the level of the m1 disappears; posteriorly it rises again and finally becomes confluent with the medial trough.

A thin medial groove, supposed for the Meckel's cartilage, has been widely reported for a variety of Mesozoic mammaliaform lineages (De Blainville, 1838; Owen, 1871; Osborn, 1888; Simpson, 1928b,c, 1929; later contributions are summarized in Kielan-Jaworowska et al., 2004). The groove was variously interpreted as an osteological correlate of the presence of a Meckel's cartilage (e.g., Flower, 1883; Bensley, 1902) or as a scar of a nerve or artery related to the mylohyoid groove of some current mammals (e.g., Owen, 1838; Simpson, 1928c). Usually, this groove extends from the symphysis back to the level of the mandibular foramen. The arrangement in the dentary is variable in each group; even specimens of same species show differences in the location and depth of the dentary medial groove.

Repenomamus robustus from the Early Cretaceous of Liaoning (China) and Gobiconodon zofiae (Li et al., 2003) have preserved an ossified Meckel's cartilage in its natural position, providing clear data for understanding the significance of the so-called Meckelian groove in Mesozoic mammals. As was noted by Meng et al. (2003), the split of the groove in the anterior and mid-portion of the dentary represents a trace left when the dentary wraps over the cartilage during ontogeny. Only the most posterior portion of the groove was occupied by the ossified cartilage in Repenomamus (Meng et al., 2003). However, the presence of the Meckelian groove does not directly imply the persistence of the Meckelian cartilage in the adult; the Meckel's cartilage could occupy the groove only during early ontogenetic stages and later disappear, as is the case among perinatal didelphids (Maier, 1993; Meng et al., 2003). The Meckelian groove in Henosferus passes ventral to the mandibular foramen (figs. 2, 5), as in many Mesozoic mammaliaforms (e.g., Morganucodon and Repenomamus). This ventral location of the groove in relation to the mandibular foramen is also evident in the development of the mandibule of the extant eutherians. The embryo of Rattus shows the mandibular foramen appearing dorsal to the Meckel's cartilage on day 19 (Tomo et al., 1997); in Ornithorhynchus, histological cross sections of the lower jaw of an immature individual show that the Meckel's cartilage runs ventromedial to the mandibular nerve (Zeller, 1989). The same is observed in marsupials (Toeplitz, 1920; personal observation). Comparisons with fossil mammaliaforms possessing similar structures on the medial side of the dentary and with ontogenetic data strongly suggest that the medial groove of Henosfenus corresponds to a Meckelian groove, which may or may not have had a persistent cartilage in the adult form. The evidence observed in Repenomamus and Gobiconodon zofiae (Wang et al., 2001; Li et al., 2003; Meng et al., 2003) indicates that possibly the medial groove of the dentary in most Mesozoic mammals such as cladotherians does not correspond to a Meckelian groove sensu stricto, but instead is a trace remnant left when the dentary wraps over the cartilage during ontogeny (Meng et al., 2003). The known interactions during embryology of the dentary and Meckel's cartilage (Zeller, 1989; Starck, 1995; Tomo et al., 1997) provide additional support to a dual nature for the Meckelian groove of Mesozoic mammals, with a wide array of morphologies, from forms with a cartilage almost completely exposed through life and extending from the symphysis to the back of the jaw to other forms in which the “groove” is more properly a suture of two dentary lips. In summary, Henosferus has a Meckelian groove extending from the symphyseal area to the level of the coronoid process, but the presence of Meckel's cartilage in this groove in the adult form is not demonstrated but could potentially remain lodged in the posterior portion of the trough, perhaps continuous with the malleus.

Dentary Trough and Contents

The dentary trough, or medial trough, of Henosfenos is clearly defined, limited dorsally by the medial flange and ventrally by the medial crest of the angular process (fig. 5). The medial trough of Henosferus is considerably shorter anteroposteriorly than in Morganucodon (Kermack et al., 1981) and docodonts (Krusat, 1980; Lillegraven and Krusat, 1991). Additionally, in Henosferus the condylar and angular processes are proportionately close to each other, restricting the anteroposterior extension of the groove. In contrast, Morganucodon has a condylar process lying well back from the level of the angular process, resulting in a broad dentary trough. Haldanodon (Lillegraven and Krusat, 1991), Castorocauda (Ji et al., 2006), and Docodon show an intermediate position of these processes with regards to Morganucodon and Henosferus. The dentary trough of Henosferus is clearly divided into two surfaces: one posterior to the mandibular foramen and leading into it, the other ventrally located to the surface mentioned above, connecting anteriorly with the Meckelian groove (fig. 5). There is a low, blunt ridge that separates these surfaces. Basal mammaliaforms such as Morganucodon, Haldanodon, and Castorocauda are known to have postdentary bones attached to the lower jaw and clear surfaces for their attachment (Kermack et al., 1973, 1981; Allin, 1975, 1986; Jenkins et al., 1983; Lillegraven and Krusat, 1991; Allin and Hopson, 1992; Clack and Allin, 2004; Ji et al., 2006). Morganucodon has a diagonal ridge running from the anterior end of the medial flange to the lower edge of the mandibular foramen, separating the prominent dentary trough from the posterior wall of the mandibular foramen (Kermack et al., 1973). A clear diagonal ridge is absent is Henosferus, but the blunt ridge between the two surfaces of the lateral trough seems to be homologous with the diagonal ridge of Morganucodon, so that a less complete subdivision into two surfaces is present in Henosferus. The area posterior to the mandibular foramen is interpreted here as being for the passage of the inferior alveolar nerve and artery that supply the body of the mandible and teeth and continue anteriorly as cutaneous branches to exit the dentary as a series of mental nerves through the mental foramina. If living mammals serve as analogs, these soft structures would reach the mandibular foramen from the posterodorsally located infratemporal fossa of the skull (see Zeller, 1989, for monotremes, and Spatz, 1964, for eutherians). The medial flange is slightly notched, or less developed, immediately distal to the mandibular foramen, a trait also present in Morganucodon, which suggests the path of a neurovascular bundle extending toward the mandibular foramen. The ventral facet, continuous with the Meckelian groove, requires further explanation. This facet widens posteriorly and, at the level of the angular process, is bordered ventrally by a distinct crest. In this region, the medial flange extends inward to approximately the same degree as the medial crest of the angular process, thus defining the dorsal border of the facet. The angular process is conspicuously wide transversely and concave posteriorly; the peculiar morphology of the process results in the formation of a restricted post-mandibular area, which, when considered in unison with the likely position of the angular (tympanic), hints to the presence of a middle ear diverticulum in connection with the postdentary tympanum, similar in position to that suggested for basal mammaliamorphs (Allin, 1986; Allin and Hopson, 1992; Luo et al., 2001b). Morganucodon (Kermack, 1973) bears a clear facet bordered dorsally by the medial flange and ventrally by a low crest, which supports the articular–prearticular and angular–surangular complex. Even though proportions and degree of crest development differ between Henosferus and Morganucodon, these taxa share a common morphology, absent in more derived mammaliaforms, that strongly supports the idea of the presence of postdentary bones in the specimen from Cerro Cóndor. It is probable that the postdentary complex was enclosed between the medial flange and the medial crest of the angular process and that the concave angular process was likely closely associated with the reflected lamina of the angular process where the tympanic membrane was attached. The relatively anteroposteriorly shorter medial trough suggests the presence of relatively smaller postdentary bones in comparison with Morganucodon, but we believe the close morphological similarities warrant the assumption that all major postdentary and paradentary elements were still present and substantially anchored to the dentary. The paradentary elements, coronoid and splenial, were not particularly large; the scar for the coronoid is distinct but occupies a small area at the base of the coronoid process (fig. 5), while the evidence for the splenial is less distinct. The splenial would be expected to cover medially the distal portions of the Meckelian groove (and cartilage if present) and extend posteriorly below the proximal portions of the dentary trough, but the ridges and rugosities in these areas are not particularly distinct, and, therefore, although we are convinced about the presence of the splenial, we are not certain about its extent. The postdentary elements would be proportionately smaller than in Morganucodon, Haldanodon, and Castorocauda and more loosely attached to the dentary, because distinct facets are not present in Henosferus. We believe this to be a natural feature because of the exquisite preservation of the Henosferus jaw MPEF 2353 (figs. 2, 5). Size and attachment of the postdentary complex has direct implication on the predicted dual cranio-mandibular joint. In Henosferus the dentary condyle is distinct, robust, and certainly the major structural link between skull and jaw; the predicted mandibular location for the tympanum in Henosferus has as a corollary the involvement of prearticullar (goniale), articular (malleus), and quadrate (incus) in a dual auditory and suspensory function. The degree of participation of the archaic mandibular articulation on the distribution of masticatory forces would depend on the strength of their attachments to the dentary (see below).

Unlike Morganucodon, but similar to Haldanodon (Lillegraven and Krusat, 1991) and Docodon (Simpson, 1929), the medial flange of Henosferus lies at the blevel of the inferior edge of the condylar process. If present, the postdentary bones would be supported dorsally by the medial ridge, and therefore the quadrate (incus)/articular (malleus) articulation would occur substantially below the major axis of the dentary condyle. The presence of two articulations that are not coaxial poses a biomechanical challenge whose details we are unsure how to resolve at present. Obvious models for a dual mandibular articular system are offered by embryos and perinatal mammals, in which the middle part of Meckel's cartilage and the first arch derivatives (such as the articular/malleus) are reduced or resorbed late in ontogeny (e.g., Maier, 1987, 1990, 1993; Zeller, 1989; Rowe, 1996; Meng et al., 2003), and by exceptional fossils (Wang et al., 2001; Meng et al., 2003). Coaxial dual articulations are mechanically simple and have been present in the temporomandibular joint (TMJ) in most of the basal mammaliaforms (Allin, 1975, 1986; Crompton and Hylander, 1986; Allin and Hopson, 1992; Rosowski, 1992). However, alternative noncoaxial contacts between the postdentary elements and/or Meckel's cartilage are likely to have occurred in fossils (Meng et al., 2003; Ji et al., 2006) and are present at least during some stages of development of recent mammals, thus offering a possible model for fossils like Henosferus. The tree topology obtained in this study includes Henosferus as a basal australosphenidan; monotremes appear to be a terminal group of this clade and would constitute the closest models for Henosferus. In fact, Rich et al. (2005a,b) have recently postulated the retention of postdentary elements in Mesozoic toothed monotremes, which makes consideration of monotreme anatomy even more relevant for the understanding of australosphenidan morphology.

Rich et al. (2005a) proposed an independent origin of the middle ear bones in monotremes and therians based on their interpretation that a basal toothed monotreme (i.e., Teinolophos trusleri) possessed middle ear bones still attached to the dentary. Rich et al. (1999a, 2002), however, do not accept links between australosphenidans and monotremes, but consider australosphenidans as members of basal Eutheria (Woodburne, 2003; Woodburne et al., 2003).

The presence of postdentary elements in Teinolophos (and by extension, ancestrally in monotremes) was challenged (Bever et al., 2005; Rougier et al., 2005) and reasserted by the original authors (Rich et al., 2005b). All of the known specimens of Teinolophos are fragmentary and represented only by dentaries; therefore, the actual morphology of any putative postdentary element is not known, but deduced from the morphology of the dentary. There is no objection against this practice (we ourselves have done so to understand the morphology in Henosferus), but the quality of preservation of the dentary is crucial for extrapolating the anatomy of the postdentary elements. In our view (Rougier et al., 2005), the specimen assigned to Teinolophos lacks unambiguous facets indicating the presence, or location, of any of the middle ear bones (articular, goniale, and incus), and even if the argument is relaxed to include the angular/tympanic (not a middle ear ossicle), we see no compelling evidence of a facet for the prearticular–articular complex. The presence of a “facet” for the angular/ectotympanic is ambiguous at best; based on all of the specimens, we are confident that the surface identified as a “facet” for the angular, in fact, extends inside the mandibular canal, a trait not seen in other nonmammaliaform cynodonts and mammaliaforms. The arrangment proposed by Rich et al. (2005a,b) is not supported by any model either living or fossil and can be more plausibly understood as the bottom of the enlarged mandibular foramen of monotremes that transmits a hypetrophied trigeminal system (Griffiths, 1978; Kuhn and Zeller, 1987; Zeller, 1989). Teinolophos has the distinct medial process that overhangs the mandibular foramen and determines an inordinately large mandibular foramen. A very large mandibular foramen is a rare condition in mammals but present in living and fossil monotremes (Musser, 2003), indicating the presence of a hyperdeveloped trigeminal system in the Cretaceous monotremes. This morphology lends support to our (Rougier et al., 2005) interpretation of the “facet” and flattened area of Rich et al. (2005a) as preservation artifact and floor of the mandibular canal, respectively. Despite troublesome preservation, we accept that ridges may be present in the back of the Teinolophos jaw. However, it does not show a suitable morphology to provide attachment of a sizable postdentary complex, no real medial flange is present (contra Rich et al., 2005a,b), and the weak ridge runs directly into the mandibular foramen, a fact more easily explained in connection to soft structures than postdentary elements.

In addition, Bever et al. (2005) remarked on the ambiguity of assigning the new specimens of Teinolophos to this species and argued for separated treatments of the specimens as different terminal taxa. We have argued (Rougier et al., 2005) that, even accepting the assumptions of the original authors (Rich et al., 2005a), the optimization is equivocal; under one of the two possible optimizations of the cladogram originally presented, independent origin of the middle ear in monotremes and therians is not supported. Therefore, the same cladogram also supported a single origin for the postdentary elements. In cases with ambiguous support, as here, it is a safe systematic practice not to regard only one optimization as supporting a given character transformation. The results of the optimization of this feature in the cladogram presented originally by Rich et al. (2005a) becomes a moot point when a larger set of relevant taxa, as we do here, is studied. If indeed Henosferus and monotremes are both members of Astralosphenida, the conclusion that the freeing of the middle ear elements in monotremes and therians is independent seems unavoidable. Rougier et al. (2005) recognized “other Mesozoic forms may question the monophyletic origin of the mammalian middle ear”; our more extensive study of Henosferus here further substantiates this challenge. According to our view, as deduced from the cladogram, the henosferids would retain a generalized mammaliaform arrangment of postdentary elements, reduced from the condition in Morganucodon and allies but still essentially mandibular in nature. Later australosphenids including ausktribosphenids and monotremes would have already achieved free, or mostly free, ear ossicles.

To sum up, Henosferus and the australosphenidans are interpreted as one of the basal branches of Mammlia and potentially forming part of the stem lineage leading to monotremes. Eutherian affinities for Australosphenida as a whole or for any of its members are here not supported and therefore a Southern continet origin for placentals is considered unlikely. The medial trough of Henosferus and its associated structures suggest the presence of postdentary bones at least partially attached to the dentary. These elements would be reduced in comparison with those of Morganucodon (Kermack et al., 1981). Henosferus presents the strongest evidence of postdentary elements for any australosphenidan, and, under the present cladogram, implies independent detachment of the angular and articular bones from the dentary of basal australosphenidans and therians. Obviously, the inclusion of Monotremata among australosphenidans is crucial for the postulation of independent liberation of postdentary elements among members of the crown group Mammalia. The inclusion of Monotremata in Australosphenida is a controversial topic (Luo et al., 2001a, 2002; Rich et al., 2002; Woodburne, 2003; Woodburne et al., 2003; Martin and Rauhut, 2005); our cladogram supports this membership (although the support of this branch is low; Bremer suport: 2), but we do not regard this study as a thorough exploration of this problem. Several preservational issues, such as the lack of upper dentitions, cranial, or postcranial material for any australosphenidans, make our results tentative. Better specimens of australosphenidans have the potential to drastically change the topology of the tree and the conclusions that can be drawn from it.