Oldfieldthomasia (Unnamed Basal Typothere Group, Oldfieldthomasiidae)

Oldfieldthomasia has long been placed in its own family (Simpson, 1967; Billet, 2011), to which a handful of other Paleogene taxa of debated affiliations (e.g., Acropithecus rigidus) have gradually accreted, thus possibly rendering the family paraphyletic (see García López, 2011; Reguero and Prevosti, 2010; Billet, 2011). Available cranial material of this genus is decidedly poor, but retains historical importance because of Simpson's (1936) study of Oldfieldthomasia cf. debilitata AMNH-VP 26800, discussed here in some detail (see figs. 1, 5–7, 20; appendix 1). This specimen is (or was) a partial skull from the locality of Cerro Blanco, now known to be of Middle Eocene age (cf. Reguero and Prevosti, 2010). Virtual sections of the preserved left hemicranium are of limited use because so much of the fossil has been lost or damaged (figs. 1, 7). Indeed, Simpson may have chosen this particular specimen to work on because it was already much battered and destructive sampling could be justified. This may have been the wrong choice for experimentation, however, for his results were predictably poor and he never attempted to prepare another South American ungulate in this manner.

TABLE 1

List of Notoungulate Specimens (E= Early; Eoc. = Eocene; L. = Late; M. = Middle; Mioc. = Miocene; Pleist. = Pleistocene)

Continued

Other Specimens Studied. Also examined were the holotype skull of Oldfieldthomasia debilitata (MACN 10376) and another cranial specimen in the AMNH-VP collection (AMNH-VP 28896), but because of their condition (which Simpson [1932a: 7] unaccountably described as “well preserved”) they supplied nothing of value to the interpretation effort.

Paedotherium (Hegetotherioidea, Hegetotheriidae, Pachyrukhinae) and Relatives

The target genera Paedotherium and Pachyrukhos (nec Pachyruc[h]us, Pachyrucos [McKenna and Bell, 1997: 466]), which are morphologically extremely similar, contain many named species. As Cerdeño and Bond (1998) have shown, sorting out which nominal species belongs to which genus is not a straightforward matter, and in the past generic assignment often seems to have been based on age rather than morphology. In any case, apart from the teeth, discernible differences in the crania of these taxa are few and possibly largely size related, with definite Pachyrukhos being somewhat more robust than definite Paedotherium. For the purposes of this study, little of value is lost by frequently ganging them together as “pachyrukhines.”

Pachyrukhinae is an indisputably monophyletic clade of hegetotheres whose known range extends from Deseadan or even Tinguirirican to ?Ensenadan time (Cerdeño and Bond, 1998; Reguero et al., 2007; Reguero and Prevosti, 2010; Billet, 2011). Pachyrukhinae is invariably paired with Hegetotheriinae to form Hegetotheriidae, the sister group of which is the grouping of “late” nominal archaeohyracids identified by Billet et al. (2009; Billet, 2011). No cranial material of earlier members of Pachyrukhinae (Deseadan Prosotherium and allies; see Reguero and Prevosti, 2010) has been described.

Specimens Studied. Available specimens of Paedotherium (figs. 8–12) were in much better condition than those of Pachyrukhos. The sample examined in this paper includes AMNH-VP 45914, an excellent skull of Paedotherium chapadmalense (probably a synonym of P. typicum) and some good material of P. bonaerense (as relimited by Cerdeño and Bond [1998], here including MLP unnumbered 1 and P. insigne MLP 99-X-2-1). A number of other Paedotherium specimens in table 1 (e.g., MLP unnumbered2), mostly lacking dentitions, cannot be validly referred to species and are thus listed as “sp.” Other specimens: Pachyrukhos typicus MACN 1251–52; Pachyrukhos sp. MACN 5886. A representative of the related taxon Hegetotherium andinum (MLP 12–2914a) was examined but not scanned. (This genus needs revision, and this species name may not be valid; G. Billet, personal commun.)

Mesotherium pachygnathum AMNH-VP 14509, representing a second family, Mesotheriidae, was only cursorily examined and was judged a poor candidate for scanning. However, the late survival of this genus suggests that good material should exist and should be sought for future work.

Cochilius (Interatherioidea, Interatheriidae, Interatheriinae) and Relatives

Cochilius volvens, a moderately early member of the Interatheriidae, is represented in the comparative set by AMNH-VP 29651 from a locality “south of Lago Colhué-Huapí” (Simpson, 1932b: 1), not otherwise identified. Structural details are well preserved at the microanatomical level, with few seriously fragmented areas (figs. 13–15). Other specimens examined: Cochilius volvens MLP 2339 (ex coll. Lillo).

Protypotherium is common in Patagonian Santacrucian localities and thus well represented in museum collections. MLP 12–2780 is waterworn and partly encased in matrix; although damaged on exposed surfaces, scanning revealed that matrix-covered areas were quite intact. A mineral sinter, investing all surfaces, is particularly noticeable on trabeculae filling the epitympanic sinus; digital removal did not provide acceptable results, so images are presented as they appear in the original scans (figs. 16–17).

Other specimens examined: Interatherium (= Icochilius) extensum MACN 9738–39; Interatherium rodens MACN 9857–58.

Progaleopithecus tournoueri, which has a complicated taxonomic history, is of interest because it is part of a group standing as sister to Cochilius, Protyp other ium, and Interatherium (see Hitz et al., 2006). Unfortunately, the only available specimen with a caudal cranium (AMNH-VP 29603) turned out on scanning to be much battered and figures little in this account.

Notopithecus adapinus AMNH-VP 28949 and 28673, representing the earliest conventionally recognized interatheriid subfamily (Notopithcinae), are very poorly preserved, although virtual sections of the skull were of some use in connection with gathering data on the presence of an entotympanic and septum bullae in notoungulates.

Other Placental Taxa

Interpretation of the highly derived cranial organization of pachyrukhine notoungulates was aided to some degree by studying taxa to which they have been historically and are still occasionally compared, as examples of detailed convergences among placentals arising from vastly different phylogenetic progenitors (figs. 18, 19). The list of useful comparisons could be long, but as this is not a functional anatomical study, attention was focused on examining selected young and/or damaged specimens of a few informative taxa in the AMNH-M's extensive holdings of Glires.

Specimens studied (all AMNH-M): Leporidae: jackrabbit, Lepus californicus 177068, 5887. Dipodomyidae: Merriams kangaroo rat, Dipodomys merriami 8598, 182081; Heermanns kangaroo rat, D. heermanni 124172; painted spiny pocket mouse, Liomys pictus annectens 190257; long-tailed pocket mouse, Chaetodipus formosus 248737; Merriams pocket mouse, Perognathus merriami 188632.

Epitympanic Theca: Definition and Development

As Roth (1903) was the first to adequately document, middle ear pneumatization is notably extensive in notoungulates, and a capacious, dorsally prominent paratympanic space on the caudal cranial sidewall has long been counted as a derived character shared by all but the most primitive members of this group (Roth, 1903; Patterson, 1936; Cifelli, 1993; Gabbert, 2004; Reguero and Prevosti, 2010; Billet, 2011). In comparative morphology this space is regarded as a derivative of the epitympanic recess, the excavation that typically lodges the heads of the malleus and incus as well as the dorsalmost part of the membranous sac that fills the adult middle ear, the cavum tympani (van der Klaauw, 1931; Fleischer, 1973; MacPhee, 1981; Macrini et al., 2010). The dorsal wall of the recess is normally, but not always, formed by the petrosal (tegmen tympani) and a shelf (or epitympanic wing) of the squamosal, or both projections in proportions that vary considerably among mammals (van der Klaauw, 1931; MacPhee, 1981; Maier, 2013).

In the assumed primitive therian condition (van Kampen, 1905; van der Klaauw, 1931; Starck, 1967; Novacek, 1977, 1993; MacPhee, 1981; MacPhee and Cartmill, 1986), the recess is merely a dimple in the tympanic roof, but in many mammals it becomes a major paratympanic space under the pervasive influence of middle ear pneumatization. Great expansion of the recess is generally regarded as part of an adaptation for low-frequency sensitivity (e.g., Fleischer, 1973, 1978; Webster, 1975; Webster and Webster, 1975). Although at the cellular level the remodeling process is fundamentally the same wherever it occurs in the skull (Fawcett, 1997), the “same” paratympanic space may be positioned in somewhat different locations in different taxa, which has led, perhaps inevitably, to the coining of a plethora of synonyms in the mammalogical literature (e.g., “mastoid bullae” in notoungulates [Sinclair, 1909]; “anterior mastoid sinus” in Dipodomys [Howell, 1932]; “dorsal bulla” in Cavia [Gacek, 1975]).

In notoungulate studies the traditional term for both the volume and its bony covering is epitympanic sinus, but its use applied to the latter element is objectionable because a sinus is by definition a vacuity, not a solid structure. I therefore propose the alternative term “theca² epitympanica” to refer to the bony shell, of whatever composition, that encloses the epitympanic sinus and communicates with the tympanic cavity proper by an aditus (= foramen pneumaticum).³ Billet and de Muizon (2013) identified a small excavation in the tympanic roof of an isolated notoungulate petrosal from Itaboraí as the epitympanic recess. There is no suggestion of an aditus, suggesting that the epitympanic sinus as a paratympanic space was primitively absent.

FIG. 5.

Oldfieldthomasia cf. debilitata AMNH-VP 28600, right hemicranium, parasagittal sections (after Simpson, 1936: fig. 5; original magnification, cited as natural size); rostral to right. This plate includes the only section illustrations of elements Xa and Xp in Simpsons paper (ss. 10, 15, 20, and 23; cf. comparable virtual parasagittal sections in fig. 6). Neither of these putative elements passes tests for recognizing independent ossifications; see text and appendix 1. Key (Simpsons nomina in parentheses, if different from usages preferred here; periods in original omitted): caf, facial canal (canalis facialis); cpg, retroarticular canal/foramen (postglenoid canal); crty, crista tympanica; cty, tympanic cavity; ets, epitympanic sinus; fcbr, cerebral fossa; fsmp, stylomastoid foramen (foramen stylomastoideum primitivum); fg, glaserian fissure (fissura glaseri); fpg, retroarticular canal/foramen (postglenoid foramen); fpn, aditus of epitympanic sinus (foramen pneumaticum); fsm, stylomastoid foramen; fv, venous foramen; fvpl, posterolateral cerebral venous foramen; mae, external acoustic meatus; oc, occipital; pa, parietal; pae, porus acusticus externus; pe, petrosal (periotic); pet, squamosal theca (pars epitympanica of squamosal); ppg, retroarticular process (postglenoid process); ppo, paroccipital process; ppt, posttympanic process; ret, epitympanic recess; set, epitympanic sinus; sg, mandibular fossa (glenoid surface); sq, squamosal; svpl, posterolateral venous sinus; ty, ectotympanic (tympanic); vph, hyoid recess (vagina processus hyoidei); Xa, element Xa (anterior adventitious element); Xp, element Xp (posterior adventitious element).

Epitympanic Theca: The Question of Composition

In Roth's (1903) opinion, the notoungulate epitympanic sinus lay within, and was thus covered by the “pars mastoidea,” which he regarded as homologous with the mastoid area of humans. Although it had long since been accepted that the human mastoid is simply an integral part of the petrosal, and not a phylogenetically independent ossification,⁴ Roth (1903: 17) invited much later confusion by asserting that, at least in notoungulates, his pars mastoidea was part of (?or incorporated) a distinct ossification that was separated from the squamosal proper by features that he named “sutura zygoma-serrialis” and “sutura squamoso-mastoidea (escamo-serrial).” Additionally, because Roth also claimed that additional sutures (e.g., “sutura mastoideopetrosa”) separated the serrialis-mastoid from the petrosal, it followed that much of the dorsal sidewall of the notoungulate skull—including the epitympanic theca, roof of the external acoustic meatus, and nearby processes—had to be formed by this unusual element, homologous in his view with the prosquamosal (= supratemporal) of anomodontian therapsids (fig. 2).

Roth's claim to have discovered an element heretofore undetected in the mammalian cranium was a major challenge to conventional morphological thinking, and it did not go unchallenged. To be sure, some investigators working on notoungulates in the early part of the last century, such as Sinclair (1909), Scott (1912a, 1912b), and Patterson (1932, 1936), accepted one or another of Roth's claims, albeit with reservations. Sinclair (1909: 71), for example, thought he had found one of Roth's “extra” sutures in a specimen of Hegetotherium, and also noted that in Pachyrukhos “the greatly distended mastoid bulla[e] … show but slight contact with the squamosal,” thereby implying that the epitympanic thecae were largely enclosed by petrotympanic material, if not the serrialis per se. Scott (1912b) was more dubious, although he utilized the term “pars serrialis” (but “of the squamosal,” he insisted, not “of the mastoid,” as Roth had asserted). Patterson (1932: 18–19) also spoke uncertainly of “the pars serrialis of the squamosal, that bone having retained the sutures [sic] between its three centers of ossification in some of his [i.e., Roth's] specimens. Between this pars serrialis and the mastoid Roth never made a clear distinction.” Patterson kept looking for evidence, however (e.g., Riggs and Paterson, 1935: 190), concluding in the case of one aged specimen of the isotemnid toxodontian Pleurostylodon (E. Eoc.) that, despite the absence of any traces whatsoever, “[adventitious] elements may be present.”

FIG. 6.

Oldfieldthomasia cf. debilitata AMNH-VP 28600, left hemicranium, virtual parasagittal sections (for approximate locations see fig. 1A); rostral to right. All to scale in C. A and B, sections through lateral cranial wall and external acoustic meatus: single white asterisk, definite suture between rostral crus of ectotympanic and alleged element Xa; double white asterisks, fracture zone (not a suture) passing through mandibular fossa, same as feature 3 in figure 1A. C, section through lateral portion of tympanic cavity, tegmen tympani: single black asterisk is adjacent to definite suture between caudal crus of ectotympanic and alleged element Xp; dashed white arrow extends from facial canal into aperture of stylomastoid foramen, to mimic route of cranial nerve VII; narrow dashed lines suggest trend of thecal walls, very little of which is left. All other discontinuities are interpretable as fractures. As is evident, this side of the skull of AMNH-VP 28600 is in worse condition than the side that Simpson (1936) prepared. Although Simpson stated that he could detect sutures that completely, or almost completely, delimited his elements Xa and Xp from surrounding bones, he failed to adequately label these sutures on section drawings. In this figure, the labels “Xa” and “Xp” are placed approximately where Simpson situated them in his published drawings. Ignoring fractures, the area of Xa is seen to be completely continuous with the preotic or retroarticular portion of the squamosal, of which it must be considered an inherent part. Because of breakage, the precise relationships of Xp and the petrosal tegmen tympani, caudal crus of the ectotympanic, and epitympanic theca are uncertain. The interpretation favored here is that Xp is indistinguishable from (the rest of) the retrotympanic process and is therefore mostly or entirely squamosal. Note that, as Simpson delimited it, Xp is not much more than a bony conduit for the transiting facial nerve. Perhaps differences in bone density in this area contributed to Simpsons belief that an independent element was present (see fig. 20). He was, however, correct in concluding that the epitympanic theca is mostly formed by the squamosal. The condition of the bulla in this specimen precludes identification of an entotympanic, if any, in Oldfieldthomasia. A few of the many broken (br) areas are specially noted.

Simpson (1936: 9) was more definitive in his assessment:

For Simpson (1936: 8)—and, given his influence, for virtually every subsequent student of the notoungulate cranium—the only plausible conclusion was that “the epitympanic sinus is not in the mastoid but in part of the squamosal.”

The two specimens to which Simpson (1936) referred in this quotation were not further identified; however, the examples reproduced in figures 2–4 are ones for which Roth specifically identified bounding sutures (Toxodon, Pachyrukhos) or the position of the serrialis (Nesodon). With regard to Toxodon, Roth must have investigated material in which sutural markings were (for that genus) remarkably clear, because basicranial sutures on most toxodontian skulls tend to be obliterated, or at least difficult to trace, even in the rare event that the skull itself is well preserved (cf. van Kampen, 1905: 617; Patterson, 1932).

By contrast, Paedotherium and its close relative Pachyrukhos are excellent candidates for the present investigation. Key sutures on pachyrukhine caudal crania are easily discerned, as they evidently remained open over the entire life span, and both genera are well represented in major paleontological collections. Indeed, it is to conditions in pachyrukhines rather than Oldfieldthomasia that Simpson (1936) should have turned when undertaking his investigation of cranial composition in notoungulates, as they unquestionably possess the most derived squamosal seen in this group. If their thecal composition is nevertheless closely comparable to that of the other notoungulates examined here, as I maintain, then it is extremely unlikely that other, unstudied taxa will radically differ.

Paedotherium

Despite the derived nature of Pachyrukhinae, the three specimens (AMNH-VP 45914, MLP 52-IX-28-29, and MLP unnumbered2) on which this section is chiefly based provide excellent insights into the morphology of the hegetotherioid caudal cranium. As in the case of the other specimens reviewed in this paper, all are fully adult.

The existence of unexpected elements cannot be discounted merely because their distinctiveness is hard to establish in the adult stage. This is particularly important to emphasize in the case of pachyrukhines, because in these hegetotherioids conditions are complicated. At first glance the squamous portion of the squamosal appears to be reduced to an insignificant rod, the squamosal bar, and thus seems an unlikely source for the thecal covering. In fact, the rostral part of the squamosal is directly connected to the caudal cranium at only one place, the tiny area of continuity between the squamosal bar and the rostral thecal wall (fig. 8D). Sections support this impression (fig. 12A, C). Although frequently broken in specimens, this area of continuity is not the site of a detectable suture (fused or otherwise) and is no wider than the bar itself, i.e., ∼2 mm. To explain how such remarkably derived conditions may nevertheless conform to those found in other notoungulates requires considerable morphological exploration and comparisons with other placentals with similarly modified squamosals, provided in the rest of this section.

Pachyrukhine notoungulates are chiefly famous for their remarkable morphological and presumed ecological convergence on certain members of Glires, particularly certain leporids and caviomorphs (Sinclair, 1909; Kraglievich, 1936; Scott, 1937; Cifelli, 1985; Dozo, 1996, 1997; Reguero et al., 2007). Leporid similarities are especially noteworthy and include such features as the long, narrow rostrum bearing lateral rarefactions, lengthy diastemata, markedly hypsodont/hypselodont cheek teeth, ever-growing incisors, voluminous orbits, basicranial kyphosis, persistent fontanelles or cranial hiatuses, deep mandibular ascending rami, highly reduced and barlike squamous portion of the squamosal, and possible leaping adaptations (Sinclair, 1909; Scott, 1937; Cerdeño and Bond, 1998; Cassini et al., 2010).

A previously unexplored example of leporid/ pachyrukhine convergence—the presence of a potential intracranial joint (ICJ)—has, on investigation, turned out to be basic to the proper interpretation of the pachyrukhine caudal cranium. The existence of an ICJ in certain lagomorphs (e.g., jackrabbit, Lepus californicus) was first hypothesized by Bramble (1989), who pointed out that, in the midcranial area, a complex of aligned sutural and synchondrosial dense connective tissues intervenes between bone territories in such a way that bone-on-bone contact within the joint is minimized. Providing a kind of syndesmotic “O-ring” of soft tissues, this arrangement divides the skull into two functional moieties: the rostral, comprising all the elements lying in advance of a semicoronal plane through the sphenooccipital synchondrosis, and the caudal, comprising the rear part of the cranium together with the auditory regions. In L. californicus there is probably very little actual movement at the ICJ; its actual purpose, as inferred by Bramble (1989), is to act as part of a shock-absorbing apparatus to prevent deformation of the orbits (and therefore visual impairment) during the strike phase of ballistic leaping (see also Stott et al., 2010). On the other hand, short-limbed, nonleaping Ochotona has some of the osteological correlates noted above for Lepus in even more exaggerated form (see Wible, 2007), but with much larger paratympanic spaces, suggesting isolation of the auditory apparatus may also be significant for other reasons as well.

FIG. 7.

Oldfieldthomasia cf. debilitata AMNH-VP 28600, left hemicranium, virtual coronal section; for approximate location see figure 1B; lateral to right. In this section, the only suture found on preserved part of lateral wall of auditory region is the squamoso-ectotympanic suture; retroarticular process formed jointly by squamosal and ectotympanic, with no evidence of additional element Xa. Petrosal material contributes modestly to medial margin of aditus to epitympanic sinus and therefore to thecal wall. Matrix (see fig. 1) filling epitympanic sinus has been partially removed to facilitate labeling. Auditory ossicles—fragments of malleus and incus—are out of position in aditus to epitympanic sinus.

Regardless of putative ICJ function, the alignment of relevant sutures, synchondroses, and hiatuses in pachyrukhines is certainly similar enough to their leporid counterparts to make comparison worthwhile (cf. figs. 8–9 and fig. 18). Sutures concerns us first. Immediately subjacent to the mandibular fossa the squamosal and alisphenoid form a delicate preotic flange of bone (not to be confused with the preotic crest, a quite different feature found in some placentals; MacPhee, 1981). In well-preserved Paedotherium fossils this flange touches, but is not fused to, the surface of the auditory bulla. (The flange in the specimen depicted in figure 8B is somewhat damaged.) Although it barely qualifies as such, topologically the contact between flange and bulla can be described as comprising the glaserian fissure in this taxon. Termination of the preotic part of the squamosal at the glaserian fissure would normally mean that the squamosal does not contribute to structures lying directly caudal to this point, such as the epitympanic theca.

FIG. 8.

Paedotherium sp. AMNH-VP 45914, stereopair views of caudal cranium (with keys) in ventral (A), right lateral (B), right caudal (C), and dorsal (D) aspects, on this and subsequent pages. All to scale in A. In key diagrams, squamosal (including external aspect of epitympanic theca) is white; other elements colored differing shades of gray. Question marks indicate indistinct or inferred borders between squamosal theca and some other element; hachure indicates major breakage. Delicate ventrocaudal extremity of paroccipital process is usually broken off in pachyrukhine skulls (see Sinclair, 1909: pl. X, text-fig. 4). Key to A: 1, petroso-exoccipital suture; 2, basioccipito-bullar suture; 3, exoccipito-bullar suture (on paroccipital process); asterisk lies in foramen sometimes identified as “foramen lacerum medium” or “carotid canal” in notoungulate studies (but see Gabbert, 2004). Note small, cylindrical tympanostyloids situated within hyoid recesses. Unidentified large basioccipital foramina (multiple on specimens left side) lying close to the midline are likely venous ports. Key to B: asterisk in dorsal midcranial hiatus. Klinocrany is obvious in this aspect; compare less derived Cochilius (fig. 13B). Note within dorsal hiatus the slight exposure of petrosal (question mark) and tentorial processes of supraoccipital and parietal. Key to C: 1, supraoccipito-thecal suture; 2, petroso-exoccipital suture.

Precise positions of sutures on caudal aspect of skull notably vary among Paedotherium specimens because overgrowths (alae) from surrounding elements may develop in such a way that they produce internal or self-sutures, thereby falsely suggesting presence of separate elements (e.g., “mastoideo-posttympanic suture” between theca and petrosal in Hegetotherium mirabile AMNH-VP 9223; Sinclair [1909: 71, text-fig. 14]). In this specimen, probable position of lines of fusion among elements indicated by question marks. Externally, aperture of posttemporal canal actually lies at junction of three bone territories (petrosal, squamosal, and exoccipital), but how the aperture is framed in the adult varies. Petrosal also produces alae, which may overgrow the posttemporal dehiscence in such a way that in some specimens (as here) the aperture appears to be wholly encased by petrosal material. In pachyrukhines the paroccipital process is a broad flange that incorporates not only exoccipital material but also ectotympanic and petrosal. Size of individual contributions also varies (cf. figs. 9B, 12F, G). In AMNH-VP 45914, both ectotympanic and petrosal appear on lateral side. As usual there is no sutural or other identifiable boundary between supraoccipital and exoccipital. Key to D: 1, supraoccipito-thecal suture; 2, parieto-thecal suture (along squamosal bar); asterisk, ?lateral interparietal (see text).

Continued

Continued

Continued

FIG. 9.

Paedotherium MLP unnumbered2, stereopair views of isolated left auditory region (with keys) in lateral (A) and medial (B) aspects. Both to scale in A. Shading convention as in figure 8. Specimen not in life position (see fig. 8B for proper orientation). This specimen, evidently adult, is relatively complete except for loss of thecal spine (single asterisk in A) and damage to exoccipital. Many features can be identified that are otherwise very difficult to see in intact skulls. Sutural divisions can be detected where expected in most areas; however, squamosal and ectotympanic are smoothly continuous (probable position of boundary indicated by question marks). Ectotympanic probably forms most or all of meatal roof; entotympanic participation has not been confirmed in pachyrukhine bullae and is thus omitted (see text). Paroccipital process is largely missing; reconstructed outline based on AMNH-VP 45914 (fig. 8B; note differences in position of petrotympanoexoccipital suture in these two specimens). The numerous vascular foramina punctuating dorsal part of auditory bulla are of uncertain homology. In B, double asterisks identify major vascular grooves. Note rugose sutural surfaces for supraoccipital and parietal on medial aspect of theca. Large black arrow is situated in transverse dehiscence between tentorium and medial thecal wall, connecting transverse sulcus with posttemporal canal (not preserved). Note large, pillarlike petrosal crest, which contributes to tentorium.

Continued

FIG. 10.

Paedotherium sp. MLP 52-IX-28-29. A is virtual coronal section through external auditory canal, medial to right. B and C locate planes of sectioning for sections illustrated in figures 10–12 (exposed matrix colored gray). On skulls, ectotympanic material appears to contribute to dorsal wall of external acoustic meatus, although no suture between ectotympanic and squamosal theca can be identified in adult and actual situation therefore uncertain. Note incus (out of position) and prominent anular bridge. Asterisk ( ^* ) below bulla indicates fracture (not ectotympano-entotympanic suture).

In pachyrukhines the epitympanic sinus is so large that it bulges caudally as well as dorsally, and this has several morphological consequences. Virtual sections and microscopic examination of specimens establish that no bones other than the squamosal and petrotympanic directly face onto the airspace within the theca. However, the sutural picture is made complex by the growth of wings or alae from various elements that slightly overlap the swollen thecal walls as well as nearby structures (figs. 12B–E). This is particularly evident in the case of alae developed from the petrosal (or petrotympanic), producing a far greater caudal exposure of this element than is typical for notoungulates (cf. Cochilius, fig. 13; Billet, 2011).

On the dorsal aspect of the skull (fig. 8C, D), the theca enters into sutural relationships with the other bones in the caudal cranium (supraoccipital, parietal, and interparietal). As all relevant sutures are patent in the available material, it is certain that these other bones are separate from, and therefore do not contribute to, the thecal walls. Sutural contacts continue onto the caudal surface of the cranium, between the rear of the auditory region and divisions of the occipital (including the interparietal, which is partly fused to the supraoccipital; see Interparietal Complex). Endocranially (fig. 9B), the exoccipital is in lengthy sutural contact with the petrotympanic, and the medial and lateral ends of the tentorial process of the supraoccipital articulate with the petrosal crest (figs. 11D, 12C). All of these elements are firmly locked together but only weakly joined to the rest of the skull along the caudal perimeter of the ICJ.

Returning to the midcranial region, it may be noted that bone-to-bone contacts are interrupted by two lengthy gaps, denoted here as the piriform fenestra and the dorsal midcranial hiatus (fig. 8A, B). The mammalian piriform fenestra as defined by MacPhee (1981) is present in adult stages of many groups, although it varies greatly in size and conformation (Wible, 2007). In Paedotherium the outlet for the mandibular nerve (foramen ovale) is not separable osteologically from the larger piriform fenestra; this joint aperture is sometimes identified as the sphenotympanic fissure (Gabbert, 2004), but here their theoretically separate identities will be maintained. The fenestra is typically positioned, i.e., on the basicranial floor, lateral to the central stem and sandwiched between the auditory region caudally and preotic portions of the alisphenoid and squamosal rostrally.

The dorsal midcranial hiatus is the much larger vacuity situated on the dorsolateral aspect of the skull between the preotic flange and the squamosal bar (fig. 8B, D). Roth (1903) did not describe the hiatus as such, although it can be clearly made out in his illustration of the skull of Pachyrukhos (fig. 3). Significantly, he showed the hiatus as bounded dorsomedially by his “sutura squamoso-mastoidea,” the existence of which comprises his principal evidence for inferring that another, nonsquamosal element must cover the epitympanic theca. Although Roth doubtless had several pachyrukhines to examine, the one depicted in his plate is misleading: as may be seen by comparing figures 3 and 8D, the indicated suture is actually the one between the parietal and the theca, the squamosal bar having been lost in his specimen. (This may or may not be the same specimen as the one figured by Lydekker [1893: pl. 1], who thought the squamosal bar was part of the parietal. Simpson (1936) probably could not have fathomed any of this from the accounts of these authors, and these misidentifications must have added to his overall frustration.)

FIG. 11.

Paedotherium sp. MLP 52-IX-28-29, four virtual parasagittal sections, left auditory region, rostral to right. All to scale in B. Sections are in lateromedial sequence, beginning with slice through external acoustic meatus and progressing through medialmost part of the tympanic cavity and bullar wall (see locator diagram, fig. 10B). In A, single white asterisk, sulcus for transverse sinus, grooving medial aspect of theca; double white asterisks, ?vascular conduit of uncertain homology, traveling within mastoid canaliculus. The dorsal hiatal ridge appears twice in this section, curving across rostrolateral aspect of theca. In B, single black asterisk identifies adital margin of tegmen tympani. Tymp ano styloid has retained ontogenetic connection with tegmen tympani, despite sculpturing effects of massive middle ear pneumatization. In C, discontinuity seen in squamosal bar is a fracture, not a suture. Note small vertical septum running between tympanic floor and cochlear Promontorium. In D, white asterisks in transverse sinus and its tributaries. Note extensive fracturing of petrosal.

Continued

FIG. 12.

Paedotherium sp. MLP 52-IX-28-29, seven virtual horizontal sections on this and subsequent pages, rostral to right. All to scale in A. For orientation, rostral portion of cranium is shown in A but cropped out in succeeding sections. In A, note continuity between transverse sulcus and posttemporal canal (white asterisks) and fracture (br, not a suture) passing through squamosal bar. Sections B–G are in dorsoventral sequence, beginning with a plane passing slightly beneath cranial roof and progressing through occipital condyles. Contributions of major elements to tentorium osseum, thecal alae, and walls of the transverse sinus are easily distinguished. Note extensive caudolateral exposure of petrosal material, unusual in a notoungulate.

Continued

Continued

Despite Roth's (1903) inadequate descriptions, it is important to consider whether the dorsal hiatus, given its very large size in pachyrukhines, might have held an element, or part of an element, during life. This possibility is nominally suggested by the presence of the low, rugose, dorsal hiatal ridge on the rostral surface of the thecal wall (figs. 8B, 9A). The ridge is denticulated like a typical cranial sutural margin, although less so than (for example) the suture formed between the theca and supraoccipital (figs. 9B, 11D). It runs in a tight, downturning arc from the position of (and in the same convex plane as) the squamosal bar and ends facing on (and in the same convex plane as) the preotic ends of the alisphenoid and squamosal. The fact that these features are oriented in such a specific way strongly implies that something stretched between them in life, thereby shutting off the dorsal hiatus. There are several possibilities:

1. Closure by bone originally present. There is no indication in any of the pachyrukhine skulls examined that an independent ossicle was ever present in the space occupied by the midcranial hiatus, even though the caudal margin of the preotic flange is probably not entirely complete in any specimen examined. If during life the flange made contact with the ventral section of the hiatal ridge, the result would have been the closing off of a part, but certainly not all, of the hiatus.

In Lepus californicus the elongated caudoventral end of the interparietal projects under the squamosal bar and into the hiatus (fig. 18B), where it freely inserts between the squamosal and petrosal, effectively locking them together. Bramble (1989) called this projection the hiatal plate and interpreted it as another, albeit passive, component of the jackrabbit's ICJ: because it does not articulate suturally with the petrosal, it “does nothing to impede movement along the intracranial joint” (Bramble, 1989: 307). In wellpreserved pachyrukhine skulls the shelflike tentorial processes of the supraoccipital and parietal/ interparietal can be seen through the hiatal opening; they do not actually project into the latter, but do form a strong sutural bond with the petrosal crest in the tentorium osseum (figs. 11D, 12C). Rather than having anything to do with the hiatus as such, then, perhaps in both leporids and pachyrukhines the arrangement of these outgrowths acts to reinforce the bony framework of the caudal cranium. This interpretation does not, however, explain the presence or orientation of the organized array of bony edges and spicules that constitutes the dorsal hiatal ridge.

2. Undifferentiated dense connective tissue. Rather than by bone, the midcranial hiatus might have been closed by membrane, continuous with the primitive ectomeninx (cf. MacPhee, 1981). Such tissues cover the midcranial hiatus in leporids as well as the piriform fenestra in shrews and some other eulipotyphlans (Gasc, 1963; MacPhee, 1981; personal obs.). The appearance and radial disposition of the pachyrukhine dorsal hiatal ridge recalls the kind of marginal ectopic ossification that may occur at bony attachment points of dense connective tissues under significant, repeated stress. Muscle attachment to such tissues might produce sufficient stress to initiate the ossification of the latter; this is often a pathological process (Standring, 2009) but of course need not be (e.g., production of sagittal crest for temporalis m.).

A feature of possible relevance in this regard is the stout projection (thecal spine) on the rostral thecal wall, situated in an exposed position just above the hiatal ridge (figs. 8B–D, 12A) and often damaged or broken in fossils (fig. 9A). The spine is in line with and appears to be functionally related to the linea temporalis for the origin of temporalis m. Alternatively, it may have acted as an attachment for a well-developed pinnal extrinsic such as parietoauricularis m. or its equivalent—an attractive argument if pachyrukhines possessed large, heavy external ears. The spines prominence may have been due to the large number of muscle fibers arising from it, perhaps because there was insufficient area for a more rostral attachment. This might have been the case if the hiatus were exclusively covered by membrane.

3. Membranous port for blood vessels. In all notoungulates examined, the medial wall of the theca is deeply incised by the large sulcus for the transverse sinus (fig. 9B, 15B, 17C). This sinus and its various tributaries presumably drained to the jugular/occipital system of veins via several ports, including the posttemporal, jugular, and retroarticular foramina. In pachyrukhines, however, this last aperture is missing as such, because the preotic portion of the cranial wall is largely unossified. Apparently, the retroarticular sinus simply passed through the midcranial hiatus in order to join the jugular system, as suggested by the presence of truncated sulci that suddenly end on the immediately adjacent thecal wall. (Misinterpreting the relevant morphology, Sinclair [1909: 89] stated that he was unable to find a “postglenoid” foramen in Pachyrukhos, as all the skulls he examined displayed a “fracture” in the relevant area.) From one standpoint it could be said that pachyrukhines differed from other notoungulates in possessing a greatly enlarged, membranous retroarticular foramen (which may have carried an artery as well as a vein). However, hiatal width in pachyrukhines is surely much larger than any plausible diameter for transiting blood vessels, indicating that this cannot be the complete explanation for the gaps presence.

4. Rarefaction zone within the squamosal. Whatever the functional reasons for their appearance, the various fenestrations or rarefactions characteristic of leporid crania are not random in their disposition or patterning (Wible, 2007; Moss and Feliciano, 1977). This point applies not only to the specific elements affected, but also to features consistently lying within their territories (such as the large dorsal foramen consistently seen in the fenestrated nuchal area of Lepus californicus; fig. 18B). Fenestration in pachyrukhine cranial bones is more restrained, but it exists (see Sinclair, 1909: pl. X, fig. 1) and is likewise patterned. This raises the question whether the midcranial hiatus could be interpreted as a particularly large and continuous zone of orchestrated bone loss within the territory of the squamosal, without having to infer that its presence was due to the transmission of an outsized vessel. In this argument, absent the hiatus and with normal ossification patterns restored, the pachyrukhine cranial sidewall—including the epitympanic theca—would now be in seamless continuity with other squamosally derived material, just as in other notoungulates in which the hiatal opening is lacking. This explanation is attractive primarily because it solves the morphological paradox of inflating a large epitympanic sinus entirely from within the confines of a tiny bone territory (squamosal bar) situated far from the primordial location of the competent tissue apparently responsible for inducing middle ear expansion (epithelial lining of the developing cavum tympani; see Thompson and Tucker, 2013).

The first three interpretations of the morphological significance of the midcranial hiatus are uncontroversial in the sense that they do not require any unusual developmental processes, whatever their likelihood may be on other grounds. However, they presuppose a very small squamosal component in the theca, perhaps no larger than the area circumscribed by the hiatal ridge (fig. 9B). By contrast, the fourth interpretation asserts that the squamosal was actually of normal size and relations in pachyrukhines, but for unknown reasons developed an extensive rarefaction zone within its caudal portion.

FIG. 13.

Cochilius volens AMNH-VP 29651, stereopairs of left side of caudal cranium (with keys) in ventral (A), oblique ventrolateral (B), and dorsal (C) aspects, on this and subsequent pages. Shading convention as in fig. 8. Key to A and B: 1, squamoso-alisphenoid suture (preotic area); 2a/2b, squamoso-ectotympanic suture (incl. glaserian fissure; question mark indicates boundary in meatal area uncertain); 3, putative ectotympano-entotympanic suture; 4, exoccipito-ectotympanic suture; 5, petrotympano-squamosal suture (question mark indicates uncertain boundary); 6, squamoso-exoccipital suture (on paroccipital process). Key to C: 1, squamoso-parietal suture; 2, inferred squamoso-interparietal suture; 3, speculative reconstruction of parieto-interparietal suture (externally obliterated, actual confirmation unknown); 4, supraoccipito-(inter)parietal suture; 5, squamoso-zygomatic suture. In A, petrosal (asterisk) slightly exposed in basicapsular fissure due to deformation. Pars canalicularis of the petrosal contributes to floor of posttemporal canal, but as it is not exposed externally to any significant extent (see fig. 15D) it is not colored as such in key diagrams. Although entotympanic (ENT?) participation in bulla is considered highly likely, decisive evidence would have to come from early stages (see text). This specimen is well preserved, but it nonetheless exhibits several fractures that superficially mimic sutures. In C, feature a is situated in a location similar to the “sutura squamoso-serrialis” in Roth's (1903) illustration of a young Toxodon (fig. 2). Another break, nearly the mirror image of this one, occurs in the same position on the right side (not shown). The explanation for their presence may be taphonomic: large vascular channels (transverse sinus and tributaries, situated just beneath the cranial roof in this area) may increase the likelihood of postmortem fracturing (e.g., fig. 14B).

Continued

Continued

Another taxon that sheds light on the organization of the pachyrukhine caudal cranium is the kangaroo rat Dipodomys, which exhibits hypertrophied paratympanic spaces in combination with a highly reduced squamosal but lacks the morphological correlates of the ICJ. Thanks to Webster's (1975; see also Webster and Webster, 1975) exhaustive studies of ear region ontogeny in this heteromyid, there is no question about the individual contribution of bone territories to the formation of the epitympanicum (= epitympanic sinus/theca of this paper). In the kangaroo rat this volume is exclusively covered by the petrotympanic (both elements participating); the squamous squamosal is reduced to a bar (suprameatal spine of the squamosal; fig. 19A), which is applied against, but not fused to, the dorsal wall of the external acoustic meatus. The parietals, interparietals, and supraoccipital make no contribution to the walls of the epitympanicum (Beer, 1965).

The demonstrable separateness of the squamosal throughout ontogeny in Dipodomys is a key difference from Paedotherium, in which, as virtual sectioning shows (figs. 10–12), most of the theca is squamosally derived. This observation does not establish that the pachyrukhine hiatus is therefore due to rarefaction in the manner discussed above, but it does allow the presumption that this group differs from other notoungulates only in the degree of definitive squamosal development, and that in no regard is it necessary to invoke the existence of an independent serrialis to explain the origin of the thecal covering.

On this point comparison to heteromyids is again useful, for they likewise vary in the degree of squamosal participation in the cranial sidewall (Webster and Webster, 1975; Nikolai and Bramble, 1983). In heteromyines (e.g., Liomys), the squama of the squamosal is still small compared to that of most eutherians, but it is large enough to contribute to the cranial wall dorsal to the meatus. In dipodomyines and also in perognathines (Perognathus, Chaetodipus) this portion is always small, never amounting to much more than the tiny rod seen in Dipodomys. Interestingly, this difference is correlated with the scale of paratympanic pneumatization: in heteromyines inflation is relatively conservative, but in some dipodomyines, of which Dipodomys deserti is the outstanding example (Webster, 1975; Best et al., 1989), pneumatization is so extensive that the combined volume of the two middle ears exceeds that of the braincase.

Cochilius

Of the three interatherioid skulls submitted for CT scanning (Cochilius volvens AMNH-VP 29651, Progaleopithecus tournoueri AMNH-VP 29603, and Protypotherium sp. MLP 12-1280), only the first produced good results and therefore comments will be largely limited to this specimen.

Macroscopically, Cochilius volvens AMNH-VP 29651 (figs. 13–15) is solidly constructed and, like other interatherioids (fig. 16), lacks the large piriform fenestra and dorsal midcranial hiatus seen in pachyrukhine hegetotherioids. In volume the epitympanic sinuses are in fact quite large, but the outline of their thecae on the caudal cranium is subdued because of the extreme robustness of the temporal ridges (fig. 13C).

A number of cranial sutures in AMNH-VP 29651 are either open, or, if fused, still identifiable in section by indicia (fig. 14C). On the dorsal surface of the skull, the external suture between the squamous part of the squamosal and the parieto-interparietal is apparent and punctuated by large foramina for rami temporales (fig. 13C). Sutural relationships are as in notoungulates generally, in that there is no evidence of any atypical sutures or subdivisions of bone territories (but see Entotympanics and Septum Bullae). The squamosal and the rostral crus of the ectotympanic are fused, and the sutural line between the bulla and the petrosal has largely disappeared (although its trace is still identifiable; fig. 13A).

Virtual sectioning of the skull (figs. 14, 15) reveals that the thecal covering is separate from all other bone territories except the squamosal, which unequivocally forms most of the theca; and the petrotympanic, which is continuously fused with the thecal wall in a small zone around the aditus and along the medial wall of the notably inflated retrotympanic process (fig. 15B, C). Roth (1903) claimed that the “protuberancia mastoidea” was pneumatized in notoungulates, but this is probably not correct judging from conditions in Cochilius. In this taxon (fig. 15D) the caudal end of pars canalicularis presents large, trabeculated spaces, possibly for hematopoeic tissue, but these lack a connection with either the tympanic cavity or the epitympanic sinus. Unlike Paedotherium (fig. 12), alae are not significant and the petrotympanic has only limited exposure on the caudal aspect of the skull, where it helps to frame the aperture of the posttemporal canal. Protypotherium sp. MLP 12-2780 is apparently similar in these respects (fig. 17B, C).

FIG. 14.

Cochilius volens AMNH-VP 29651, three virtual parasagittal sections, left side, arranged in lateral to medial order on this and subsequent pages; rostral to right. All sections to scale in C. Diagram A indicates planes of sectioning in this figure and figure 15. B, section through external acoustic meatus; note incus in epitympanic sinus, out of position post mortem (also seen in fig. 15B). Although remodeling has erased most suture lines, still partly represented is suture between ectotympanic s rostral crus and preotic portion of squamosal (squamoso-ectotympanic suture, large white asterisk). Suture between caudal crus and retrotympanic process is no longer identifiable. Upper leader to mand fos (mandibular fossa) points to damage comparable to that seen in a similar place in Oldfieldthomasia AMNH-VP 28600 (feature 3 in fig. 1A); it is not evidence of a suture. C, section through towered tip of cochlear promontory. Simpson (1936) thought the small septum (small white asterisk) on rear wall of bulla conducted internal carotid in Oldfieldthomasia; in Cochilius this feature is better interpreted as conduit for tympanic nerve. D, section through medial aspect of middle ear; the part of the bulla articulating with petrosal is probably of entotympanic origin (see text). Rostralmost end of septum bullae in this section is confluent with bullar wall; more caudally it is distinct and connects with promontorium, as in other notoungulates in which this feature has been detected. Note long tongue of tentorial process of supraoccipital (tent pr soc) meeting tentorial process of parietal (tent pr par). Exoccipital and supraoccipital are completely fused, but the division between parietal and interparietal can still be identified.

Continued

FIG. 15.

Cochilius volens AMNH-VP 29651, four virtual coronal sections, left side (rev.), arranged in rostral to caudal order (for approximate locations of sectioning planes, see fig. 14A). All to scale in D. A, section through rostralmost end of bulla. Note fracturing (br) on medial surface of squamosal, which mimics a true suture. Arrow, discontinuity on rostral lip of bulla thought to represent nearly obliterated ectotympano-entotympanic suture. B passes through external acoustic meatus (note incus in epitympanic sinus, out of position post mortem). Arrow identifies caudal end of same groove seen in A. Except for petroso-bullar suture, other discontinuities are fractures. Asterisk identifies the large longitudinal fracture, seen on dorsal aspect of skull and positioned in roughly the same place as Roth's “sutura squamoso-serrialis” (see fig. 13C). Note that fracture intersects the very large transverse sinus, also longitudinally positioned. In C (section through caudal portion of epitympanic sinus and retrotympanic process), it appears that rearmost part of petrosal was pneumatized in concert with retrotympanic process; thus, sidewall of epitympanic sinus in this region is probably both squamosal and petrosal in origin. This should be distinguished from conditions in D: here caudal portion of pars canalicularis helps to frame part of posttemporal canal, but trabeculated space (asterisk) seen internally within “mastoid” is a marrow cavity, unconnected to middle ear. Caudalmost part of epitympanic theca appears suspended in space, separated from more medial parts of skull by relatively enormous posttemporal canal and sulcus for transverse sinus.

Continued

FIG. 16.

Protypotherium sp. MLP 12-2780, caudal cranium in left lateral (A), dorsal (B), and caudal (C) aspects. All to scale in C. Caudal cranium of Protypotherium strongly resembles that of close relative Cochilius (fig. 13). Both of these interatherioids possessed large epitympanic sinuses, although thecal prominence is relatively less than in pachyrukhines. Few sutures can be identified on the caudal cranium of this particular specimen, but there is no reason to think that it differed from other notoungulates in sutural content or element composition. Note broken areas above posttemporal canal. Large size of foramina for rami temporales is noteworthy (size varies within taxon; cf. Sinclair, 1909, pl. III, fig. 2).

Continued

In summary, the caudal cranium of adult Cochilius bears no evidence of a serrialis or unusual sutural relationships suggestive of additional elements. The epitympanic theca is largely formed by the squamosal, with small contributions from the compound petrotympanic. Because of crushing it is not clear whether the rostral crus of the ectotympanic contributes to the lateral part of the meatal roof. The caudal crus definitely does not, as there is a marked fissure (retrotympanic fissure; fig. 13B) between the two. As far as may be ascertained this last statement applies to other interatherioids as well.

Oldfieldthomasia

As already noted, Oldfieldthomasia AMNH-VP 28600 was not a happy choice for Simpson's (1936) purpose, not only because of its general condition, but because it was not one of the taxa that Roth (1903) specifically noted as possessing the serrialis. Despite its greater geological age compared to that of most other notoungulates known to Simpson in the early 20th century, in its caudal cranial construction Oldfieldthomasia cannot be described as especially primitive (as compared, for example, to the definitely more primitive henricosborniid Simpsonotus [Late Paleocene; see Billet, 2011]).

Although Simpson (1936) maintained in his text that the thecal walls (his “pars epitympanica”) of Oldfieldthomasia were squamosal, some section drawings show apparent continuity between thecal walls and bone that is surely ectotympanic in origin; others show a connection with his putative adventitious elements (see The X Elements). However, ignoring these, there is no reason to doubt that the squamosal contributed most of the thecal covering in Oldfieldthomasia, with the petrotympanic contributing modestly to the adital area where bulla, theca, and tegmen tympani anatomically meet and where intense remodeling probably first began during ontogeny. Evidence for a serrialis is accordingly completely lacking.

Entotympanic Presence in Cochilius volvens AMNH-VP 29651

I looked for signs of entotympanic participation in the best-preserved notoungulate skulls in the AMNH, MLP, and MACN paleontological collections but found only one specimen—Cochilius volvens AMNH-VP 29651—for which a reasonably compelling argument can be made (figs. 13A, B). Interestingly, this specimen was apparently fully mature at death according to Simpson (1932b), who compared it to another specimen (AMNH-VP 29657) of the same species in which milk teeth were present.

In Cochilius AMNH-VP 29651 the tympanic floor in ventral aspect consists of three morphological portions: a rather flattened medial section, a swollen central part, and, completing the lateral region, the canal of the external acoustic meatus. Emphasizing the division between the two first-named portions is a shallow groove, running more or less parasagittally (feature 3 in fig. 13A, B). Externally, the groove extends from the rim of the aperture for the auditory tube to the intersection of the bulla with the paroccipital process, where it disappears from view. Grooves on the right and left sides are symmetrically disposed, unlike most of the small thrust fractures or hairline breaks seen elsewhere on the skull. Bullar surface textures on either side of the groove differ slightly, and serve to divide the larger, rather smooth-walled and well-inflated central portion from the slightly rugose, less-inflated medial section. Although the groove and surface texture differences can scarcely be described as conspicuous, they were evidently obvious enough to Simpson's artist, who suggested them with contour lines on the ventral surface of the skull (cf. Simpson, 1932b: fig. 2). Simpson (1932b: 6) himself, however, noted only that “in this apparently fully adult specimen all sutures are still open and the bullae not fused with surrounding elements.”

In coronal section (fig. 15B) the difference in inflation between the medial and central portions is quite noticeable, the changeover being marked by a distinct “knee” or inflection on the internal (tympanic) surface of the bulla. Here the external groove, which follows the trend of the inflection, is barely identifiable as a slight dimple on the bullar wall except in relation to the paroccipital process, near which it becomes more prominent (arrow in fig. 15B). However, there is only one place—the margin of the aperture for the auditory tube—where a discontinuity interpretable as an unfused sutural remnant can be convincingly identified (fig. 15A). Small disjunctions in the caudal bullar wall also exist, but all seem better interpreted as artificial breaks. The only other suture related to the tympanic floor lies as expected between the bullar floor and the otic capsule (figs. 15B, C). It is still patent and can be continuously traced in virtual sections along the entire contact zone between these bone territories.

A final point of interest is that Cochilius AMNH-VP 29651 possessed a septum (fig. 14C) that corresponds in location to the “vertical” septum bullae identified in hegetotherioids by Patterson (1936; see also Billet, 2011). The septum runs from the rostral pole of the promontory to the bullar wall, creating thereby a small rostromedial diverticulum adjacent to the bony channel for the auditory tube. There is another, smaller septum (fig. 14B) in the position of the one that Simpson (1936) thought conducted the internal carotid into the tympanic cavity in Oldfieldthomasia, but in both taxa the structure in question and its associated foramen (fig. 14C) probably carried the tympanic nerve, not the internal carotid (cf. inferences similar to Simpson's by García López [2011] and Billet and de Muizon [2013]).

Entotympanic Presence in Other Notoungulates

Conditions in other well-preserved specimens available for study were equivocal. The tympanic floors of another specimen of Cochilius (MLP 2339, ex coll. Lillo) displayed slight textural differences, similar to that seen in AMNH 29651, but clear evidence of the groove was lacking. Interatherium (= Icochilius) extensus MACN 9738-39 and Interatherium rodens MACN 9857-58 exhibited bullae conformationally much like that of Cochilius volvens, but no suture line or textural difference was observable. In Progaleopithecus AMNH-VP 29603 there is no evidence, in the tubal region or elsewhere, of an entotympanic. Scans of Notopithecus adapinus AMNH-VP 28949, despite the very poor preservation of cranial architecture in this specimen, show a small rostral septum like that of Cochilius but no unequivocal evidence of a suture.

On relatively intact skulls of Mesotherium pachygnathum (AMNH-VP 14509) and Hegetotherium andinum (MLP 12-2194a) there was again a tiny fold on the rim of the tubal aperture, suggestive of a suture, but the fold did not extend caudally as a groove. Skulls of Pachyrukhos (e.g., MACN A1251-52) sometimes displayed surface differences on their bullar walls, similar to those observed in Cochilius AMNH-VP 29651, but once again distinct grooves were not present. Although in very good material sites of suspected sutural fusion may be detected by reference to histological features of the bone matrix (MacPhee and Cartmill, 1986), as already mentioned the resolution of conventional micro-CT instruments is generally inadequate for such a purpose.

With regard to the septum bullae, Patterson (1936) stipulated that this structure was absent in two interatherioids he studied (Interatherium and Protypotherium), although “vestiges” might remain. Although often discriminated in character analyses, I suspect that such “vestiges” are probably universal within notoungulates, but scored only when well developed. Protypotherium MLP 12-2780 exhibits a double medial bullar wall (fig. 17B), but this is not evidence of an entotympanic as such. Instead, this appearance is due to the expansion of a paratympanic space (medial paratympanic cavity) that inflated the entire medial bullar wall up to the point where the bulla meets the paroccipital process (which the cavity did not invade). Its aditus is a small aperture on the medial wall of the bony channel for the auditory tube (not shown). Extension of the middle ear cavity into the substance of the bulla in this manner is not rare in mammals generally (MacPhee, 1981), although this appears to be the first example to be reported in a meridiungulate.

Finally, as Billet (2011) has noted, the saliency of the crista tympanica within the tympanic cavity varies, projecting medially in some notoungulates taxa as a wide, curved shelf (e.g., Paedotherium, fig. 10A). The presence of this shelf (anular bridge) has nothing to do with the occurrence of an entotympanic, nor does it represent yet another novel element (a separately ossified “anulus membrane”). Its form and width are simply a function of the amount of remodeling activity occurring in the subtympanic recess, as has been documented for various primates and rodents in which a similar feature occurs (MacPhee and Cartmill, 1986; MacPhee, 2011).

In summary, the symmetrical, parasagittal grooves on the bullar floors of Cochilius volvens AMNH-VP 29651 can be plausibly interpreted as marking the position of an ectotympano-entotympanic suture in this taxon. In other specimens such grooving is absent, possibly because it was normally obliterated during ontogenetic remodeling of the bullar wall. Symmetrical presence of grooves is not an absolute guarantee that postmortem crushing can be excluded as an explanation (cf. “symmetrical” cracks in dorsum of caudal cranium of same specimen, fig. 13C), but it seems unlikely in this case because virtual sectioning shows that the grooves are just that, rather than breaks passing entirely through the bullar bone. Although AMNH-VP 29651 is but a single example, it offers about as clear a case of ectotympanic/entotympanic fusion in an adult fossil eutherian as I have seen, and permits the reasonable conclusion that Cochilius possessed a compound bulla made up of both elements. In this interpretation, the central and lateral portions of the tympanic floor are certainly ectotympanic because they include the crista tympanica; the striplike medial part of the bulla must be derived from entotympanic material—practically by definition, as all other basicranial constituents can be accounted for.