Institutional Abbreviations

Dentition

Scott gave the formula 4.1.3.3/4.1.3.3; Patterson suggested 5.1.2.4/4.1.2.4. Regarding the incisor count, Patterson was correct. Posterior to the intact upper I1s there are clearly four alveoli, including a partial right I5 in YPM-PU 15065 fig. 1. In addition, the left side of YPM-PU 15699 preserves the anterior four single-rooted incisors, with a root fragment of left I5 just anterior to the canine; on the right side I5 is entirely intact fig. 2; Patterson, 1958: 6. Identification of incisors vs. more posterior teeth is possible using morphology and origin, with incisors rooted at least partially in the premaxilla, and canines, premolars, and molars in the maxilla. YPM-PU 15065 shows breaks just anterior to the upper canines, within the fossa that receives each lower canine, which we believe correspond to the premaxilla-maxilla boundary fig. 1. All of the aforementioned upper incisors occur within the region we identify as the premaxilla. Lower incisors are more difficult to homologize. Of the anterior four lower teeth i.e., the putative incisors, the first is the largest, showing a root extending approximately to the base of the jaw below the lower canine. The succeeding three teeth are small and non-hypsodont, and in YPM 15699 appear to show enamel restricted to the crown, in contrast to Necrolestes' hypsodont cheek teeth. We follow both Scott and Patterson in identifying the pair of hypertrophied, trenchant lower antemolars as canines based on their form.

Figure 1

Skull of Necrolestes YPM-PU 15065 in stereo A dorsal, B ventral, C lateral views with D closeup of dorsal maxillary region showing hypsodont, rootless cheek teeth. Arrows in C and D indicate location of internal temporal space and course of infraorbital canal, respectively. Scale bars = 5 mm.

Figure 1

Continued

Figure 2

Necrolestes YPM-PU 15699 in stereo A dorsal, B ventral, and C lateral views, with inset showing anterior fragment of nasal bone in dorsal view. Scale bars = 5 mm.

Figure 2

Continued

The best criterion to distinguish molars from premolars is tooth replacement, a feature for which we unfortunately do not have any direct evidence in the YPM-PU specimens. However, new material described by Goin et al. in press shows an erupting premolar anterior to three more posterior cheek teeth, suggesting that Scott 1905 was correct in identifying three molars. We would therefore argue that the correct dental formula for Necrolestes is 5.1.3.3/4.1.3.3.

A break in the maxilla posterolateral to the lacrimal foramen on the right side of YPM-PU 15065 exposes the dorsal extremes of two posterior cheek teeth M1 and M2 following Goin et al., in press. These teeth are hypsodont, without roots, and retain enamel on their lingual aspect dorsally into the maxilla fig. 1D. When viewed from an anterior perspective, the cheek teeth of Necrolestes are curved from dorsal apex to occlusal surface, showing a lateral concavity. From a lateral perspective the cheek teeth are also curved, showing a posterior concavity fig. 1C.

Jaw

Patterson 1958: 8 suggested that although the angular process was small, it “bears an internal ledge and this is inflected and concave dorsally.” Ameghino 1891: 303 also described the angular process with a small and internally directed inflection. However, we do not regard this kind of inflection to be comparable to that of, for example, didelphids or borhyaenids. Rather, it shows some resemblance to that of Gypsonictops Clemens, 1973: figs. 2, 3. In fact, and somewhat ironically, the dentary of the placental afrotherian Chrysochloris shows a much more prominent, medially directed angular process than that of Necrolestes fig. 3, considerably broader than that of the marsupial mole Notoryctes fig. 3D. The morphology of the chrysochlorid angular process is no doubt influenced by its articulation with the hyoid apparatus Bronner, 1991.

Figure 3

Mandibles of A, B Necrolestes patagonensis A shows coronoid process of YPM-PU 15699; B shows jaw of YPM-PU 15384, C Chlorotalpa leucorhina ZMB 31505, and D Notoryctes typhlops BMNH 39.4210 in external top, internal middle and occlusal bottom views. Scale bars = 5 mm.

Scott 1905: 369 suggested that the dentary of Necrolestes is “exceedingly” like that of Chrysochloris. Because the coronoid process of the YPM-PU 15384 Necrolestes dentary fig. 3B is broken off close to its base, it shows a somewhat chrysochlorid-like form. However, not only is the coronoid process prominent in an intact specimen YPM-PU 15699, but it also shows an additional, posteriorly directed process close to its apex fig. 3A. Furthermore, compared with the jaw of Chrysochloris fig. 3C, Necrolestes has a more gracile mandibular angle, a prominent masseteric fossa, a more inferiorly situated mandibular foramen, and a shorter, dorsoventrally deeper mandibular symphysis. Hence, we disagree with Scott's assessment of similarity in jaw structure between the two taxa.

Cusp Homologies

Patterson 1958:4 referred to the main cusp of the upper molars of Necrolestes as “presumably paracones,” an interpretation that at first glance might be shared by many contemporary paleomammalogists. However, we note that the main cusp in molars of zalambdodont marsupials such as Notoryctes has been interpreted to be the metacone, not the paracone Archer et al., 2000; Long et al., 2002: 67; Asher and Sánchez-Villagra, 2005. Similarly, Goin and Candela 2004 recently described the putative metatherian genus Kiruwamaq based on a zalambdodont-like upper cheek tooth, showing a large metacone and diminutive paracone. Occlusal relations can elucidate some of these cusp homologies: for example, the paracone occludes in or adjacent to the lower ectoflexid, lateral to the talonid basin; the protocone occludes within the talonid basin; and compared to the paracone, the metacone occludes closer to the paracristid of the more posterior lower molar e.g., M1 metacone-m2 paracristid. With the caveat that occlusal relations across tooth loci are obviously contingent upon jaw mobility, and that noninterlocking upper and lower teeth are not as canalized as those that interlock Polly et al., 2005, we tentatively note that the main upper cusp of Necrolestes could be the metacone. As discussed by Asher and Sánchez-Villagra 2005: fig. 2, the main upper cusp of each molar in YPM-PU 15699, which preserves uppers and lowers in occlusion, shows closer proximity to the preparacristid of the next most-posterior lower molar than to any structure on the lower molar of the same tooth locus. We regard this occlusal pattern as suggestive that the main upper molar cusp in Necrolestes is the metacone, not the paracone.

Enamel Microstructure

An isolated right molar from YPM-PU 15384 was embedded in resin and sectioned longitudinally and transversely to obtain information about its enamel microstructure. Enamel terminology follows Koenigswald and Sander 1997a and Martin 1999a, b; for preparation techniques, see Martin 2004. The schmelzmuster of Necrolestes fig. 4 was compared to that of various Mesozoic taxa, as well as to data on small placentals and marsupials drawn from a large body of literature see Koenigswald and Sander [1997b] and literature cited therein. In addition, we compared the pattern of Necrolestes with that of sectioned right upper molars of Chrysochloris ZMB 76872, Notoryctes ZIUT SZ10068, and Potamogale ZMB 71592.

Figure 4

Scanning electron microscope SEM photos of sectioned and etched right M1s showing enamel microstructure of A, B Necrolestes YPM-PU 15384; occlusal surface to the left in A, C, D Notoryctes ZIUT SZ10068; tip of protocone to the right in C, E, F Chrysochloris ZMB 76872, and G, H Potamogale ZMB 71592; occlusal surface to the top in G. Longitudinal sections are listed in the left column, cross sections in the right. EDJ, enamel-dentine junction; IPM, interprismatic matrix; OES, outer enamel surface; P, prisms; RE, radial enamel; TE, tangential enamel. Scale bars = 50 µm.

The enamel distribution on the Necrolestes upper molar is very asymmetrical. Enamel is thickest 140 µm on the lingual side and very thin or even missing on the labial side. In the longitudinal section fig. 4A two zones are evident. The inner zone consists of radial enamel with the prisms inclined about 40° apically. After three fifths of the enamel thickness, prisms turn simultaneously in horizontal direction inclination 0°, forming an outer layer of tangential enamel. Prisms and interprismatic matrix IPM run parallel in the outer zone and therefore are difficult to distinguish. Prisms and IPM seem to be at least partially confluent, forming a prismless external layer PLEX. In the cross section fig. 4B, prisms are cut transversely in the inner zone and longitudinally in the outer zone. In this section the distinction between prisms and IPM in the outer zone is somewhat more clear. From the cross section it is evident that the PLEX apparently is restricted to the outer 10–20 µm of the enamel layer. The radial enamel of Necrolestes with a single turn of prisms is very similar to that of the other small placentals and marsupials that have been studied.

In longitudinal section fig. 4C, the marsupial mole Notoryctes exhibits a schmelzmuster with radial enamel and steeply apically inclined prisms in the inner zone and a simultaneous antapical turn of prisms in the outer zone. In the cross section fig. 4D, the prisms are cut obliquely transversely in the inner zone and almost longitudinally in the outer zone; the thick IPM is well exposed. The golden mole Chrysochloris has a very similar schmelzmuster with radial enamel in the inner zone and tangential enamel in the outer zone; prisms and IPM tend to be confluent. In the longitudinal section of the protocone, the radial prisms of the inner zone are cut longitudinally fig. 4E, and in the cross section they are cut transversely fig. 4F. The rounded-prism cross sections are surrounded by thick IPM. In the outer zone, prisms and IPM are confluent and hardly distinguishable. The otter shrew Potamogale has a generally similar schmelzmuster with apically inclined prisms in the inner zone and a simultaneous antapical turn in the outer zone figs. 4G, 4H. Prisms and IPM are more clearly distinct in the outer zone than in the other taxa studied.

The schmelzmuster of Necrolestes and the other taxa studied here represent the plesiomorphic therian schmelzmuster type that is typical for small marsupials Peradectia and Didelphimorphia and placental insectivorans Koenigswald, 1988, 1997a, b; Koenigswald et al., 1987; Koenigswald and Goin, 2000. A very similar schmelzmuster has also been detected in the molars of various Mesozoic non-therian taxa Wood and Stern, 1997; Wood et al., 1999 as well as in the incisors of zalambdalestids, pseudictopids, and early gliriforms such as Eurymylus and Eomylus Martin, 1999a, b, 2004. Evidently derived characters of gondwanatheres Koenigswald et al., 1999, such as an IPM that runs at a distinct angle to the prisms in the radial enamel of the molars and prisms increasing in thickness towards the exterior of the tooth, are not evident in our observations of Necrolestes. The schmelzmuster of Necrolestes represents the plesiomorphic therian condition that characterizes many small placentals and marsupials.

Basicranium

Scott 1905: 368 referred to an “ossified and moderately inflated” auditory bulla. However, we agree with Patterson 1958 that Scott was actually referring to the ventrum of the pars cochlearis itself. None of the Necrolestes material shows any sign of ossifications enclosing the middle ear figs. 1, 2. In fairness, the YPM-PU specimens do not have completely intact, articulated basicrania. The most complete is on 15699, which has a petrosal still articulated with the skull, bounded laterally by the jaw joint and anteromedially by the basisphenoid; however, regions posterior and medial to the petrosal are missing fig. 2. YPM-PU 15384 preserves a fragment consisting of petrosal, squamosal, and a part of the still-articulated mandibular condyle fig. 5. Neither specimen preserves an ectotympanic. They nevertheless clearly lack alisphenoid, basisphenoid, and petrosal contributions to an ossified bulla. Terrestrial mammals with prominent, ossified ecto- or entotympanic bullae tend to have strong coossifications and/or articulation scars, such that the contribution of tympanic elements to a bulla can often be inferred. Since the basicranial remains of Necrolestes show no sign of an ecto- or entotympanic bulla, we agree with Patterson 1958 that it lacked an ossified bulla.

Figure 5

Stereo views of Necrolestes YPM-PU 15384: A right petrosal and glenoid region in ventral view; B left petrosal fragment in dorsal view showing internal acoustic meatus bottom. Anterior is at top, lateral towards left. Note that the articulated mandibular condyle in A has been slightly displaced medially.

CT scans of the inner ear of YPM-PU 15699 and 15384 indicate that the cochlea was coiled, making nearly a full turn. While this degree of coiling is not as tight as in some placental mammals, it is considerably more than that seen in monotremes. Interestingly, hedgehogs, sea cows, vombats, and the marsupial mole Notoryctes are among the only extant therians that also show reduced cochlear coiling Gray, 1907; Sánchez-Villagra and Schmelzle, in press.

Adult monotremes, didelphids, dasyurids, and other metatherians such as Pucadelphys and Deltatheridium, but not Dromiciops, Mayulestes, or borhyaenids, possess a prootic canal Wible and Hopson, 1995; Wible et al., 2001. In didelphids, this structure transmits venous blood from the prootic sinus Sánchez-Villagra and Wible, 2002, passes through an osseous canal connecting the ventrum of the pars canalicularis of the petrosal and a small foramen immediately lateral to the secondary facial foramen, and leaves the tympanic region posteroventrally adjacent to the facial nerve. Except for petrosals assigned to the eutherian taxon Prokennalestes Wible et al., 2001 and specimens assigned to the taxonomically ambiguous “zhelestids” Ekdale et al., 2004, the prootic canal has not been explicitly documented in any eutherian mammal. The anteroventral pars canalicularis of Necrolestes is smaller than that of didelphids and shows no sign of a prootic canal fig. 5.

Wible et al. 2001: character 153 noted that the state of the internal acoustic meatus “shallow, with thin prefacial commissure” comprises a synapomorphy for Eutheria in an analysis of selected crown and stem Theria, focusing on extinct taxa Rougier et al., 1998. However, they noted in addition that the “shallow-thin” state also occurs in certain metatherians e.g., Turgiodon, deltatheridians, Pucadelphys, Marmosa, Didelphis, Dromiciops, and dasyurids. Necrolestes also exhibits a relatively thin prefacial commissure fig. 5, similar to that of Monodelphis.

Neither Patterson nor Scott mentioned the unusual character of the Necrolestes skull near the mandibular glenoid articulation. Although YPM-PU 15065 is missing most of the basicranium fig. 1B, it still retains the right glenoid articulation for the jaw, and shows considerable similarity to the jaw articulation in golden moles and Notoryctes. As opposed to the morphology seen in Monodelphis fig. 6, and resembling that of golden moles and Notoryctes, the mandibular glenoid fossa in Necrolestes is situated posteriorly, lateral to the pars cochlearis of the petrosal fig. 2B and probably also dorsal to the external auditory meatus, although the latter structure is not intact in any of the YPM-PU specimens. In addition, the postglenoid process in Necrolestes is diminutive, much smaller than that of Monodelphis.

Figure 6

Monodelphis domestica ZMB 35522 skull in stereo ventral view. Scale bar = 5 mm.

Dorsal to the mandibular glenoid fossa, the external sidewall of the braincase in YPM-PU 15065 shows a deep excavation, creating a space internal to the squamosal root of the zygomatic arch fig. 1C. As in chrysochlorids such as Eremitalpa, this space probably served to increase surface area for the attachment of temporalis musculature.

Immediately lateral to the exit foramen for the facial nerve in the tympanic roof, a space is evident that we believe housed the epitympanic recess fig. 5A. This region is best preserved with a damaged roof in the semiarticulated petrosal-squamosal fragment of YPM-PU 15384, and is consistent with the enlargement of an element within the ossicular chain, as in most golden moles. However, because the internal temporal region fig. 1C dorsal to the mandibular glenoid remains relatively flat, a hyperinflated malleus of the kind observed in Chrysochloris asiatica Mason, 2003 was not present in Necrolestes.

Scott 1905: 370 stated that “the condylar and carotid foramina are not displayed in any of the specimens,” whereas Patterson suggested the presence of the carotid foramen in YPM-PU 15065. We agree with Patterson regarding the presence of a carotid foramen, but on the basis of a different specimen. Unfortunately, except for an isolated right petrosal, the “good deal of the structure of the basicranium and auditory region” mentioned by Patterson for YPM-PU 15065 now appears to be lost. Nevertheless, YPM-PU 15699 retains an articulated right petrosal and part of the basisphenoid, within which are right and left foramina anteromedial to the petrosal, and which correspond closely in position and morphology to carotid foramina in Monodelphis. In contrast to Monodelphis, the carotid foramina in Necrolestes open directly onto the dorsum of the basisphenoid, whereas in Monodelphis each carotid foramen defines a short, anteroposteriorly directed canal within the basisphenoid. In addition, Monodelphis possesses a transverse canal that is situated anterior to the carotid foramen and is intramural, not exposed within the braincase see character 2 of Sánchez-Villagra and Wible, 2002. In Necrolestes, slitlike openings on the dorsum of the basisphenoid are evident medial to the dorsal exposure of each carotid foramen fig. 7; these are continuous with foramina exposed on either side of the midline anterior to the carotid foramina and comprise the intracranial apertures of each transverse canal fig. 2B.

Figure 7

Dorsal view of mid-braincase floor in Necrolestes YPM-PU 15699, showing intracranial apertures of carotid foramina and transverse canals. Scale bar = 5 mm.

YPM-PU 15384 preserves fragments of both occipital condyles, one of which is associated with a partial atlas see below. Neither is sufficiently well-preserved to document the presence/absence of a hypoglossal = “condylar” foramen.

Both Scott 1905 and Patterson 1958 noted the unusual internal nares of Necrolestes, which show a narrow, posteriorly expansive element dividing the internal nasal aperture into two halves figs. 1B, 2B. A clear suture with the basisphenoid is evident immediately posterior to the rodlike element dividing the internal nares. The identity of this element is unclear. In Leptictis Novacek, 1993: fig. 9.5, Monodelphis Wible, 2003: fig. 5, and Euphractus Wible and Gaudin, 2004: fig. 2, the ventral, midline element immediately anterior to the basisphenoid has been labeled the presphenoid; whereas in another figure of Monodelphis Novacek, 1993: fig. 9.4, it is depicted as the vomer. Starck 1995: 41 labeled as “praesphenoid” a small, midline-exposed element anterior to the basisphenoid in a newborn Mirounga Carnivora; on the same figure he depicts as the “orbitosphenoid” the adjacent ossification that also shares a ventral suture with the basisphenoid.

According to McDowell 1958: 127, the presphenoid comprises a “median structure of the endochondral cranium anterior to the basisphenoid and formed by the fusion of mediad extensions of the two orbito-sphenoids.” McDowell goes on to state that in golden moles, some species of which e.g,. Amblysomus gunningi, TM 42117 resemble Necrolestes in the morphology anterior to the basisphenoid, the orbitosphenoids are “absent”. In fact, the orbitosphenoids are present in golden moles, but are too small to exhibit a ventral midline junction Broom, 1916. Similarly, Starck 1995: 42 states that the presphenoid results from the ossification of midline, ventral extensions of the cartilaginous “Alae orbitales” and “Radices pro- und metopticae” of the developing mammalian central stem. Starck 1995: fig. 31 repeats the conclusion of Roux 1947 that golden moles possess the cartilaginous precursor of a presphenoid.

Debate on the homologies of the elements comprising the “central stem” of the mammalian cranial base dates from Parker 1885, Broom 1927, and Roux 1947, among others. Both Broom and Roux argued for the presence of a presphenoid in golden moles based on their observation of a median, unpaired ossification center, independent of the orbitosphenoids, anterior to the cartilaginous precursor of the basisphenoid. The term “parasphenoid” has been used in reference to an element in nonmammalian amniotes, hypothesized to be the homologue of the vomer and/or presphenoid in mammals e.g., DeBeer, 1937: 433; see Novacek, 1993: 453–454. The vomer itself comprises much of the bony nasal septum throughout mammals, but does not typically articulate with the basisphenoid. Hence, for present purposes, we use the term presphenoid for the skull-base element anterior to the basisphenoid in Necrolestes.

In Monodelphis, Euphractus, Leptictis, and most other therian mammals, the bony roof of the posterior nasal fossa, corresponding to the ventrum of the posterior lamina transversalis or posterior-most “choanal passage” of Cave 1948: fig. 73, is flat and undivided; division of the nasal fossa occurs more anteriorly, within the nasal fossa proper. Necrolestes, in contrast, shows a median ridge extending posterior to the internal nares, probably composed of the presphenoid and, more anteriorly, the vomer. This ridge appears to be continuous with the median nasal septum within the nasal fossa.

Lacrimal and Orbit

Patterson 1958: 7 stated that the lacrimal foramen is “very similar to that of various didelphids as regards both position and size.” In our view, however, the two taxa are roughly similar only in that both have foramina dorsal to a relatively long infraorbital canal, as is common in several marsupial and placental groups. Didelphids typically show multiple lacrimal foramina Sánchez-Villagra and Asher, 2002, whereas only one is evident in Necrolestes fig. 2. Furthermore, didelphids show lacrimal foramina at the anterior margin of the orbit, facing posteriorly Wible, 2003: fig. 3; Necrolestes, in contrast, shows the lacrimal foramen situated in a groove dorsal to the infraorbital canal, medial to a broad emargination in the maxilla figs. 2C, 2D.

Patterson 1958: 7 also stated that “the optic foramen appears to be confluent with the sphenorbital fissure.” However, the orbital region is poorly preserved in all available specimens, and we can confirm neither the contour of the sphenorbital fissure nor the confluence of an optic foramen with it.

Jugal bone

Although he acknowledged that sutures in the zygomatic region were obliterated, Patterson 1958: 6 suggested that in Necrolestes, the jugal extends posterior to the glenoid surface of the jaw, and regarded this morphology as indicative of marsupial affinities. However, without intact sutures, neither the posterior nor anterior extent of the jugal bone in Necrolestes can be confidently identified. Hence, we do not believe this element can be interpreted as didelphid-like. Necrolestes resembles the marsupial Monodelphis in showing a preglenoid process anteriorly buttressing the mandibular condyle in its glenoid articulation. As noted by Patterson, the immediate postglenoid region in Necrolestes is relatively flat see above, whereas Monodelphis shows a large postglenoid process fig. 6. Several eutherian groups also possess processes of the jugal contributing to the glenoid jaw articulation Asher et al., 2003.

Palate and Rostrum

We agree with Patterson's 1958: 7 statement that “the long, slit-like incisive foramina are very didelphid-like”. The incisive foramina appear anterior to a roughly symmetrical, slightly distorted break that we believe corresponds to the maxilla-premaxilla suture fig. 1. We also agree with his comment that Necrolestes' palate is solid and lacks palatal vacuities. YPM-PU 15065 is the only specimen to expose this region fig. 1B. It shows a break in the posterior palate just right of the midline, but this is not repeated on the left side, indicating that in life the hard palate was not perforate.

Scott 1905: 368 claimed the “nasals are long and tapering” in Necrolestes. We suspect he was referring to the general contour of the rostrum. As Patterson 1958 pointed out, actual sutures for most of the dermal cranium, including the nasals, are completely fused. Hence, it is not possible to infer the extent to which the posterior nasal bones widen posteriorly, or if they articulate with the lacrimal, as is often the case in marsupial groups e.g., Didelphis; see Horovitz and Sánchez-Villagra, 2003.

Cervical spine

In most mammals e.g., Chrysochloris, fig. 8C, the atlas is composed of a right and left lateral mass, connected dorsally by a neural arch and ventrally by an intercentrum and/or a ventral arch. Several diprotodont marsupials, some specimens of the peramelid Echymipera e.g., ZMB 37012, and some Mesozoic therians Kielan-Jaworowska, 1977; Kielan Jaworowska et al., 2004: 483 show an incomplete atlantal ventral arch and no intercentrum but still show a fused neural arch Horovitz and Sánchez-Villagra, 2003. Juvenile monotremes display an unfused or only partially fused suture between left and right neural arches Griffiths, 1978; Lessertisseur and Saban, 1967a, b, as do the presumably juvenile specimens of the basal prototribosphenidan Vincelestes neuquenianus Rougier, 1993. Later in ontogeny, the neural arches of monotremes fuse to each other; adult monotremes possess a complete, ring-like atlas. In Vincelestes the intercentrum is missing, but may have been present in life. Separate atlantal hemiarches and a separate intercentrum have also been reported in several close relatives of mammals, such as nonmammalian cynodonts Jenkins, 1971, the basal mammaliamorph Oligokyphus Kühne, 1956 and the basal mammaliaform Megazostrodon Jenkins and Parrington, 1976.

Figure 8

Atlantes of A Necrolestes YPM-PU 15065, B Echymipera ZMB 37012, C Chrysochloris ZMB 76896, and D Chlamyphorus ZMB 6007. For each specimen, anterior view is at top, posterior middle, and lateral bottom. Scale bars = 5 mm.

McKenna et al. 2000 reported that a juvenile specimen of the Cretaceous eutherian Daulestes may have possessed unfused ventral and neural arches of the atlas; whether this condition characterized adult specimens is unclear. Ontogenetic studies of the development of beagle dogs report fusion of the atlantal hemiarches by postnatal day 106 and fusion of the intercentrum to the massae laterales by postnatal day 115 Watson et al., 1986, cited in Evans, 1993; ontogenetic studies of Didelphis also indicate differential timing in fusion: fusion of the ventral arch occurs “much later” than that of the dorsal arch Oliveira et al., 1998: 117.

Based on five preserved hemiatlantes and/or fragments thereof, representing all three of the Hatcher-Peterson specimens, Necrolestes may be added to the very short list of mammals with completely separate right and left atlantal halves fig. 8A. Neither Scott 1905 nor Patterson 1958 mentioned any details of atlantal morphology in Necrolestes, perhaps because its atypical appearance hindered identification. Nevertheless, we believe this attribution is correct for two main reasons:

Figure 9

A atlas of Necrolestes YPM-PU 15699 in anterior top, posterior middle, and internal bottom right views. B associated right atlantal and occipital fragments from YPM-PU 15384 in posterior view. Right atlantal fragment has been displaced postmortem and does not occlude naturally with occipital condyle; however, association between atlantal and occipital fragments is genuine. Scale bar = 5 mm.

Unfortunately, there remains some uncertainty as to the correct provenance of certain elements among the three YPM-PU specimens. The right atlas fragment that we now associate with YPM-PU 15699 fig. 9A based on preservation, color, and nonduplication of elements had been in the same box with elements from YPM-PU 15384. We do not believe these atlantal elements are associated with 15384 because the atlantal fragment shown in figure 9B, associated with YPM-PU 15384 based on preservation and association with a fragmentary petrosal on which “15384” is written, duplicates the right atlantal lateral mass shown in figure 9A. The black-colored right lateral mass in figure 9A is in our view a better candidate for association with the YPM-PU 15699 skull, and in addition is associated with a left atlas fragment preserving most of the neural arch fig. 9.

The left neural arch of the YPM-PU 15699 atlas and the right neural arch of the YPM-PU 15065 atlas fig. 8 preserve undistorted apices without signs of breakage. This is consistent with an unfused soft-tissue joint between right and left atlantal neural arches. The neural arch is missing on the YPM-PU 15384 atlantal fragment.

The process for the cranial articular fovea on the atlas protrudes anteriorly beyond the neural arch. There are no lateral vertebral foramina, just a groove on the medial side of the root of the process for the cranial articular fovea on the left and right sides as in Notoryctes and Amblysomus. There are no wings per se, but just ridges along the lateral sides of the massae laterales, a morphology similar to that found in Amblysomus, Scapanus, and Condylura. A foramen on each side perforates the lateral ridges and opens dorsally in the anterior area of the neural arch just medial to the processes for the cranial articular facets, to the area just ventral to the lateral ridges, also in the anterior area of the vertebra.

The right neural arch is preserved in YPM-PU 15065. It is narrower anteroposteriorly than the massa lateral, with a dorsal tubercle insinuated in the medial-most end of what is preserved of the neural arch. The caudal articular fovea is preserved in both specimens. It is almost flat, with a roughly triangular outline. It converges ventrally with the cranial articular fovea, from which it is separated by a very narrow surface pointing medioventrally. This narrow surface bridges the separation between the cranial and caudal articular facets and is smooth like an articular facet; it was therefore not part of an unfused suture with an intercentrum. It may have made contact with an unpreserved intercentrum or with the dens of the axis.

A partial axis is preserved for both YPM-PU 15384 and 15065 fig. 10. The dens, cranial articular surface, part of the body, and a very small portion of the left neural arch are preserved. The ventral surface of the body is smooth with no traces of ridges or median keel. The cranial articular facets end at the dens and do not extend ventral to the dens. Not enough of the neural arches are preserved to assess the presence of an enclosed transverse foramen. However, the body of the axis is anteroposteriorly much longer than those of other vertebrae, including those in the thoracic and lumbar regions fig. 11. We believe this is due to fusion. Contrary to Scott's statement 1905: 371 that the cervical vertebrae of Necrolestes do not resemble the fused cervicals of Notoryctes, we believe that the cervical spine of Necrolestes was extensively fused, possibly including C2 to C6 fig. 11. In addition to its fragmentary but relatively elongate axis, YPM-PU 15384 shows another cervical fragment with an antero-posteriorly elongate body. This does not show a clear fit to the 15384 axis, but may still have been fused to it, separated by fragments that are now missing fig. 11. Both elements are considerably more elongate than an isolated vertebra that we interpret to be a posterior or posterior-most cervical.

Figure 10

Axes of Necrolestes A YPM-PU 15065, B YPM-PU 15384, in dorsal top, ventral middle, and anterior bottom views. Scale bars = 5 mm.

Figure 11

Ventral views of cervical vertebral skeleton in A Amblysomus Asher's collection, B Notoryctes ZMB 35694, C Necrolestes atlas, YPM-PU 15065; axis and cervical vertebrae, YPM-PU 15384, and D Chlamyphorus ZMB 6007. Scale bars = 5 mm.

Thoracic and Lumbar Spine

Scott 1905: 371 stated that in both chrysochlorids and Necrolestes, spinous processes throughout the vertebral column extended posterodorsally; neither taxon possessed anticlinal extending dorsally at a right angle to the vertebral column or anterodorsally extending spinous processes. We agree with Scott based on the material available to us. The lack of anticlinal and anterodorsally directed vertebral spinous processes is also evident in Dasypus, but contrasts with Talpa, which shows anterodorsally directed spinous processes in the anterior lumbar and posterior thoracic region. Articulations between vertebrae in this region are limited to zygapophyses and vertebral centra, as in other non-xenarthrous mammals.

Sacrum

see “pelvic girdle” section under “hind limb”.

Tail

Patterson 1958: 9 noted a “well-developed tail” in Necrolestes as a similarity to didelphoids. Here again it seems that some of the YPM-PU Necrolestes material may have been lost, because we can identify just two caudal vertebrae, both from YPM-PU 15384 fig. 12. One of these is broad at one end and may have been the proximal-most caudal vertebra, adjacent to the sacrum. Both caudal vertebrae are considerably larger than any vertebrae in the tails of Amblysomus hottentotus and Talpa europaea, both of which are similar in overall size to Necrolestes but have reduced tails. Necrolestes may have resembled Notoryctes in terms of tail size, but unlike the latter shows relatively small transverse processes and neural spines on the two preserved caudal vertebrae. We cannot infer too much detail about the tail in Necrolestes, but agree with Patterson that it was probably longer and more robust than that of Talpa or golden moles.

Figure 12

Dorsal left and ventral right views of caudal vertebrae in Necrolestes YPM-PU 15384. Scale bar = 5 mm.

Scapula

For our descriptions we orient the scapula in the anatomical position of most quadrupedal tetrapods, rather than humans. That is, the “dorsal” margin of the scapula refers to its vertebral margin; its ventral aspect includes the glenoid articulation with the humerus; and its lateral surface anchors the scapular spine.

Only the humeral articular regions of the right and left scapulae from YPM-PU 15384 remain intact fig. 13A. These show a scapular spine defining infra- and supraspinous fossae; however the spine itself is incomplete, missing the metacromion and acromion. In didelphids and xenarthrans e.g., Dasypus, Zaedyus, Chlamyphorus, the scapular spine transversely bifurcates the lateral scapular surface into cranial supraspinous and caudal infraspinous regions of similar size; i.e., the spine is situated in roughly the middle of the lateral surface. Necrolestes resembles the condition shown in Notoryctes fig. 13B and golden moles fig. 13C in which the scapular spine originates closer to the cranial border of the scapula. Furthermore, as in Amblysomus and the talpid Galemys, the root of the scapular spine in Necrolestes extends ventrally to a point close to the glenoid fossa, unlike that of xenarthrans, didelphids, Metacheiromys, and Notoryctes, which stops well dorsal to the scapular neck. Another similarity to Amblysomus, Galemys, some xenarthrans e.g., Tamandua, and epoicotheres e.g., Metacheiromys is the caudal border of the scapula, which curves and approaches the scapular spine, giving the infraspinous fossa a tube-like appearance fig. 13A. Otherwise, the scapula of Necrolestes does not resemble the elongate, narrow scapula of talpids. Furthermore, it shows no sign of a coracoid process. Overall, Necrolestes most closely resembles the golden mole Amblysomus in its scapular morphology.

Figure 13

Scapulae in A Necrolestes YPM-PU 15384, lateral fragment in lateral top and inferior bottom views; B Notoryctes ZMB 35694 in lateral top and posterior bottom views, and C Chrysochloris ZMB 76897 in inferior left, lateral top right and posterior bottom right views. Scale bars = 5 mm.

Humerus

Right and left humeri are at least semi-intact in two specimens: YPM-PU 15384 and 15065. The best preserved is the right humerus of 15384 fig. 14A. The head of the humerus is mediolaterally compressed, reminiscent of that of moles, golden moles Horovitz, 2004, and the epoicotheriid palaeanodonts Xenocranium pileorivale, Epoicotherium unicum Rose and Emry, 1983: fig. 8B and D respectively, and Dipassalus oryctes Rose et al., 1991: figs. 5A, C and D. Greater and lesser tubercles are similar in size, but the greater tubercle is located more proximally than the lesser tubercle. The head projects proximally beyond both of them. There is a conspicuous depression on the anterior side of the head, in the area of transition into the diaphysis, medial to the lesser tubercle. The diaphysis of the humerus is straight for the most part in medial/lateral view, except for the proximal end that curves forming a concavity under the head. The deltopectoral crest is raised and convex in the proximal area of the diaphysis, and its convexity is continuous with that of the head and more distally with the greater tubercle. It becomes distally thinner, in the shape of a crest with a concave edge directed laterally, and protrudes terminally in a large process on the midshaft. This process is missing its distal tip in all four YPM-PU specimens. The shape of this crest resembles that of Plesiorycteropus MacPhee, 1994 and Dasypus novemcinctus, except that in the latter two the crest is not continuous with the head, but with the proximal end of the greater tubercle. In addition, Necrolestes shows an eminence running distally from the deltopectoral crest towards the medial epicondyle, forming a bridge over the entepicondylar foramen, and comprising a very prominent feature of the humeral shaft proximal to the articular surface of the trochlea and capitulum fig. 14. Amblysomus possesses a similar, more gracile structure, also extending distally from the deltopectoral crest towards the medial epicondyle, defining the entepicondylar foramen fig. 14B.

Figure 14

Humeri of A Necrolestes YPM-PU 15384, B Chrysochloris ZMB 76897, and C Notoryctes ZMB 35694 in posterior left and anterior right views. Scale bars = 5 mm.

The medial epicondyle is directed mediodistally and protrudes distally beyond the trochlea and condyle, similar in this respect to that of Epoicotherium unicum Rose et al., 1991: fig. 6. In Necrolestes there is a small process next to the trochlea most evident in the right humerus of 15384 and a concave, distomedially directed edge between this small process and the medial epicondyle. Both of these features are absent in Epoicotherium, where the lateral edge of the medial epicondyle starts next to the medial edge of the trochlea. In Notoryctes and Amblysomus there are very large medial epicondyles as well, but in Amblysomus the epicondyle is directed almost straight medially and in Notoryctes it protrudes slightly distally, and is shorter. In addition, Notoryctes lacks an entepicondylar foramen fig. 14C.

The supinator crest is large and laminar, reaching proximally to midshaft, close to the level of the deltopectoral process. Its proximal end extends abruptly out of the humeral shaft, at a level just inferior to the deltopectoral crest, creating a flange similar to that seen in Amblysomus fig. 14 and even more strongly developed in Palaeanodon Rose, 1999: fig. 2, Metacheiromys Simpson, 1931: fig. 16, Epoicotherium, and Xenocranium Rose and Emry, 1983: fig. 8.

The trochlea is cylindrical, and shows a subtle transition to the capitulum defined by a waisted neck. The capitulum is also cylindrical, but broadens laterally, and is mediolaterally wider than the trochlea. The trochlea is similar in shape to that of Epoicotherium; however, the capitulum of Necrolestes differs in being cylindrical, rather than sphere-shaped as in Chrysochloris and Amblysomus. The anconeal process of the ulna articulates in a mediolaterally long depression, running dorsal to the posterior articular surface of the trochlea and capitulum.

In Amblysomus, the articular surface on the distal humerus for the radius and ulna is fairly restricted to a lateral position. As in other fossorial taxa, the medial epicondyle provides greatly expanded surface area for attachment of extrinsic hand musculature; in golden moles, it provides the proximal support for the ossified digital flexor tendon. This “articulation” is not synovial; rather, an unossified, tendinous insertion connects the more distally ossified flexor tendon with the medial epicondyle.

Forelimb

One of the most conspicuous features of the Necrolestes forearm is the medially curved olecranon process of the ulna Rose and Emry, 1983: fig. 9. This feature is present to an even greater degree in the palaeanodonts Xenocranium and Epoicotherium Rose and Emry, 1983: fig. 10. Golden moles, Notoryctes, and Euphractus show a similarly curved ulnar olecranon; in golden moles this process bends in a slightly more posterior than medial direction. The anterior face of the olecranon process in Necrolestes is relatively flat. A facet for the radial head is evident lateral to the ulnar coronoid process. The ulnar shaft is substantial and broadens slightly toward the carpus, coming to a point only at the styloid process on its distal extreme, which extends from the posterior surface of the distal ulnar margin. Relative to the distal radius, the ulna comprises a small component of the carpus-forearm articulation.

Golden moles possess one of the most remarkable forelimbs among Mammalia, showing three long bones of the forearm: radius, ulna, and an ossified flexor tendon that approaches the humeral epicondyle and articulates with the carpus fig. 15D; see Dobson, 1883: 121. Based on the description of Leche 1907, who called the palmar sesamoid of Notoryctes “unverkennbar homolog” unmistakably homologous with the ossified flexor tendon of golden moles, W.K. Gregory 1910: 256 stated that the “third lower arm bone” is at least partly present in Notoryctes. As described in Stirling 1891, 1894, Gadow 1892, Wilson 1894, and Carlsson 1904, and as evident in an articulated skeleton available to us ZMB 35694, the forearm of Notoryctes does possess a large palmar sesamoid, into which M. flexor digitorum inserts Wilson, 1894. However, this element does not extend proximal to the carpus, and resembles palmar sesamoids present in, among others, Chlamyphorus and Talpa fig. 15A, B more than it does the ossified flexor tendon in golden moles fig. 15D. In fact, Carlsson 1904: 116 stated explicitly that the marsupial mole lacks a third-forearm bone, contra Leche 1907 and Gregory 1910.

Figure 15

Hands of A Chlamyphorus ZMB 6007, B Mogera ZMB 77008, C Necrolestes YPM-PU 15384, and D Amblysomus TM 43872. Roman numerals I–V indicate digital rays. For carpal terminology we follow Horovitz and Sánchez-Villagra 2003. Scale bars = 5 mm.

An articulated right ulna, radius, and some carpals are preserved for YPM-PU 15384 fig. 15C, referred to by Patterson 1958 as “various bones of the fore foot”. Other small elements with articular surfaces are present, including several phalanges, a metacarpal, and indeterminate distal carpal elements, one of which is still partly embedded in matrix. At some point these elements may have had clearer associations with one another; however, at present we restrict our discussion of carpal anatomy to those elements still articulated to the distal radius.

The right radius is articulated with a proximal carpal, probably the lunate fig. 15C. A moderately large pisiform is also present, adjacent to the styloid process of the ulna. Between pisiform and distal radius, another ossification is evident, larger than both lunate and pisiform. This element extends from a position adjacent to the lunate towards the ulnar shaft fig. 15C, and possibly provided for the insertion for M. flexor digitorum. This proximal carpal ossification in Necrolestes is considerably more elongate than the palmar sesamoid of Notoryctes and shows a greater similarity to the ossified flexor tendon of golden moles. Scott 1905: 374 also interpreted the forearm of Necrolestes to have had a chrysochlorid-like flexor tendon, “though only the distal portion of it is ossified”.

The phalanges show elongate fissures extending longitudinally on either side of each terminal phalanx fig. 16, resembling those of Xenocranium figured by Rose and Emry 1983: fig. 15. We cannot be sure if these belong to the manus or the pes. If the former, then they would comprise an important difference from the manus of both Notoryctes and golden moles. In most of these species, digit III is considerably larger than the others; digit IV is reduced to a stub; and digit V is lost altogether Hickman, 1990: fig. 4. In Notoryctes, digits III and IV are enlarged relative to the others Wilson, 1894. Most golden moles except Eremitalpa, which shows similarly sized claws on digits I, II, and III and Notoryctes do not show three terminal digits of the hand that resemble each other in size. Such would be the condition in Necrolestes if the three phalanges in figure 16 belong to distinct digits of the manus. However, the lack of an articulated specimen prevents certainty on this point.

Figure 16

Terminal phalanges of Necrolestes YPM-PU 15384. Scale bar = 5 mm.

Pelvic Girdle

YPM-PU 15384 retains a fragmentary sacrum articulated with seven more anterior vertebrae fig. 17. Right and left os coxae are also preserved but not articulated; the proximal tip of the left ilium remains preserved in the matrix anterodorsal to the articulated sacrum and lumbar spine. Both Scott 1905 and Patterson 1958 described a well-developed pubic symphysis for Necrolestes. Although none of the YPM-PU specimens available to us retains a complete pelvis, we agree that Necrolestes had a moderately broad pubic symphysis based on the right os coxae of YPM-PU 15384. In this specimen, the pubis broadens distally and extends in a ventromedial direction from the acetabulum fig. 17. In contrast, the pubis of Dasypus, Zaedyus, Talpa, and chrysochlorids narrows distally and extends in a caudal direction from the acetabulum, rather than ventromedial. We find no evidence to support Patterson's 1958: 9 statement that Necrolestes had epipubic bones. The epipubic bone of Didelphis has a broad origin along the cranial margin of the pubis, an area which is relatively intact on the right os coxae of YPM-PU 15384 fig. 17 and that shows no signs of articulation with another bony element. The epipubic bone of Notoryctes and the symphysis itself fig. 18 are reduced but still associated with a prominent emargination along the dorsal rim of the pubic symphysis. We find in Necrolestes no sign of any “small fragment of bone attached to the [right pubis] by matrix [that] could be a remnant of the epipubis itself” Patterson, 1958: 9.

Figure 17

A lumbar spine, sacrum, and left ilial fragment and B isolated right os coxae of Necrolestes YPM-PU 15384 in internal top, external middle, and anterior bottom views. Scale bars = 5 mm.

Figure 18

Pelves of A Amblysomus Asher's collection and B Notoryctes ZMB 35694 in ventral top, dorsal middle, and lateral bottom views. Lateral view in Notoryctes is reversed image from left side. Scale bars = 5 mm.

Necrolestes exhibits many characters that typify fossorial mammals in several orders, particularly in its forelimb, skull, and cervical vertebrae. However, compared to those of fossorial xenarthrans, Notoryctes, talpids, manids, and Orycteropus, the sacrum and pelvis of Necrolestes are unremarkable. The articulated sacrum in YPM-PU 15384 is slightly distorted and incomplete caudally, but consists of three fused sacral vertebrae, with an unfused articulation between centra and zygopophyses of the proximal sacrum with the last lumbar vertebra fig. 17. The neural spine of the last lumbar vertebra projects posterodorsally and is not fused with adjacent neural spines, unlike the fused dorsal keel evident in European talpids, for example. The only articular surface evident on the sacrum of YPM-PU 15384 fits a corresponding surface on the internal aspect of its ilium, as in golden moles and most nonfossorial mammals. Because the ischia on the two preserved innominates are both incomplete, it is impossible to rule out an ischial-sacrum articulation, as seen in many xenarthrans that possess an accessory obturator foramen MacPhee, 1994. However, given the gracile sacrum, we believe such accessory articulations were unlikely and that Necrolestes possessed a relatively small pelvis without the hyperossifications seen in certain other fossorial mammals e.g., Notoryctes, see fig. 18B. Pelves of Metacheiromys and Alcodontulum Palaeanodonta described respectively by Simpson 1931: fig. 18 and Rose et al. 1992: fig. 4 also resemble that of Necrolestes in showing moderate sacral fusion and a simple articulation with the vertebral column at the ilium.

The right ilium is missing the wing, although part of the auricular surface is visible on the medial side. On the lateral side the ilium ends posteriorly in a raised region on the edge of the acetabulum. This eminence is likely to be for the attachment of the M. rectus femoris. This position is similar to that in the palaeanodont Alocodontulum atopum Rose et al., 1992: fig. 9 rather than in Escavadodon zygus Rose and Lucas, 2000: fig. 11, where the eminence is located anterior to the acetabulum. This eminence contributes some articular surface to the acetabulum in Necrolestes. The lunate surface is therefore very wide anteriorly and becomes narrower dorsally and posteriorly. The caudal border of the lunate surface ends high up posteriorly because of breakage. The ventral edge of the cranial region of the lunate surface is damaged, and it is not possible to determine its position. The pubis has suffered some deformation; however, it is possible to discern an iliopubic eminence, as found in such mammals as Palaeanodon ignavus Emry, 1970: fig. 29, Escavadodon zygus, and Alocodontulum atopum.

Femur

Left and partial right femora are available for YPM-PU 15384 fig. 19A. The right femur missing the femoral head had been in the same box as YPM-PU 15699 in a capsule labeled “15699 or 15384”, but is marked with “15384” on its diaphysis. Its potential association with 15384 may have previously been viewed as problematic based on the absent femoral head, potentially representing an unfused epiphysis of a younger individual than the left femur. However, we believe the epiphysis on the right femur had been solidly fused to the diaphysis, but broken off as a simple post-mortem artifact unrelated to age. We therefore presume that both femora belong to YPM-PU 15384.

Figure 19

Femora of A Necrolestes patagonensis YPM-PU 15384, B Notoryctes typhlops ZMB 35694 with inset showing lateral aspect of Notoryctes' left knee; and C Chrysochloris asiatica ZMB 76897, in anterior left and posterior right views. Hole in Chrysochloris femoral neck is a postmortem artifact. Scale bars = 5 mm.

Necrolestes lacks a femoral neck, with the femoral head projecting anteriorly as well as medially from a proximal femur compressed in an anteroposterior dimension fig. 19. The greater and lesser trochanters are not distinct; and there is no trochanteric fossa. A very shallow trochanteric fossa has been noted in Xenocranium and is absent in Manis and Amblysomus, among other mammals Horovitz, 2004.

A flattened proximal femur and absent trochanteric fossa are also found in Notoryctes fig. 19B. The A-P compressed proximal femur in Notoryctes owes its peculiar shape to the expanded greater trochanter. As documented by Carlsson 1904, this area provides a large area for attachments of gluteal musculature and powerful internal and external vasti, the latter two inserting onto a greatly enlarged patella fig. 19B. These muscles contribute to hip flexion, adduction, and crus extension. Uniquely among fossorial mammals, and particularly noteworthy for a marsupial which with the exception of peramelids generally lack patellae, the large size of the patella in Notoryctes may indicate some function in the animal's subterranean environment. Despite its enlarged patella, the distal femur of Notoryctes is smooth without a well-defined patellar trochlea.

Both Notoryctes and Necrolestes show femoral condyles that are wider mediolaterally than proximodistally. The lateral condyle in both taxa is proximodistally narrow and shows more protrusion laterally than its medial counterpart. Necrolestes shows a slight fossa that may correspond to a patellar trochlea, but is not as clearly defined as the trochlea in Chrysochloris. We have not identified a patella among the YPM-PU material attributed to Necrolestes.

Distal Hind limb

Patterson 1958: 3 reports the presence of the “…tibia, fibula, calcaneum, astragalus, and cuboid” for YPM-PU 15699. Unfortunately, almost all of this material appears to have been lost since Patterson's study. The only element distal to the knee remaining for any of the three Hatcher-Peterson specimens is a proximal fibula fragment fig. 20, currently in a box with elements from YPM-PU 15384. Because it is mediolaterally flat and anteroposteriorly broad, this specimen resembles the proximal fibula of Didelphis, Notoryctes, and Manis. Unlike chrysochlorids and xenarthrans, the proximal fibula and tibia do not appear to have been fused.

Figure 20

Proximal fibula of Necrolestes YPM-PU 15384. Scale bar equals; 5 mm.

Scott 1905: 378–379 also discusses material of the hind limb that at one point was part of the Hatcher-Peterson specimens, and goes into more detail on the structure of the foot than Patterson:

Australosphenida

The clade encompassing monotremes, Ausktribosphenos, Bishops, and other southern-hemisphere tribosphenic mammals, either on the branch including the Asian taxon Shuotherium or immediately distal to it, shows two optimized characters in common with Necrolestes: a triangular alignment of the main cusps of the anterior lower molar, and pronounced shearing between molar postvallum and prevallid. Significantly, these characters are also present in northern tribosphenic mammals, and reflect the convergent acquisition of this kind of occlusion in holarctic vs. southern radiations of Mesozoic mammals Luo et al., 2001. Otherwise, Necrolestes does not share any of the unambiguously optimized synapomorphies, all of which are dental, with the stem clade encompassing monotremes.

Theria

Necrolestes shares with crown Theria several characters of the skull and skeleton, including a coiled cochlea of the inner ear, absence of a septomaxilla, and following the descriptions of Scott [1905] presence of an astragalar neck. The scapular spine, present in Necrolestes, has also been discussed as a therian synapomorphy Jenkins and Weijs, 1979; Sánchez-Villagra and Maier, 2003. Furthermore, Necrolestes shows a therian-like glenoid articulation for the humerus fig. 13, with no sign of the independent coracoid ossification present in monotremes. On the other hand, according to Luo et al., 2002: character 134, a supraspinous fossa defined in part from a scapular spine is present not only in therians, but also in some dryolestoids Henkelotherium, triconodont Gobiconodon, and amphilestids Jeholodens. According to their phylogeny, the scapular spine does not optimize as a therian synapomorphy.

Necrolestes lacks many of the characters optimized for Theria. Due to its zalambdodont dentition and lack of a protocone, nearly all of its upper molar occlusal surface corresponds to the stylar region of the tribosphenic molar. Hence, many of the dental characters listed by Rougier et al. 1998 and Luo et al. 2002 as diagnostic for Theria are absent e.g., extent of pre- and postprotocristae, molar conule morphology. In addition, unlike the character reconstruction hypothesized for basal Theria in Rougier et al. 1998: supplementary data, Necrolestes shows only a single mental foramen on the exterior of its mandible, inferior to its canine.

Metatheria

Of the characters optimized unambiguously as metatherian synapomorphies by the data sets identified above, Necrolestes shares several: lack of a stapedial artery sulcus, palatal process of the premaxilla reaching the canine alveolus, and presence of three premolars see discussion of dental homologies above. Of the features of the jaw hypothesized by Rougier et al. 1998: supplementary data to be derived for Metatheria, Necrolestes shares three, all of which are “absence” characters: it lacks a labial mandibular foramen Kielan Jaworowska and Dashzeveg, 1989: 18, a Meckelian groove, and a coronoid facet.

Several other characters contrast with those optimized for the metatherian common ancestor, such as a double-rooted upper canine, and a noninflected mandibular angle a derived feature of some marsupial species such as the koala. Without ontogenetic data, we cannot infer the condition for dental replacement Horovitz and Sánchez-Villagra, 2003: character #175. However, based on the observation of Goin et al. In Press that the jaw of Necrolestes shows an erupting posterior premolar anterior to three molar loci, one may infer that this animal had three upper and three lower molars, unlike the four reconstructed for the common ancestor of metatherians.

Marsupialia

Compared to the reconstructed common ancestor of crown marsupials, Necrolestes shares a ball-shaped distal = styloid process of the ulna. It also shows two small foramina anteromedial to the carotid foramina, also exposed on the dorsum of the basisphenoid. We interpret these to be transverse canal foramina fig. 7. Otherwise, unlike the marsupial common ancestor, it lacks an alisphenoid tympanic process, shows a solid, nonperforated palate, and the capitulum of the distal humerus has an articular surface extending farther proximally than the trochlea.

Eutheria

The inferred presence of three molars, all similar in size, fits the morphology of the eutherian common ancestor as reconstructed by Rougier et al. 1998. In addition, characters from the matrix of Asher et al. 2003, 2005 show additional similarities: the zygomatic process of the squamosal is posteriorly gracile; the incisive foramina are relatively small; and the angle of the jaw is not inflected fig. 3. Features of Necrolestes that do not match the reconstructed eutherian ancestor include the molariform appearance of the premolars, absence of protocones and protocristae, absence of a caudal tympanic process of the petrosal, and a reduced coracoid process of the scapula fig. 13A.

Placentalia

The only two unambiguously optimized characters of crown Placentalia from Asher et al. 2003, 2005 are absence of epipubic bones and a single hypoglossal = “condylar” foramen. Patterson 1958 believed that the Hatcher-Peterson Necrolestes specimens indicated presence of epipubic bones. However, as summarized above, we see no sign of these in any of the specimens available to us. Similarly, fragmentary occipital remains of the Necrolestes material do not clearly show the morphology or number of hypoglossal foramina that may have been present.