The Problem of Antillean Heptaxodontidae

Depending on author and definition, as many as six—or as few as two—species comprise the West Indian membership of Heptaxodontidae. With one or two very dubious exceptions, all known fossils attributed to these species come from Quaternary contexts, although their history on the islands doubtless goes back much further (MacPhee, 2009). In rough order of naming and known range, currently accepted species are: Amblyrhiza inundata (Cope, 1868; Anguilla/St Martin, or Anguillea1), Elasmodontomys obliquus (Anthony, 1916; Puerto Rico), Quemisia gravis (Miller, 1929; Hispaniola), Clidomys osborni (Anthony, 1920; see also G. Morgan and Wilkins, 2003; Jamaica), Xaymaca fulvopulvis (MacPhee and Flemming, 2003; Jamaica), and, possibly, Tainotherium valei (Turvey et al., 2006; Puerto Rico). The first four have long been known to grossly resemble one another in exhibiting multilamellar or “platetooth” cheektooth organization. The affinities of the remaining two taxa, based respectively on a single, nearly edentulous jaw (MacPhee and Flemming, 2003) and a single, somewhat damaged femur not formally assigned to any family (Turvey et al., 2006), may be regarded as particularly unsettled. It is pertinent to mention that, although most platetooth taxa are founded more or less exclusively on teeth, the dentition tends to be a poor guide to higher-level relationships within Caviomorpha because of rampant convergence in dental characters (Landry, 1957: 52; see also Patterson and Wood, 1982; Woods, 1989a, 1990).

The nomina Heptaxodontinae/Heptaxodontidae have a confusing nomenclatural history. The nominotypical genus, Heptaxodon Anthony, 1917, is based on specimens that were shown by Stehlin and Schaub (1951), and in greater detail by Ray (1964), to be juveniles of Elasmodontomys Anthony, 1916. Anthony (1917: 186), who thought that in Heptaxodon bidens he had discovered a member of a wholly new and unusual group of caviomorphs, referred his monotypic subfamily Heptaxodontinae to Chinchillidae, together with the equally monotypic Amblyrhizinae and Elasmodontomyinae, on the ground that “[i]t is expecting too much of parallelism to so consistently develop identical tooth structure in three hystricine [i.e., hystricognath] genera.” (Even though Heptaxodon is now regarded as a junior synonym of Elasmodontomys, as Ray [1964, 1965] originally noted Heptaxodontinae remains valid as a family-group name under ICZN Art. 40.1 [synonymy of the type genus, validity of family-group names not affected].)

Anthony's conclusions, although very briefly stated, established the tone for much later thought on the relationships of these taxa to one another and thence to other caviomorphs (see also Anthony, 1926, 1940). When Clidomys and Quemisia were discovered in the 1920s, their general similarity to Elasmodontomys in dental features suggested to some authorities (e.g., Kraglievich, 1926; Landry, 1957) that they could all be accommodated in one phylogenetically and biogeographically cohesive package, together with a number of mid-Cenozoic South American taxa that also exhibited multilamellar molars and large body size. Simpson (1945) followed this line of reasoning, brigading the Antillean taxa (as Heptaxodontinae) with the mainland South American groups Potamarchinae, Eumegamyinae, and Neoepibleminae, and placing the entire lot under the banner of Heptaxodontidae, superfamily Cavioidea. Citing Kraglievich (1926, 1932) as authority, Simpson (1945: 212) noted that “the once very abundant, wide-spread, and varied genera here united as †Heptaxodontidae” constituted a holophyletic group, which in his concept meant that they were “related to one another and not referable, as is often done, to the Dinomyidae or other recent families.” His reasoning for sharply separating heptaxodontids from dinomyids was not given, although in his classification they were placed adjacent to each other. Some later students of the group took a different stand (e.g., Fields, 1957), arguing that some of Simpson's heptaxodontids (e.g., Tetrastylus) could even be viewed as standing in the direct line of descent of Dinomys. Ray (1965) also reviewed the case for the dinomyid affinities of Amblyrhiza and Elasmodontomys.

Recent views have been more diverse, but with little character evidence in play other than that provided by the dentition, systematic assignments have remained weakly founded. Thus, Woods (1989a, 1990; see also Woods and Kilpatrick, 2005) proposed to split the Antillean taxa (his “giant hutias”) into two groups, Heptaxodontinae and Clidomyinae (see also Pascual et al., 1990). His concept of Heptaxodontinae included Quemisia, Elasmodontomys, and Amblyrhiza, while Clidomyinae was reserved for Clidomys alone. Earlier, Woods (1982), expanding on a view also espoused by Ray (1964, 1965), thought of possibly placing heptaxodontines and capromyids together in a separate monophyletic group, but in later papers he abandoned this conclusion and simply stated that the heptaxodontines and capromyids were very closely related, possibly sharing an unspecified common ancestor that probably lived in the West Indies in the Miocene (Woods, 1989a; Woods and Kilpatrick, 2005). Where the ancestors of that ancestor came from is unstated; in his various publications, Woods only tangentially considered Kraglievich's (1926) perceptive argument that Antillean heptaxodontids might be more closely related to certain of the large-bodied rodents of the Patagonian Neogene than to any other group. Patterson and Wood (1982) essentially threw up their hands: they were not willing to commit to any specific affiliation for Heptaxodontidae, leaving this taxon as incertae sedis as to superfamily—the only nominal family of caviomorphs that they treated in this manner.

In their classification McKenna and Bell (1997) virtually emptied Heptaxodontidae sensu Simpson and redistributed its parts among all three of the standard superfamilies. Thus, Potamarchinae and Eumegamyinae (as subfamilies within their concept of Dinomyidae) were left in Cavioidea, while Neoepibleminae (now as a family) was moved to Chinchilloidea. Heptaxodontidae proper was rotated into superfamily Octodontoidea ( =  Octodontidae, Myocastor, Echimyidae, and Capromyidae2) and its content reduced to Elasmodontomys, Amblyrhiza, Clidomys, and the poorly known Patagonian taxa Tetrastylomys and Pentastylomys. This association recalls Kraglievich's (1926, 1932) scheme in part, but seems otherwise to be their own. The inclusion of heptaxodontids within Octodontoidea echoes Landry's (1957) conclusion, later reinforced by Woods (1982), that Capromyidae—regarded as octodontoids in all recent classifications—is the sister group of heptaxodontids. More recently, Negri and Ferigolo (1999) have relimited Neoepiblemidae to include Neoepiblema, Phoberomys (a eumegamyine according to McKenna and Bell [1997]), and Eusigmomys (a potamarchine according to McKenna and Bell [1997]). Classificatory shufflings of this magnitude may well be warranted, but they also indicate that consensus on the phylogenetic placement of these rodents has not been reached.

Comparative Set

The comparative set (appendix 1) assembled for this study includes representatives of all currently recognized ctenohystrican families having extant representation (i.e., Ctenodactylidae + all families traditionally placed in Hystricognathi; see Huchon and Douzery, 2001; Huchon et al., 2007; Churakov et al., 2010). Although the concept that Caviomorpha is strictly monophyletic has been challenged from time to time, at present the possibility of paraphyly seems remote and I do not deal with it further here (Nedbal et al., 1994; Huchon et al., 2007; Carleton and Musser, 2005; but see Jenkins et al., 2005; Coster et al., 2010).

As already illustrated, there is a continuing flux in ctenohystrican systematics regarding sister-group relationships, organization of superfamilies, and the positioning of a number of fossil taxa, and doubtless this grouping will undergo further internal changes with the application of new data sources and new methodologies. However, to ease the need to use a plethora of possibly evanescent names, the supergeneric entities presented in this paper generally follow the most recent version of the classification of Ctenohystrica as provided by the Paleontology Database (currently housed at  http://museumu03.museumwww.naturkundemuseum-berlin.de/cgi-bin/bridge.pl), with modifications where noted. For simplicity I maintain the conventional division of extant caviomorphs into the four superfamilies Erethizontoidea, Cavioidea, Chinchilloidea, and Octodontoidea. The hystricognaths of Eurasia and Africa present a more complex picture; at present I follow the convention of simply grouping Bathyergidae, Thyronomyidae, Hystrichidae, Petromuridae, and Diatomyidae under the banner of Phiomorpha. Ctenodactylidae is traditionally left outside this scheme, as here, within its own superfamily.

The total number of specimens per taxon examined (see appendix 1) is necessarily small because whenever possible I selected “no data” skeletons and skulls with broken basicranial regions for study and dissection. Notes on the dissection of the infratemporal fossa and sidewall of the auditory region of Dinomys branickii AMNHM 201638 are presented separately (see Original Observations).

With respect to the basicrania of fossil dinomyids and other extinct chinchilloids, which are of special interest here, published observations of any sort are rare and do not shed much light on the interpretation of the specific structures of interest here. For example, in regard to the feature here named the tympanic fenestra, Fields (1957: 282) notes for the La Venta taxa Olenopsis and Scleromys only that the “accessory rounded ventral opening [is] present as in Dinomys.” Negri and Ferigolo (1999: fig. 14) mention the presence of another feature, an apparent posttympanic foramen, on an excellent skull of Neoepiblema ambrosettianus, but they allude to it as a supernumary “forame para um dos ramos do nervo facial (VII)?”—an occurrence otherwise unknown in rodents. Vucetich (1975), who described middle ear features of the related genus Perimys incavatus, noted the presence of the tympanic fenestra as well as some other systematically pertinent basicranial features. The partial skeleton of Late Miocene Phoberomys pattersoni, a truly giant megafaunal taxon recently analyzed by Horovitz et al. (2006; see also Sánchez-Villagra et al., 2003), includes cranial material. However, the basicranium is not in a condition in which it can be profitably studied (M. Sánchez-Villagra, personal commun.). The related, and even larger, species Josephoartigasia monesi (Rinderknecht and Blanco, 2008) is represented by a skull, but no description of the auditory region has yet been published. Fortunately, during the course of this study an opportunity arose to study the basicranium of MLP 41-XII-13-237 (figs. 9–11), a well-preserved partial skull of the Pliocene dinomyid Eumegamys ( =  Eumegamysops; Alvarez, 1947; Fields, 1957). While not a substitute for the much larger study required of Neogene chinchilloids in general, it is hoped that the evaluations presented here will lay a proper basis for further work and give some indication of what may be accomplished with appropriate material.

Most anatomical structures and ontogenetic terms mentioned in the text are defined in MacPhee (1981), which should be consulted for details not covered here. For accessing general mammalian cranial morphology or finding answers to specific queries, also helpful are databases and published papers viewable on the Morphobank site ( www. morphobank.org). MacClade 4 (Maddison and Maddison, 2003) was utilized for the character optimizations presented in figures 27–Fig. 28.29.

Hypodigm of Amblyrhiza inundata

AMNHVP and AAHS hold almost all of the available material of this species, whose distribution was limited to the major islands of the Anguilla Bank (Anguilla and St. Martin). Naturalis (National Natural History Museum, Leiden) also possesses a small number of Amblyrhiza fossils, as detailed by Schreuder (1933; see also van der Geer et al., 2010). AMNHVP specimens largely derive from Edward Drinker Cope's personal paleontological collection, which was purchased by the AMNH in the 1890s. Cope's Amblyrhiza specimens were, in turn, originally collected for him by H. van Rigersma, a Dutch colonial physician who for many years provided interested scholars in the United States and Europe with natural history specimens from several of the northern Lesser Antilles (Cope, 1883; McFarlane and MacPhee, 1989). The AMNH mammalogist Harold E. Anthony, who visited St. Martin and Anguilla in 1926 for the express purpose of collecting additional remains of Amblyrhiza, recovered a number of specimens for this museum in cave localities on both islands. Although he never published on these fossils, his fieldnotes (Anthony, ms.) are available in the Mammalogy archives. The AAHS collection was amassed between 1988 and 1995 in cooperation with Donald McFarlane (Claremont McKenna College) and various AAHS members (for locality and other information, see McFarlane and MacPhee, 1989; Biknevicus et al., 1993).

Although reasonably good specimens of the jaw, rostrum, palate, auditory region, and many isolated teeth are available for study, there are no intact skulls. Much the same applies to the postcranium: although proximal and distal ends of most long bones exist, there are no complete specimens. Some of the long bones in Cope's original collection are broken in a way that suggests that they may have been intact when recovered during phosphate mining operations, but were poorly handled thereafter.

Hypodigm of Elasmodontomys obliquus

Most of the known specimens of this species are housed in the AMNHVP, with smaller collections in the Florida Museum of Natural History, Museum of Comparative Zoology at Harvard, and elsewhere. Remains of this rodent were first encountered in 1915, during the course of the New York Academy of Sciences' wide-ranging scientific survey of Puerto Rico and the Virgin Islands (Woods, 1996). During the following year, H.E. Anthony made prodigious collections in caves in the western and central parts of the island; virtually all of the AMNHVP specimens of Elasmodontomys derive from his fieldwork (Anthony, 1918; Woods, 1996). Since then few additional specimens have come to light, although with renewed interest in active collecting of extinct West Indian mammals (e.g., Turvey et al., 2006, 2007), major new discoveries of material will doubtless occur. The hypodigm for this species is extensive and, unlike that for Amblyrhiza, contains large numbers of complete specimens, including postcranials and skulls.

Character Analysis

Table 2 provides a list of characters and recognition criteria for discriminating among the character states of interest here. Character states have been formulated on the basis of the morphological observations provided under Original Observations (see below), and are referenced in the text by a consistent convention. “C1:2”, for example, refers to “Character 1: Absence/presence of tympanic fenestra,” with character state 2, “tympanic fenestra present as large dehiscence, isolated from meatus by suture-delimited area” determined on the basis of associated recognition criteria. Consult figures 27–Fig. 28.29 to see how character distributions map onto a predetermined scaffold tree. Although I have framed the interpretation of character states in a developmental framework where feasible, I emphasize that all character states should be understood to refer to adult or definitive conditions.

Table 2

List of Basicranial Characters and States

Incidence and Ontogeny of the Tympanic fenestra

Although the structure identified here as the tympanic fenestra is described under a variety of names in the mammalian morphological literature, its anatomy and development have been traced in detail only in Homo sapiens (most frequently under the names foramen tympanicum or foramen of Huschke; e.g., Wang et al., 1991; Lacout et al., 2005). Gacek (1975: 363), who gave a brief description of this feature in adult Cavia, describes it as the “ventral tympanic canal” (cf. “ventral accessory opening” of Fields [1957]; “segundo meato alargado” of Vucetich [1975]). However, I prefer tympanic fenestra because this dehiscence rarely resembles a typically rounded foramen or canal. Although it is unknown whether the persistent tympanic fenestra affects audition in any taxon that exhibits it, it is noteworthy that this feature is typically situated directly across from the umbo, site of the tympanic membrane's maximum amplitude.

Viewed over the timescale of ontogeny, the part of the fetal mammalian ectotympanic that bears the tympanic membrane (crista tympani) stops growing relatively early, but in most mammals the inner and outer margins of the original crura continue to expand radially along the fibrous membrane, albeit to varying degrees in different clades (MacPhee, 1981; MacPhee and Cartmill, 1986). In all mammals, lateral growth frames the bony external acoustic meatus more or less completely; in rodents and many other groups, growth along the ectoympanic's medial aspect also produces most or all of the auditory bulla. The ectotympanic always ossifies as a membrane bone in mammals and never exhibits primary or secondary cartilage (MacPhee, 1981). Ossification defects like the tympanic fenestra therefore cannot be ascribed to imperfect replacement of a cartilaginous preformation.

The conventional description of fenestral formation in the human specifies that, during the first year of life, the rostral and caudal crura of the U-shaped ectoympanic send out platelike expansions that meet centrally to form the floor of the developing external acoustic meatus and tympanic cavity (Anson et al., 1955). Typically, a small, unossified zone remains for a time between these expansions. The gap usually disappears by the age of five, but a persisting dehiscence is not uncommon. Lacout et al. (2005) found a patent hole in 4.5% of their sample of 103 temporal bone examinations.

Lacout et al. (2005) also state that the fenestra of Homo does not transmit anything at any stage of development. With regard to large neurovascular structures, this also appears to be the case in ctenohystricans. However, in a number of taxa in this group, fine vascular perforations occur within the same circumscribed region on the lateral bullar wall as the patent tympanic fenestra found in other members of this clade. This suggests that the vessels in question are probably a reasonably constant feature of lateral wall development. Whether or not they are eventually entrapped in bone depends on the nature of final ectotympanic development in this region. On incompletely cleaned skulls, dried vasculature may often be found in the immediate vicinity of the meatus and outer rim of the tympanic membrane. In Homo, these areas are fed by branches from the posterior auricular, superficial temporal, and maxillary arteries; as it is not evident which of these major trunks send emissaries to the lateral bullar wall in ctenohystricans, here they will simply be designated as meatal innominate vessels. They are not homologous with the ramus posttympanicus (see below), which, when definitely present, appears to supply a different area, one situated well within the middle ear cavity.

I emphasize that, although it is often continuous with the porus meatus, the fenestra is not an empty gap. In life it always seems to be occluded or walled off by a plug of meatal soft tissues ultimately derived from the fibrous membrane of the tympanic cavity (or the membranous meatus, with which the fibrous membrane is continuous). In several incompletely cleaned osteological specimens utilized for this study, the plug displays—unsurprisingly, given its continuity with the meatal lining—structures suggestive of glandular tissue (Gacek, 1975; cf. fig. 25, Dasyprocta punctata AMNHM 41394).

If every occurrence of an aperture in the lateral bullar wall were counted as a tympanic fenestra, character discrimination would be pointless. Accordingly, cases in which the tympanic fenestra is reduced to a single, rounded vascular foramen for innominate vessels (e.g., Myocastor coypus AMNHM 80097; fig. 22A) are scored the same as complete absence (C1:0). (Although the point is of no direct pertinence here, in this scheme Homo would also be scored as C1:0.)

Although obliteration of the tympanic fenestra is a normal aspect of temporal bone ontogeny in Homo, its incidence in the ontogeny of most other mammals is simply unknown. Part of the reason for this is that previous investigators have noted the presence of a fenestra only in instances in which it was comparatively large. Van der Klaauw (1931), for example, described finding it as an opening or ventral cleft in the meatal margin only in certain members of extant Herpestinae, Canidae, and Caviomorpha. Reports of this aperture in fossil mammals are even rarer; in one of the few instances on record, Simpson (1967: 222) described what appears to be a patent fenestra (as a “deep narrow notch”) in the meatal floor of the trigonostylopid astrapothere Trigonostylops wortmani.

In the specific case of caviomorphs, van der Klaauw (1931) documented a certain amount of variation in fenestral shape. The fenestra often has the form of an irregular gap (distal expansion of this paper) at the end of a narrow slit (ventral cleft of this paper) in the lateral bullar wall, but the expression of these features varies. Thus, van der Klaauw (1931) noted that Hydrochoerus, Dasyprocta, and Dinomys display a prominent ventral cleft that extends through the meatal lip, while in Cavia, Lagostomus, Chinchilla, and Lagidium “both margins of the incision are connected distally, thus forming an aperture” (van der Klauuw, 1931: 161) or an external acoustic meatus “doble con los orificios separados” (Vucetich, 1975: 488). In Myocastor, “only a notch (Einschnitt) in the border of the meatus is found, which by further growth is closed by the union of the margins [of the ectotympanic]. The development is much the same in man.” Although his wording is not explicit, van der Klaauw appears to have concluded that ectotympanic development was actually rather similar in all of these cases, the only differences being in the details. In one sense this is correct, but, as we shall see, variation is complex enough to warrant the discrimination of at least three character states (table 2). In evaluating these states, a certain amount of emphasis is placed on how the margins of the dehiscence mature. The definitive condition of the meatal area is, of course, achieved just as it is elsewhere on the bulla, i.e., by the deposition of bone matrix along existing bone margins on the scaffolding provided by the fibrous membrane. This is fully analogous to the closure of cranial fontanelles by mineralization along the primitive ectomeninx (MacPhee, 1981). Typically, however, approaching bone territories on the cranial vault eventually meet and form sutural tissues; in the case of the closure of the tympanic fenestra, sutural formation is seen much more rarely, probably because the single element involved (the ectotympanic) forms an internal suture only under special conditions. Unfortunately, very young specimens with largely unossified skulls are rarely found in traditional osteological collections, and thus the earliest phases of ectotympanic development in the taxa described here remain partly conjectural.

Posttympanic Foramen, Posttympanic Canal, and Ramus Posttympanicus

The posttympanic foramen and canal (C2, C3) are recognized as anatomical entities for the first time in this paper. Only four taxa—Amblyrhiza inundata (figs. 1, 4), Dinomys branickii (figs. 12–14), and the extant erethizontids Coendu mexicanus and Erethizon dorsatum (fig. 19)—in the comparative set were found to exhibit both the external aperture (posttympanic foramen, C2:1) and the intratympanic conduit related to it (posttympanic canal or septum, C3:1). (The foramen is present in Eumegamys paranensis, but the canal could not be identified with certainty.) The posttympanic foramen is situated on the external bullar wall, slightly posteroventral to the stylomastoid foramen in Dinomys. The posttympanic canal is a large, discrete tube or, sometimes, a grooved intratympanic septum that runs in close relation to the bony frame encompassing the tympanic membrane, to terminate in the vicinity of the stapedius fossa or cochlear windows. That these structures transmit an artery (here named the ramus posttympanicus) is conclusive for extant Dinomys (see Original Observations) but inferential for extinct taxa such as Amblyrhiza and Eumegamys. The posttympanic canal should not be confused with the canaliculus chordae tympani (MacPhee, 1981); the latter nerve also passes freely into the rear part of the tympanic cavity, but then wends its way through the ossicular chain, unlike the posttympanic canal. Consideration of the possible homologies of the ramus posttympanicus is left for the last part of this paper.

Some extant members of Chinchillidae were also found to exhibit a probable posttympanic foramen, in or near the stylomastoid foramen (C2:2), but because they lacked an identifiable posttympanic canal within the middle ear, the presence and destination of the vessel transiting the foramen must remain somewhat uncertain until dissection evidence is available. The true stylomastoid artery may be implicated here instead.

The posttympanic canal is clearly vascular and therefore not the homolog of the meato-cochlear bridge formed by tympanohyal-mastoid fusion, as described by Meng (1990) for the Eocene North American protrogomorph Reithroparamys. However, because the true posttympanic foramen is normally situated close to the base of the tympanohyal, it could easily be (and perhaps has been) misinterpreted as a sheath for that structure.

Notice is also made of the occurrence of pessuli (C4),3 or small struts of bone within the obturator foramen of the stapes that are interpreted by some authors (e.g., Guthrie, 1963) as homologous with the bony stapedial canal seen in some primates and a variety of other, mostly small, mammals. Pessuli are thus relevant to the question whether parts of the primitive stapedial system persist in hystricognaths (see appendix 2).

Basicranial Morphology of Amblyrhiza inundata

Two specimens of this species preserve the auditory region in substantially complete form. The first, AAHS 95044, is an isolated left tympanopetrosal from Tanglewood Cave, Anguilla, collected by Don McFarlane and the author (see McFarlane and MacPhee, 1989). The specimen includes the floor of the ectotympanic bulla but lacks the lateral parts of the external acoustic canal and related structures. The tympanic floor was intact at the time of discovery, but it has since been dissected (by carving a “window” into it) to facilitate the present investigation (figs. 1–3). The other specimen, AMNHVP 11842, is a basicranial fragment consisting of the left petrosal, exoccipital, and a portion of the ectoympanic bulla with the meatal region preserved (figs. 4, 5). This fossil, part of the AMNHVP Cope collection, lacks a stated provenance but was presumably collected by van Rigersma on either St. Martin or Anguilla. It is significantly larger than AAHS 95044 and somewhat differently proportioned, suggesting that the former specimen may have belonged to a juvenile animal. Or the difference may be a consequence of significant variability in body size in this species, evidence for which is discussed in detail by Biknevicius et al. (1993). Although for the purposes of this paper the two specimens were photographed in broadly similar aspects, slightly different viewing angles were chosen so that complex features could be seen in at least one ear region.

In ventral aspect the banana-shaped bulla of Amblyrhiza presents three projections. The most prominent is the external acoustic canal, which in AMNHVP 11842 is drawn out into a long tube that (when intact) would have projected upward and slightly backward (fig. 4). The other two projections (anteromedial and anterolateral processes) are simple pneumatic evaginations of the rostral end of the bulla. The more medial of the two may have articulated with the pterygoid or possibly served as an attachment point for palatal musculature (tensor veli palatini muscle). There is no indication that the lateral margin of the porus meatus supported accessory bones of the sort found in Cavia (van der Klaauw, 1931; Gacek, 1975), although the available material is not well enough preserved to decide this point. In the more complete specimen, AMNHVP 11842, both the mastoid and occipital are damaged, but judging from remaining surfaces both participated in what must have been a markedly robust paroccipital process (fig. 4).

Four prominent apertures or channels for vasculature and nerves can be identified in their expected positions in AMNHVP 11842 (figs. 1, 4)—the inferior petrosal sinus and the stylomastoid, hypoglossal, and jugular foramina. The stylomastoid foramen is situated in an elongated groove, relatively high on the sidewall of the cranium and immediately posterior to the posterior crus of the ectotympanic (or bone derived therefrom). AAHS 95044 (fig. 1), photographed from a more medial angle and without obscuring occipital material, displays several other small apertures, including the cochlear canaliculus for the perilymphatic duct and the canal for the tympanic nerve. Posteriorly, the line of petromastoid-exoccipital contact leaves space for venous drainage of the petromastoid, as in extant caviomorphs (e.g., Myocastor coypus, fig. 22B). Nearby there is a tiny channel (?mastoid canaliculus for the auricular ramus of the vagus nerve), passing obliquely through bullar material. (No exit foramen was identifiable on the bullar wall, but equally there is no indication that the channel actually perforated the tympanic cavity.)

The posttympanic foramen is situated low on the lateral aspect of the bulla, a short distance from the stylomastoid foramen and near a shallow dimple that may have lodged the tympanohyal-stylohyal articulation. This foramen opens directly into the posttympanic canal, discussed in detail below.

Of particular interest is the fact that both specimens display a teardrop-shaped tympanic fenestra, resembling that of certain extant chinchillids and dinomyids and positioned at the bottom of a deep fossa (figs. 2, 4). In AAHS 95044 the fenestra narrows into an ectotympano-ectotympanic suture, still partly open at the site of the original ventral cleft. In AMNHVP 11842 a more advanced stage of closure is seen, with the cleft completely closed. The suture line can still be identified, externally as well as internally, by a raised seam on the meatal floor (fig. 5).

Since the bullar floor was already broken in AMNHVP 11842, allowing inspection of most of the middle ear cavity, an equivalent area was removed from the floor of AAHS 95044 to permit comparison (see figs. 1, 4). In both specimens the tympanic cavity proper is undivided, although several low septa are present. Thus, the deep trough and low partial septum in the anteromedial quadrant of the tympanic roof (fig. 4) marks the passageways for the auditory tube and the belly and tendon of tensor tympani muscle. The cochlear promontorium is relatively low (i.e., not towered, as in Cavia, Dasyprocta, and some other cavioids) and unmarked by large sulci or cristae. No aperture for the tympanic nerve can be detected in either specimen, although it probably entered the tympanic cavity near the posterior pole of the promontory, as is typical of placentals (MacPhee, 1981). Epitympanic pneumatization of the dorsal part of the auditory capsule is exceptionally extensive in Amblyrhiza. Although no intact skulls of this taxon are known, inspection of isolated cranial fragments indicate that much of it was massively pneumatized, perhaps in much the same manner as the skull of Eumegamys (see below).

The incus-malleus of AAHS 95044 (not illustrated), a fused structure in many but not all adult ctenohystricans (Patterson and Wood, 1982), was retrieved when the bullar floor was opened. The stapes was not in evidence; it may have fallen out earlier through the damaged epitympanic recess. There is no indication that Amblyrhiza possessed a pessulus.

In AAHS 95044 the crista tympani is carried on a lengthy cylinder of bone that projects deeply into the tympanic cavity (fig. 1). Presumably, as the bulla inflated the segment of the ectotympanic supporting the tympanic membrane maintained both its proportions and its ontogentically primitive position near the cochlea. In van der Klaauw's (1931: 201) terminology the cylinder would be an eminentia cristae tympanicae, but I propose the simpler and more descriptive phrase tympanic collar (diameter, 7.5 mm in AMNHVP 11842) to refer to the expanded, intratympanic portion of the ectotympanic. (This structure in caviomorphs is also morphologically equivalent to the spurious ossified anulus membrane of lemurs, as MacPhee and Cartmill [1986] originally established.) Also worth noting is the large number of small, apparently imperforate septa adorning the collar's periphery. Their function is obscure, although some may have conducted blood vessels (?meatal innominate vasculature).

The degree of obliteration of the intratympanic ectotympano-petrosal suture is widely variable among hystricognaths. In the available specimens of Amblyrhiza, fusion is extensive: the medial bullar wall is seamlessly continuous with the promontorium in ventral aspect. Externally, the only indication of the original site of ectotympano-petrosal contact is a narrow crevice defining the border of the anteromedial part of the tympanic roof (which must therefore in part be derived ontogentically from ectotympanic material; fig. 3).

Three bony tubes pass through or along the tympanic roof. In homological terms the least problematic to identify is the broad prominence of the facial canal, which travels through the rear part of the cavity (fig. 1). The second canal lies anterolaterally, and can be traced from the anterior margin of the epitympanic recess to the lateral aspect of the tympanic roof. The canal passes through the roof to terminate in a foramen situated on the anterosuperior side of the bulla, in front of the external acoustic canal (figs. 2, 4, 5). Judging from conditions in osteological preparations of extant caviomorphs in which dried remnants of vessels and nerves exist in this position, the tube and its aperture most likely represent the canaliculus chordae tympani. However, it is relatively large to be carrying the nerve alone, and may have transmitted blood vessels as well.

The homology of the third tube, the posttympanic canal (fig. 1), is more difficult to determine. Morphological evidence is consistent with the proposition that the canal was occupied by an artery, the ramus posttympanicus of this paper. In Amblyrhiza the canal runs across the posterior part of the tympanic cavity in close relation to the posterior aspect of the tympanic collar—indeed, in AAHS 95044 it appears to simply be a larger version of one of the partial septa ornamenting the latter (fig. 1). It differs, however, in that it connects with an external foramen and, instead of terminating on the collar's rim, continues onto the crest behind the posterior pole of the cochlea, where the fossula fenestrae cochleae and stapedius fossa are located (fig. 1). Under high magnification, the trace of what appears to be a small sulcus can be seen leaving the canal to run into the last-named space. The groove's course (if any) thereafter could not be determined. Externally, the canal connects with its foramen on a prominence located low on the lateral bullar wall, between and slightly ventral to the stylomastoid foramen and tympanic fenestra. Conditions are similar in AMNHVP 11842, except that the posttympanic canal is relatively shorter (and unfortunately harder to see in ventral aspect).

Basicranial Morphology of Elasmodontomys obliquus

The AMNHVP collection contains five partial or complete auditory regions of Elasmodontomys. Two are still in position on skulls: AMNHVP 14171 (holotype), from Toraño Cave near Utuado (see Anthony, 1918), badly battered but retaining the left auditory region; and AMNHVP 143605, from an unidentified cave in southwestern Puerto Rico, in much better condition (figs. 6–7). The three remaining specimens are isolated tympanopetrosals collected by Anthony in 1916 but never catalogued prior to this study: AMNHVP 143606 (right side), from Toraño Cave, retaining an intact bulla; AMNHVP 143607 (left side), also from Toraño Cave, bulla broken; and AMNHVP 143608 (left side), locality unknown, bulla broken. AMNHVP 143607 is illustrated in figure 8. The condition of the other two tympanopetrosals is such that they do not warrant illustration or detailed description, but they were of some use in identifying certain features found on better-preserved material. Anthony (1918) collected and figured several other skulls retaining auditory regions, but unfortunately these are no longer in the AMNHVP collections.

The auditory bulla of adult Elasmodontomys differs from that of Amblyrhiza in several significant respects. It is a simple, bean-shaped excrescence, comparatively small in relation to skull size (fig. 7A), with a highly irregular external surface somewhat reminiscent of that of extant Cuniculus. The anterior end of the bulla is considerably roughened for reception of the medial wing of the pterygoid. As in Amblyrhiza, there is no indication that the external meatal margin bore accessory bones. There are many pinhole-sized foramina distributed over the bullar surface, but nothing corresponding to the large posttympanic foramen of Amblyrhiza.

The external acoustic canal has a complete ventral lip, with no evidence of the earlier presence of a tympanic fenestra. A vaguely defined groove on the external bullar wall runs subvertically below the external acoustic meatus (asterisk, fig. 7C). Close study of the walls of the groove indicates that it is punctuated by small foramina, but there is no indication that it was originally a gap. The deep fossa that indents the external wall of the recessus meatus in Amblyrhiza is also absent. Instead, this area has the same contour as the rest of the lateral bullar wall, as in other extant caviomorphs lacking the tympanic fenestra.

Other foraminal content and positioning is similar in the two West Indian taxa, as would be expected given the lack of variability in such features in caviomorphs generally. The stylomastoid foramen is located below and behind the meatal aperture; a deep sulcus for the facial nerve slopes ventrally away from this foramen. The cochlear canaliculus is perched on the lip of the jugular fossa, but the aperture of entry for the tympanic nerve cannot be identified. (There are two shallow grooves on the promontorium, one of which is directed into the aperture for the tubal canal, but neither can be traced back to the posterior pole.)

The auditory region is relatively less inflated than in Amblyrhiza, although there are small paratympanic spaces emanating from the epitympanic recess that pneumatize the posterior and superior portions of the petrosal. The bullar wall is exceptionally thick and internally vesiculated. There are no noteworthy differences in promontorial position or shape, although in Elasmodontomys the line of ectotympano-petrosal contact on the ventral surface of the promontory is clearly apparent, more so than in Amblyrhiza (fig. 6). Apertures or grooves for the auditory tube and tensor tympani muscle are situated in the expected places (fig. 8). A large exit foramen for the chorda tympani can be detected on the outer sidewall of the bulla, but no tube for it was visible internally. The function of the aperture identified as a “?venous foramen” in figure 7B and C is in fact unknown, but as it does not conduct anything into the tympanic cavity it cannot be a posttympanic foramen. Unsurprisingly, in view of the lack of an identifiable posttympanic foramen, there is no canal running from the lateral bullar wall along the posterior aspect of the tympanic collar in any specimen. The ramus posttympanicus is therefore considered absent (figs. 6, 8).

Basicranial Morphology of Eumegamys paranensis

MLP 41.XII.13.237 (fig. 9) is a relatively complete, albeit shattered, skull collected in 1941 by E.D. Sors in prov. Entre Ríos. The posterior cranium is in good condition, but the more anterior portions of the skull are not. The palate and much of the nasal rostrum survive as a separate piece, but the orbital, jugal, and pterygoid regions are either missing or represented by detached fragments (mandibular fossae of the squamosal, right nasal, and pieces of the right and left premaxillae).

At some point after the discovery of this specimen the floor of the left auditory bulla was carefully sawn off, exposing the middle ear cavity on that side (fig. 10). This piece could not be found when the skull was made available for study, and it is presumed lost. The sawcut was made in such a way that the external acoustic canal was transected along its long axis, thus revealing its extraordinary length (29 mm from position of crista tympani to lateral margin of external acoustic canal). Although the saw damaged the crista tympani and ventral surface of the promontorium, the other major features of the middle ear are otherwise mostly intact. Unfortunately, a thin scale of calcium carbonate hides most structures on the tympanic roof, and what appears to be casting material blocks many of the small apertures within the middle ear. Although it is clear from the paper by Negri and Ferigolo (1999) that the ear region of Neoepiblema is substantially similar to that of Eumegamys, their account lacks sufficient detail to make useful comparisons.

As in Amblyrhiza, the auditory bulla is large (∼35 mm, long axis) and ventrally flattened. The lateral bullar wall and related areas of the basicranium are extremely thick (fig. 10). The anterior end of the bulla terminates in two large processes, also as in Amblyrhiza. The more medial of the two is grooved by a large, smoothly rounded and posteriorly directed channel, presumably for a large vein connected to the pterygoid venous plexus and passing out of the enormous combined foramen lacerum + foramen ovale (11 mm, maximum width). Dorsal to this feature on the caudal lip of foramen ovale, and thus not visible in figure 9, is a large aperture which presumably housed the cartilage and ostium of the tubal canal.

The auditory region is moderately pneumatized; other than the epitympanic recess, which could not be adequately explored, there are no paratympanic spaces of large size, and only a few pneumatic cellules can be seen in the vicinity of the auditory tube, posterior wall of foramen ovale, and in the mastoid and paroccipital processes (fig. 11). By contrast, portions of the skull rostral to the auditory region, although poorly preserved (and not illustrated here), were clearly very extensively inflated. Paranasal sinus expansion evidently inflated the frontal bone in such a way that it was divided into inner and outer tables, separated by a virtually continuous airspace that extended from the nasal rostrum to the position of the coronal suture. (A similar condition can be plausibly inferred for Amblyrhiza, despite the lack of complete skulls for this taxon.) These massive pneumatic chambers nearly isolated the actual bone-lined brain case—a date-sized protrusion on the cranial floor connected by septa to the cranial roof. Interestingly, the pneumatized area stops abruptly along the line of frontal/parietal contact; caudal to the coronal suture the skull is massively thickened (as in the case of the mastoid) rather than pneumatized. Such extensive buttressing might be explained as an adaptation for transmission of forces through the neck region, perhaps because the upper incisors were used for powerful digging or cutting. The thickness of the frontal bone along its posterior edge is 2.1 mm, whereas the thickness of the parietals immediately caudal to frontal contact rises to 20.5 mm—thus thicker by a factor of 10.

The upwardly and outwardly projecting external acoustic canal is nearly intact on the specimen's right side and is framed by three gaps. The rostral gap (fig. 1, single asterisk), which opens into the epitympanic recess, is simply an artifact caused by damage to the bullar wall. The teardrop-shaped aperture caudal to the meatal opening is the stylomastoid foramen; the thin, somewhat damaged wall of bone (double asterisks) separating these two apertures is interpretable as derived from the unexpanded posterior crus of the ectotympanic. The remaining dehiscence, which intersects the ventral margin of the meatus, is the tympanic fenestra in the form of a narrow slit (ventral cleft). The bulbous rostral wall of the cleft has the appearance of being turned over on itself, almost in the form of a scroll, but there is no evidence that it represents a separate ossification. In life the tympanic fenestra and the aperture of the external acoustic meatus may have been externally separated by the same damaged sheet of bone just mentioned, much as appears to be the case in the specimen of Neoepiblema described by Negri and Ferigolo (1999), who, however, did not distinguish the fenestra from the meatus. Hypoglossal, jugular, and other constant foramina can be identified in their expected positions.

Approximately 15 mm ventral to the stylomastoid foramen there is another foramen, situated in the floor of a narrow but prominent groove (fig. 11). The foramen occupies a position similar to the posttympanic foramen of Amblyrhiza, and will be identified as such in the rest of this paper. (Neoepiblema apparently exhibits both apertures as well, although as previously noted Negri and Ferigolo [1999: fig. 14] assumed that both conducted parts of the facial nerve.) The identity of the groove is uncertain, although an obvious possibility is that it may have accommodated the cranial part of the hyoid apparatus, as did the similarly-situated feature in Amblyrhiza.

Structures on the left tympanic roof of this specimen can be studied with ease, thanks to the removal of the tympanic floor, but because of damage some points of morphological interest cannot be adequately investigated (fig. 10). This applies, for instance, to the identity of the posttympanic canal (see below). Medially and laterally the slight bulge of the promontorium is bordered by low eminences and diverticula. Some of these structures may be canals conducting nerves or arteries, but others are simply partial bony septa. The continuous ridge that borders the medial, rostral and part of the lateral side of the low promontorium appears to represent a sutural contact between the onlapping ectotympanic bulla and portions of the petrosal (promontorium + ?medial portion of the tegmen tympani) rather than a canal. Anteromedially the line of this ridge is continued forward by the low crest that divides the fossa for the tegmen tympani from the tubal canal. The area of the posterior pole of the cochlear promontorium is broken and matrix-filled, and the fenestra cochleae and vestibuli cannot be distinguished as such.

Laterally there are deep pockets in the tympanic roof surrounding the tympanic collar. The rostralmost conforms to the entrance to the greatly expanded epitympanic recess. Pockets in the rear wall of the tympanic cavity may represent the aditus of the mastoid cavity, although as already noted the mastoid part of the petrosal is itself little inflated.

Given the likely pathway of the posttympanic ramus, I have inferred that one of the apertures seen on the cut portion of the mastoid represents the posttympanic canal (here completely encased in thick bone and thus not at all canallike). However, this is far from definite as it was not possible to pass a probe through the aperture into the tympanic cavity (hence C3 is scored as “?” in appendix 2).

There are several subsidiary crests on the rostralmost part of the tympanic roof that can only be seen by strongly tilting the specimen (not illustrated). Most are partial septa, probably imperforate, although one prominent ridge that extends from the facial canal and bends rostromedially toward foramen ovale might have contained the lesser petrosal nerve. A less likely, but not irrelevant, possibility is that some of these crests conducted rostrally directed remnants of the stapedial system (see Development, Homology, and Character Analysis). There is no information on the middle ear cavity of Neoepiblema.

Basicranial Morphology of Dinomys branickii

Pacarana specimens available for this study display a large aperture on the posterolateral bullar wall, situated ventromedial to the stylomastoid foramen and in essentially the same relative position as the posttympanic foramen of Amblyrhiza and Eumegamys (figs. 12–14). To confirm homology, the bulla of Dinomys branickii AMNHM 46551 (fig. 13A) was opened and inspected. As previously noted, in this specimen a large tube, continuous with the foramen, projects from the posterior wall of the tympanic cavity and runs thence onto the caudal tympanic process of the petrosal, where it ends, just as in Amblyrhiza. The end of the tube is perforated; a tiny channel issues from it and disappears into the fossula fenestrae cochleae. However, it should be noted that a complete canal is evidently not always present; in AMNHM 70354 (fig. 13B), a large adult, the posttympanic foramen opens onto a shallow sulcus on a low septum rather than a tube.

To conclusively determine the source and nature of structures traveling within the posttympanic foramen of Dinomys, an alcohol-preserved specimen (AMNHM 201638) was selected for detailed study. After removal from the body, the head was partly skinned and both parotid regions dissected; only the right-hand side is depicted here (fig. 14). In order to provide a window of a size sufficient to visualize relevant basicranial structures, the gonial portion of the mandible was sacrified, together with muscles and veins overlying part or all of the area of interest (platysma, sternocleidomastoideus, superficial masseter, pterygoid muscles; portions of the external jugular and facial veins). The digastric (both bellies) and stylohyoid (jugulohyoid) muscles were exposed but left intact. Proximal parts of the external carotid artery, internal jugular vein, and vagus nerve within the upper neck were cleaned to the point where they passed deep to the digastric. The tough meatal tissues were carefully pruned to expose the extracranial track of the facial nerve and the structures related to the posttympanic foramen, but given the delicacy of the latter it was decided not to attempt to clear the bony meatus or tympanic fenestra. Divisions of the external carotid artery were briefly followed: lingual and facial arteries were identified, but not traced rostrally beyond the parotid region, and the occipital artery (not illustrated) was found deep to the caudal belly of the digastric. Areolar tissue cleanly separated the caudal digastric from the ventral bullar wall, which was therefore easily exposed.

Only one major blood vessel traverses the potential space between the bulla and the suprahyoid musculature. From its origin on the caudal aspect of the external carotid, the vessel ascends caudodorsally in close relation to the upper free edge of the stylohyoid, crosses the nerves to the latter and the caudal digastric, then continues dorsally, closely invested by tissues comprising the membranous meatus and pinna. In these relations it strongly corresponds to, and is therefore regarded as, the homolog of the posterior auricular artery of Homo (see also Guthrie, 1963).4 At least in the area surveyed there was no evidence of a true internal carotid artery medial to the external carotid trunk, and the former is judged to be absent (either completely or at the stage represented by this specimen).

The postganglionic end of the sympathetic system is represented by a small trunk that becomes an extensive plexus on the external carotid. A separate deep petrosal nerve, like that found in primates, eulipotyphlans, and several other groups, which travels through the middle ear to the pterygoid canal where it meets the greater petrosal nerve (MacPhee, 1981), was not seen. Since the course of the internal carotid artery is extrabullar even in the one group of caviomorphs (erethizontids) known to exhibit this vessel in the adult stage (Bugge, 1974a, 1974b), postganglionic sympathetic fibers destined for the nasal mucosa, lacrimal gland, and other cranial targets must be distributed along branches of the external carotid artery and thus never pass through the middle ear.

The chief targets of the dissection were the stylomastoid foramen and posttympanic foramen. With careful cleaning of meatal tissues adherent to the posterolateral wall of the bulla, the usual branches of the facial nerve were identified as they left the stylomastoid foramen. A few millimeters posteriorly, a small division of the posterior auricular artery was seen to penetrate the bulla at a position that corresponds to the expected entry point of the ramus posttympanicus. This interpretation is fully consistent with the inferences made on the basis of osteology. In order to preserve anatomical relationships in the dissected specimen for future reference, it was decided to end the dissection leaving the bullae intact (fig. 13).

Relevant Basicranial Features of Extant Ctenohystrican Families

In character definition and analysis it is important to be able to distinguish between structures that appear to be operationally homologous from those that are not. The purpose of this section is to briefly summarize, for comparative purposes, relevant developmental and morphological aspects of the basicrania of crown-group Ctenohystrica, with an emphasis on caviomorphs. Except for the main targets of this paper, extinct taxa are usually not referenced. Their study is well beyond my immediate purpose, and, in any case, there are virtually no paleontological studies with much relevance to the points of interest here. Although conditions in at least one member of each extant family (as recognized by Woods and Kilpatrick, 2005) are described, only a subset of these could be illustrated photographically. These are, however, augmented by short descriptions of particular specimens in the AMNHM and USNMM collections.

Ctenodactyloidea

Phiomorpha

Although tiny foramina occur in the circummeatal area in phiomorphs (figs. 16–18), the only constant, sizeable aperture on the lateral bullar wall is the stylomastoid foramen. However, notice should be made of a large external groove (sometimes taking the form of a partly bridged canal) situated behind the posterior margin of the meatus in the available specimens of Bathyergus, Hystrix, Atherurus, and Thryonomys (but not Petromus). A similar feature is seen in many caviomorphs, suggesting that it is probably primitive for hystricognaths. The vessel occupying the groove is assumed to be homologous with the posterior auricular artery, which in placentals typically ascends behind the meatal opening to feed the external ear and related portions of the scalp. Given the generally close proximity of the stylomastoid foramen to the artery's trackway, it is not unlikely that it releases small branches to the former in order to feed local structures. However, none of these taxa displayed features conforming to the posttympanic foramen or canal as seen in Dinomys. If the equivalent of the ramus posttympanicus exists in these taxa, it must travel to its targets via the facial canal (rather than through an independent tube), and is thus not detectable osteologically.

Caviomorpha Erethizontoidea

Chinchilloidea

Octodontoidea

Tympanic Fenestra and Development of the Lateral Bullar Wall

Although it would be necessary to have access to developmental series in order to thoroughly document fenestral development in different clades of ctenohystricans, major patterns can be plausibly induced from juvenile and adult conditions, as briefly treated in this section and figure 25. The present investigation indicates that the critical factor in fenestral development is the relative growth rates of the crural margins of the primordial U-shaped ectotympanic. Theoretically, if crural growth rates were roughly similar on both the lateral and the medial aspects of the ectotympanic, the bulla would tend to expand in a simple, balloonlike manner, with the primitive external acoustic meatus appearing to reduce in relative size as its floor ossifies. Further, if the floor of the meatus were to evenly ossify and remodel in concert with the rest of the lateral bullar wall, as in humans, the fenestra as a “potential space” separate from the meatal aperture should disappear by the definitive stage of development. By contrast, if growth gradients along the crura were such that, for example, one crus tended to preserve its embryonic proportions while the other became greatly expanded, the effect on the meatal floor would be a persistent defect (ventral cleft) in the presumptive margin of the external acoustic meatus. In later ontogeny the defect might remain completely open or partly close, depending on the degree of approximation of bone territories.

Other than the fibrous membrane and associated dense connective tissues, the only structures in the presumptive meatal area are the tympanic membrane and the vascular elements grouped here as meatal innominate vessels. Meatal innominate foramina are common in the target groups, and are found in taxa that exhibit the tympanic fenestra as well as those that do not. It is unlikely that local blood supply has any influence on the relative speed of ossification. Body size is likewise irrelevant to fenestral persistence. Although megafaunal rodents like Amblyrhiza, Eumegamys, and Phoberomys (?and perhaps Josephoartegasia) all possess a large tympanic fenestra in the adult stage, so do much smaller Dinomys and many cavioids. On the other hand, the size of the fenestra is positively correlated with the length of the external acoustic canal, which is itself a proxy for the hyperdevelopment of the anterior crus. This is especially evident in chinchilloids.

The reconstructed ontogeny of the three character states of C1, each developing from the indifferent or unexpanded anulus of late fetal life, is depicted in highly diagrammatic form in figure 25. The facial nerve, in its standard or primitive position passing immediately behind the posterior crus (MacPhee, 1981), is provided as a reference point. In order to make character analysis consistent, C1:0 does not distinguish between complete absence of all apertures in the circummeatal area and presence of one or more meatal innominate foramina. C1:1 and C1:2 cover the remaining possibilities, which focus on whether the fenestra is simply a notch or a separate window, and whether an identifiable ectotympano-ectotympanic suture is formed.

The tympanic floor conformation seen most frequently in mammals is C1:0 (fig. 25, top), in which the definitive, smoothly rounded meatus is the only sizeable opening in the lateral wall of the bulla. Minor variations on these conditions are seen in many caviomorphs (capromyids, echimyids, and cuniculids) and Old World ctenohystricans (e.g., Ctenodactylus gundi USNMM 325852, fig. 15; Thryonomys swinderianus AMNHM 216340, fig. 18). For this pathway the track of a meatal innominate vessel is also shown; it is assumed that such vessels are constantly present, whether or not an evident trackway can be discriminated osteologically.

C1:1 (fig. 25, middle) is possibly a wastebasket grouping, but with the material at hand I am not able to discriminate more than one character state. The recognition criterion for this character state is the notched condition of the ventral margin of the meatus. The apex of the notch is either little or not at all expanded, and shapes range from a narrow slit to a triangular gap in the meatal floor.

The only complication in recognizing this character state is that small projections from the margins of the notch may eventually grow together in some taxa or individuals, yielding a bridged condition with a large opening above (meatus) and a much smaller aperture below (tympanic fenestra). For example, notches are nearly closed bilaterally in Dolichotis sp. AMNHM 48218, an old juvenile, and only just bridged over in Galea musteloides AMNHM 213465. In adult Cavia aspera AMNHM 264468, bridging is complete, and the distance between fenestra and meatus is very wide indeed. It is not known whether the presence of so-called accessory meatal bones have an effect on fenestral development in this taxon. This last arrangement impinges, at least descriptively, on that described under C1:2, although other distinctions also apply. As explained below, the primary difference is that the bony projections of the bridged condition of C1:1 do not ordinarily form sutural tissues at an early developmental age or lead to the lengthy line of sutural fusion seen in taxa expressing C1:2.

Interestingly, some specimens display a notch on one side and a bridge on the other, which points up that these morphologies should be thought of as slightly different outcomes produced along the same general ontogenetic pathway. Thus in the dasyproctid Dasyprocta punctata AMNHM 41394, a very young animal in which only the first upper molars are fully erupted, the meatus is deeply notched (fig. 24). This also applies to a somewhat older D. punctata AMNHM 23461 (upper third molars erupting), but in the dentally mature D. punctata AMNHM 262281 the notch is bridged on the left side and but not on the right.

The hallmark of C1:2 (fig. 25, bottom) is very marked differential growth of the crura and elongation of the meatus into a long canal, as seen in certain chinchillids (e.g., Lagostomus maximus AMNHM 80208, fig. 20). Sutural tissues are inferred to form at an early stage, because in L. maximus AMNHM 70222 (47 mm skull length), a juvenile, the ectotympano-ectotympanic suture is already closed, but still detectable. In the adult stage, a low ridge or seam may be seen running along the line of fusion inside the meatus, marking the line of contact (see Amblryhiza inundata AMNHVP 11842; fig. 5). Chinchilla is essentially the same as Lagostomus, except that pneumatization is so pronounced in this genus that the original form of the crura is entirely masked. However, the closed ventral cleft can easily be distinguished on the lateral bullar wall.

In summary (see also fig. 27), there are no morphological grounds at present for inferring that the fenestra develops, and later disappears, during the ontogeny of most phiomorphs. (Petromus is the only extant contrary example, although this point should be checked in fossil taxa.) Within Caviomorpha, at the superfamily level the fenestra is absent in adult erethizontoids (i.e., inferior margin never notched, meatal innominate vessels small). In adult stages of extant octodontoids the fenestra is either absent or no more than a small aperture, in many cases bridged off from the meatus by a strap of bone. In most taxa that express a definite fenestra, margins remain rather ragged, as might be expected in an imperfectly ossified region. However, in others the aperture is small and rounded, and functions exclusively as a true foramen for transmitting vasculature (e.g., Spalacopus cyaneus AMNHM 33276; Myocastor coypus AMNHM 35626). In cavioids and chinchilloids, by contrast, the fenestra is usually fairly large relative to the size of the external acoustic meatus.

Amblyrhiza agrees more closely with conditions in Lagostomus and other chinchillids than with Dinomys, in that it possesses a markedly elongated external acoustic canal and a fused ectotympano-ectotympanic suture in the adult in combination with a patent tympanic fenestra. The skull of Eumegamys, although damaged in the critical area, agrees closely with Dinomys in retaining an open ventral cleft and much shorter external acoustic canal. Elasmodontomys resembles the majority of octodontoids in completely lacking a fenestra.

Ramus Posttympanicus as Stapedial Ramus Posterior

Is the occupant of the posttympanic canal found in some caviomorphs in fact the homolog of the stapedial ramus posterior, a vessel never previously identified in a rodent? From a comparative standpoint, this question is not as unlikely as it may seem. In several basicranially primitive placental taxa (Erinaceus, Solenodon, and several tenrecs) that preserve an intact, primitive stapedial system, the ramus posterior is a branch released by the parent vessel just before the latter passes through the obturator foramen of the stapes (MacPhee, 1981). In Erinaceus the ramus posterior is very small, supplies only the area of the stapedius fossa, and evidently involutes in early postnatal life. In tenrecs and, especially, Solenodon, it is a large vessel that travels outside the confines of the middle ear in close relation to the facial nerve (MacPhee, 1981). Recently, Wible (2008) confirmed that the ramus posterior in Solenodon supplies the area conventionally associated with the posterior auricular artery in other mammals (pinna and related structures in the rear part of the head).

In Homo, a vessel with somewhat similar relations, but lacking its own foramen and individual bony conduit, is the stylomastoid branch of the posterior auricular artery. Whether this vessel may itself be a stapedial remnant has never been properly discussed in the general morphological literature, let alone that oriented toward Rodentia. Thus, Aihara (1958) noted that a branch of the stylomastoid artery—his “stapedial artery”—feeds the stapedial muscle in Cavia. However, he did not consider whether both, either, or neither might be homologous with any part of the primitive stapedial system. As Cavia definitely lacks a posttympanic canal (see below), the situation remains as ambiguous in this taxon as in any other.

While the possible equivalency of the ramus posttympanicus and part or all of the stylomastoid artery's distributary network is worth acknowledging, there are other observations that suggest a different picture for hystricognaths. Although Wible (1984) did not encounter a ramus posterior in any of the taxa he personally examined for his monumental study of mammalian cephalic arterial systems, in his literature review he noted some additional instances of the probable or possible occurrence of this vessel. One such instance is directly relevant here: Struthers' (1930) study of cephalic arterial development in fetal Erethizon, in which this author described an “occipital” artery transitorily linked to the developing proximal stapedial. Wible (1984) commented that the stapedial “occipital” connection traveled in close relation to the facial nerve and stapedius muscle, just as in verified occurrences of the ramus posterior in nonhystricognaths. Whether the porcupine “occipital” is homologous with the posterior auricular of Homo cannot be determined from the information provided by Struthers, but that is certainly a possibility.

The question that immediately occurs is whether the arrangement seen in fetal erethizontids can be generalized to explain some of the findings presented in this paper. Figure 26 illustrates a possible, although speculative, interpretation. In some unspecified caviomorph (or hystricognath) ancestor the entire primitive stapedial system is inferred to have been present (fig. 26A), fed by the internal carotid stem, and perhaps largely functional in the adult. The ramus posterior might have already possessed the indicated connection with the posterior auricular, or this may have developed later on when the stapedial system as a whole began to involute. With involution, the distalmost segment—now as the ramus posttympanicus—would have crossed the track of the facial nerve and continued to supply the area of the stapedius fossa, possibly via retrograde blood flow from its anastomotic partner as suggested in figure 26B (which purposely mimics conditions in Amblyrhiza, although apart from the ramus posttympanicus there is no conclusive evidence for other possible stapedial vestiges in this taxon). How much of the resultant arterial pathway should be considered equivalent to the original ramus posterior and how much to the neomorphic anastomosis is impossible to say without more evidence. The best course would be to retain the name “ramus posttympanicus” for the whole structure, even though it may well incorporate material ultimately derived from more than one primitive source.

There are other clues that a vestigial stapedial network still exists in extant caviomorphs. One is the inferior tympanic artery identified in Cavia by Aihara (1958), which has some of the characteristics one would expect to find in a vestige of the primitive stapedial stem: it exhibits anastomotic links with both the middle meningeal and the vessel issuing from the stylomastoid artery that Aihara (1958) confusingly named “the” stapedial artery (see above). Another is the “ramus tympanicus” of the maxillary artery that provides a “supply to the middle ear” in Cavia according to Cooper and Schiller (1975: 151). Its position of origin seems too rostral for it to represent one of the meatal innominate vesssels of this paper. Although there are several plausible homologies, it is possible that the “ramus tympanicus” may be yet another remnant of the stapedial system (?ramus inferior)—in this case, one that has been pirated by the maxillary and which now delivers retrograde flow to some part of the tympanic cavity (for parallel cases in primates, see Diamond, 1991). Another is the apparent retention of a foramen of exit for the ramus superior (evidently maintained for meningeal supply) on the cerebral side of the petrosal in Lagostomus (cf. Bugge's [1974b: fig. 6D] illustration of an injected specimen). It would be of great interest to know whether the vessel traversing this foramen anastomosed with the ramus posttympanicus or with the stylomastoid artery (assuming these are different entities in the plains viscacha).

Also possibly relevant is Tandler's (1901: 360) alleged detection of a highly reduced internal carotid artery in Hystrix: in his interpretation, the vessel entered the middle ear through a tiny aperture on the posterior side of the bulla, then traveled in a groove transpromontorially in company with the internal carotid nerve toward the cranial cavity. Dierbach (1985) found a “carotid foramen,” but no artery traversing it, in a 25 mm CRL fetus of Cavia; from this indirect (and admittedly inadequate) evidence he inferred that the internal carotid artery was formed but had already involuted. Although carotid involution occurs during the ontogeny of many placentals, loss need not be complete and some structures may be retained as functionless or near-functionless remnants (Bugge, 1974b; MacPhee, 1981; Wible, 1984; Diamond, 1991). Cooper and Schiller (1975: fig. 4–15) claim that a highly reduced version of this artery exists in adult Cavia, but this has not been confirmed either. Indeed, whether any nonerethizontid ctenohystrican possesses an intact internal carotid artery at any stage of development remains to be properly demonstrated. By contrast, a much more credible case can be made for the persistence and repurposing of parts of the primitive stapedial system.