BACKGROUND

Colpodon propinquus Burmeister (1885) was among the first notoungulates known. Burmeister (1885) based this taxon upon an upper molar and two lower molars found near the mouth of the Chubut River by a naturalista viajero (“traveling naturalist”) from what was then known as the Museo Público de Buenos Aires. Burmeister regarded Colpodon to be morphologically intermediate between Nesodon Owen, 1853, and Homalodotherium Flower, 1873, although not in any kind of phylogenetic sense, since Burmeister believed species were anatomically fixed and immutable (Simpson, 1984). Later, the same (unnamed) traveling naturalist discovered nearly complete dentitions of Colpodon from the locality of the “type” specimen, which were described and figured (Burmeister, 1891: pl. 7, figs. 4–10).

The earlier paucity of notoungulate specimens ended with the fieldwork Carlos Ameghino (Simpson, 1984). Among his many discoveries in Patagonia was material that his elder brother Florentino described and named Leontinia gaudryi (Ameghino, 1895; the generic name was to honor Florentino's wife, Léontine [Simpson, 1984]). Although the elder Ameghino eventually described several other species of Leontinia, subsequent workers have regarded these as synonyms of L. gaudryi Ameghino, 1895, or Ancylocoelus frequens Ameghino, 1895 (Loomis, 1914; Patterson in an unpublished [but frequently cited] catalog; Paula Couto, 1983; Soria and Alvarenga, 1989; Shockey, 2005). Ameghino placed all these taxa within the family he named Leontiniidae Ameghino, 1895. He considered leontiniids to be related to the Homalodotheriidae, both of which he placed in the Order Ancylopoda Cope, 1889, which included the chalicotherioids (the clawed herbivorous perissodactyls of the Holarctic, which then were generally regarded as having ordinal status distinct from perissodactyls).

With considerable reluctance (“sólo de una manera absolutamente provisoria,” Ameghino, 1902), Ameghino briefly regarded Colpodon as a leontiniid. He noted, however, that Colpodon also had characters reminiscent of notohippids, including the lack of labial cingulae on the incisors and a symmetric astragalus (Ameghino, 1902). Ultimately, he removed Colpodon from the Leontiniidae (and the more inclusive group Ancylopoda) and placed it within its own family (the Colpodontidae Ameghino, 1906). Colpodontids were included in his Order Hippoidea (Ameghino, 1906) along with notohippid notoungulates and equids. Such unions with similar South American taxa with different Holarctic groups was internally consistent with Ameghino's strong inclination to unite South American taxa with those of the world's other continents, but at odds with the opinions of his contemporaries and later judges of relationships of South American endemics.

In contrast to Ameghino's separation of Colpodon from the Leontiniidae, Gaudry (1906) believed that there was a close relationship between Colpodon and Deseadan leontiniids. Indeed, he lumped the Deseadan taxa Leontinia gaudryi and Ancylocoelus frequens within Colpodon.

Early to middle 20th-century students of South American ungulates (e.g., Patterson, 1934; Simpson, 1945) did not follow Gaudry's system of placing the Deseadan leontiniids within Colpodon, but regarded the Colhuehuapian (early Miocene) species of Colpodon as generically distinct from the Deseadan taxa, yet still leontiniids. However, its generally notohippidlike incisors continued to cause doubts in others about its leontiniid affinities. Soria and Bond (1988), for example, referred Colpodon to the Notohippidae. Their findings evidently influenced the placement of Colpodon within the Notohippidae in the authoritative classification of McKenna and Bell (1997). By that time, however, Bond and López (1995) had already returned Colpodon to the Leontiniidae based on the discovery of a specimen possessing labial cingulae on the incisors.

Cifelli (1993) performed the first cladistic phylogenetic analysis that included Colpodon. He indicated that his initial analysis placed Colpodon as the sister taxon to a clade that included leontiniids, notohippids, and toxodontids, but in the final analysis, in which “ancestral states were defined” (Cifelli, 1993: 206), Colpodon nested within the Leontiniidae. Indeed, in that modified analysis, Colpodon was sister to Ancylocoelus, with the loss of the canine regarded as the synapomorphy linking these two leontiniids. Similar results using outgroup comparisons were reported in Villarroel and Colwell Danis (1997) and Shockey (2005).

FIGURE 2.

Photograph of exposures of the Cura-Mallín Formation at the Laguna del Laja, Miocene, Chile.

To date, the most comprehensive phylogenetic study of the Leontiniidae was that of Villarroel and Colwell Danis (1997). Their character-taxon matrix for the first cladistic phylogenetic analysis designed to elucidate the evolutionary relationships among leontiniids originated as a descriptive table in the master's thesis of the second author (Colwell, 1965), although it was not originally designed for phylogenetic analyses. That matrix was small enough that the authors were able to perform an exhaustive search for the shortest tree, yielding a phylogeny suggesting that leontiniids formed two major clades: a “tropical clade” including the low latitude Taubatherium and Huilatherium and the remaining “Patagonian” leontiniids. Shockey (2005) added two new taxa from Salla (Anayatherium ekecoa and A. fortis) to the matrix and performed an exhaustive search that suggested that the midlatitude Salla leontiniids nested among the “Patagonian” clade.

The studies by Villarroel and Colwell Danis (1997) and Shockey (2005) were based exclusively on dental characters, and included some species level ambiguities since scoring was at the generic level. For example, there is enough variability between Taubatherium paulacoutoi and T. major that a composite score for the genus is problematic (see Results and Discussion).

For the current study, leontiniid taxa were evaluated at the species level. For clarity, and to reduce the possibility of ambiguities, the individual specimens studied or specific published descriptions are documented (appendix 2). Moreover, illustrations (figs. 2 and 5–12) of several of the characters and their states are provided to facilitate replication of our observations and analyses. Characters in this study differ from all previous analyses by including features of both the dentition and the postcranial skeleton.

Major goals of this study are to: (1) document and describe two new species of leontiniids from southern South America (Chile and Argentina); (2) place these species (as well as the Leontiniidae clade) within the context of an explicit phylogenetic hypothesis; and (3) attempt to resolve the phylogeny of taxa frequently associated with the Leontiniidae but that have been problematic historically (e.g., Colpodon and the Homalodotheriidae).

MATERIALS, ABBREVIATIONS, AND METHODS

MATERIALS. The initial stimulus for this paper is the analysis of a recently discovered leontiniid from the Laguna del Laja area of the central Chilean Andes, SGOPV 5704, herein designated as the holotype of Colpodon antucoensis, new species (fig. 3). It is curated in the fossil vertebrate collection (SGOPV) of the Museo Nacional de Historia Natural, Santiago, Chile. We also describe and name a second new taxon that has been discussed in the literature for some time as “Leontinia sp.” (Patterson, 1934; Paula Couto, 1983; Shockey, 2005). This new taxon, Elmerriggsia fieldia, is from Pico Truncado, southern Patagonia (Santa Cruz, Argentina). Other specimens from which data were obtained, and publications we used for scoring characters, are listed in appendix 2.

ABBREVIATIONS. Institutions (with abbreviations used in the text) that provided access to our study of notoungulates include the American Museum of Natural History, New York (AMNH), Field Museum of Natural History, Chicago (FMNH and specimen prefixes of P or PM); the Florida Museum of Natural History, University of Florida, Gainesville (FLMNH and specimen prefixes of UF).

The “number” sign (#) is used throughout to indicate a specific character as listed by number in the character analyses. The number parenthetically associated with # represents the character state.

I, C, P, and M represent upper incisors, canines, premolars and molars, respectively (lower case notation for lower teeth). Terminology for dental characters generally follows that of Patterson (1934). Many of these terms are graphically represented and defined in figures 3, 5, 6, and 9 (postcranial characters are given in figs. 7, 8, 10, 11, and 12). We follow Smith and Dodson (2003) for orientation of dentitions, where the four cardinal directions are mesial, distal, lingual, and labial. Thus, for example, the anterolingual cingulum of previous workers will be denoted as the mesiolingual cingulum, etc. Mesiodistal dimensions (lengths) of teeth were obtained at the greatest length of the ectoloph. Transverse dimensions (widths) of cheek teeth were measured from the base of the paracone ridge of the ectoloph to the internal border lingual to the protoloph, including any cingulae.

METHODS: Parsimony analyses were performed in PAUP 4.0b10 (Swofford, 2000), using a heuristic search. All characters were treated as unordered and of equal weight. The character-by-taxon matrices include up to 27 species of notoungulates (12 of which are leontiniids) and 83 dental plus postcranial characters. To elucidate the phylogenetic position of leontiniids within toxodontian notoungulates, we included taxa from a variety of toxodontian “families” (e.g., toxodontids, notohippids, isotemnids, and a homalodotheriid). Taxa were selected based on the presence of reasonably complete dental and postcranial material. Our desire to undertake a more comprehensive analysis that includes postcranial elements limited selection to mostly post-Eocene taxa, since pre-Oligocene notoungulate skeletons are scarce. Sufficient material was available, however, to include the Eocene isotemnids Thomashuxleya and Anisotemnus and the Paleocene-Eocene oldfieldthomasiid, Colbertia magellanica. Colbertia served as the outgroup for the cladistic analyses. A second analysis was performed in which five typothere taxa were added (see appendix 2).

The age of the Colpodon antucoensis holotype was estimated from radioisotopic analyses of two ashfall tuff samples (CH-31 and CH-32; fig. 4; appendix 3) that bracket the horizon at Estero Correntoso of Laguna del Laja from which the holotype was discovered. The sample ages were determined by the ⁴⁰Ar/³⁹Ar incremental heating method at the University of California Santa Barbara Laboratory for Argon Geochronology using the general procedures and system described by Gans (1997). Standard density and magnetic separation techniques were used to generate pure separates of plagioclase from these ash layers. Taylor Creek Rhyolite sanidine with an assigned age of 27.92 Ma was used as the flux monitor. For comparison, we obtain an age of 27.75 Ma on Fish Canyon Tuff sanidine (another widely used standard). The results of these analyses are given in the “Distribution” section of the description of C. antucoensis (below).

LEONTINIID POSTCRANIAL SKELETON

Aside from the monographic work describing complete skeletons of Scarrittia (Chaffee, 1952), little has been reported regarding the postcrania of leontiniids. The sum of knowledge of postcrania of leontiniids other than Scarrittia is minimal and includes elements referred to Colpodon (astragalus noted, but not figured by Ameghino, 1902), Leontinia gaudryi (proximal tarsus figured by Gaudry, 1906: fig. 46; an atlas, axis, humerus, and ulna described and figured by Loomis, 1914), and Taubatherium paulacoutoi (proximal ulna and radius and calcaneum described and figured as “Leontinia sp.” by Paula Couto, 1983). Here we briefly summarize the known postcranials of leontiniids and add a few new observations. Our phylogenetic analysis of leontiniids includes 43 postcranial characters, a substantially greater number than included in prior analyses of this or related notoungulate groups.

AXIAL SKELETON: The most conspicuous feature of the axial skeleton of Scarrittia canquelensis was its elongated neck (Chaffee, 1952). Chaffee noted that this elongation was greater than that of the neck of the other notoungulates (the toxodontids Nesodon and Adinotherium and the homalodotheriid Homalodotherium) to which he compared it. He did not compare it to the poorly known neck of Leontinia, but that of Leontinia may be shorter, as evidenced by the greater relative total length of the axis (centrum + odontoid process) compared to the width of the axis at the condyles. This ratio is 1.60 in Scarrittia (AMNH 29584: odontoid + centrum = 163 mm/condyle width = 102 mm) but only 1.35 in Leontinia (Loomis, 1914). It also is relatively short in outgroups. Thus, until the cervical regions of other leontiniids become better known, the elongated neck of Scarrittia is best regarded as autapomorphic.

SCAPULA: The scapula of leontiniids has been documented only for Scarrittia (Chaffee, 1952; see also fig. 10). We note a scapula in the FMNH Riggs collection that is referable to Leontinia gaudryi of Cabeza Bianca (P 12684). The supraspinatus fossa is badly damaged, but the border at its medial extreme suggests that it was much smaller than the infraspinatus fossa. The high scapular spine lies at an angle such that it leans over the supraspinatus fossa and away from the relatively large infraspinatus fossa. Although the acromion sits high on the spine, it is recessed away from the region above the glenoid fossa. There is no evidence of a metacromion. The glenoid has a teardrop form in which the apex lies opposite the coracoid region. The coracoid region is blunt and rounded, lacking a distinct coracoid process. Compared to the scapula of Scarrittia, that of Leontinia differs by way of a less developed coracoid process and apparent absence of a well-developed metacromion. The available specimens also suggest that the overall outline is less rounded than that of Scarrittia, and that there are unequal relative sizes of the supraspinatus and infraspinatus fossae, with Leontinia appearing to have a smaller supraspinatus fossa.

HUMERUS: The humerus is known for both Scarrittia (Chaffee, 1952) and Leontinia gaudryi (Loomis, 1914; and AMNH 32774). They are similar in these two taxa in both having a reduced medial epicondylar processes, no entepicondylar foramen, weakly developed supinator crest, and similar antebrachial articular morphology wherein the transversely concave trochlea grades into the convex capitulum.

FOREARM: Significant components of the antebrachium of leontiniids are known for Scarrittia, Leontinia, and Taubatherium (Loomis, 1914: fig. 77; Chaffee, 1952; and Paula Couto, 1983: figs. 9, 14a, b). All three have olecranon processes that are directed in the ventral direction, as occurs in most “advanced Toxodontia” (e.g., Nesodon and Adinotherium [Scott, 1914], Rhynchippus equinus [FMNH 13410]), and is typical for ungulates having an erect stance (Van Valkenburgh, 1987; Shockey and Flynn, 2007), but differs in the notohippid Eurygenium (Shockey, 1997) and from many typothere notoungulates (e.g., Protypotherim [Sinclair, 1909] Trachytherus [Shockey et al., 2007]), which have olecranon processes that are somewhat dorsally directed (indicative of a crouching stance [O'Leary and Rose, 1995; Van Valkenburgh, 1987; Shockey and Flynn, 2007].

MANUS: Scarrittia is the only leontiniid in which the hand is satisfactorily known (Chaffee, 1952). It is mesaxonic and pentadactyl, but digits I and V are much reduced compared to II–IV. Metacarpals II–IV are subequal in length and are half again as long as the metatarsals. Two carpals of a leontiniid were collected at Pico Truncado on the Marshall Field Expedition, a right lunar (PM 411) and a right unciform (PM 412). Aside from one significant detail (below), they are of a similar size and morphology to the homologous elements of Scarrittia, so there is no real doubt as to their identification as a leontiniid, but it is uncertain to which of the leontiniids of Pico Truncado it pertains (Elmerriggsia or Leontinia). The anatomical fit is imperfect enough to suggest that they are from two individuals. The unciform is instructive, because the distal facet is a simple concavity that has subequal transverse and dorsopalmar dimensions. Unlike the unciform of Scarrittia, which has a small Mt V facet on the same plane as the Mt IV facet, there is no evidence of a facet for Mt V This suggests that Mt V is more reduced (perhaps absent) in the Pico Truncado taxon than in Scarrittia.

FEMUR: Regarding the femur of Scarrittia, Chaffee (1952) documented the prominent head, the distinct neck and greater trochanter, and the proximally placed third trochanter, but said nothing of the distal region of the element, perhaps owing to the distortions in all the available material.

A left distal femur and proximal tibia (P 13632, fig. 8A) in the FMNH collection is referred to Leontinia gaudryi. This specimen is distinctive in regard to the asymmetry of its patellar groove; the medial trochlear ridge is longer than the lateral, but it does not appear to have been significantly higher (some damage makes it impossible to assert this with great confidence). Also, proximal to the patellar groove, the lateral region of the cranial face of the femoral shaft forms a longitudinal suprapatellar keel that is much higher than the rest of the cranial face of the distal region the shaft. A cross section at this region would have the form of a blunt comma, in which the concave side of the comma would represent the cranial face of the shaft and the tail of the comma the medial ridge. How far this extends up the shaft cannot be determined from available material.

A distal femur from the Salla, Bolivia, of late Oligocene age (UF 225743; fig. 8B) is referred to Anayatherium, cf. A. fortis. Like the femur of Leontinia, it has a suprapatellar keel; however, its distal component is more pronounced and it grades into an enlarged and elongated medial trochlear ridge of the knee.

CRUS: Chaffee (1952) suspected that the proximal tibia and fibula of Scarrittia were fused in life, but he was unable to assert this definitively, due to damage of material he studied. Proximal tibiofibular fusion is unambiguously shown in the proximal tibia-fibula (P 13632) of the leontiniid crus from Pico Truncado. However, it is not certain if that specimen pertains to Elmerriggsia or Leontinia.

PES: Scarrittia is the only leontiniid for which the foot is completely known. Chaffee reconstructed it as being pentadactyl, although functionally tridactyl. It is interpreted as having been semidigitigrade, with the long plantar processes suggestive of having supported a significant footpad that resulted in the calcaneum being raised somewhat from the ground by supportive tissue (Chaffee, 1952). Proximal tarsals are documented for some other leontiniids, including Leontinia gaudryi (Gaudry, 1906), Scarrittia, Taubatherium (Paula Couto, 1983), and now Elmerriggsia (above). Several distinctive traits are common to all of these leontiniids, including some characteristics more inclusively shared with notohippids and toxodontids. These “advanced toxodont” traits include calcaneonavicular articulation (“reverse alternating tarsus” sensu Cifelli, 1993) and a strongly developed fibular facet that is dorsoventrally (not obliquely) directed. Peculiar to the leontiniids (Scarrittia, Leontinia, Elmerriggsia, and Taubatherium) is the distinctive form of the fibular facet, which has the greater transverse dimension at the proximal border compared to the distal (fig. 4).

CHARACTER ANALYSIS

We developed the following character analysis based upon our original studies of specimens, complemented by published descriptions and illustrations. Appendix 2 lists the published sources and the instructive specimens we examined.

Some data obtained in this study also augment and correct observations from previous studies of leontiniids. For example, Shockey (2005) indicated that the presence or absence of I3–P1 of the holotype of Anayatherium fortis could not be ascertained. However, AMNH pre-parator Edward Pedersen expertly removed matrix that had obscured the alveoli of the holotype, so that we can now objectively assert that A. fortis had a complete (3-1-4-3) upper dental series, including I3–P1 (fig. 9A). Also, our conclusions regarding the state of the molar mesiolingual cingulae of Taubatherium differs from that of Villarroel and Colwell Danis (1997) who reported it as absent; the molars of T. paulacoutoi do indeed have a mesiolingual cingulum (Soria and Alvarenga, 1989: fig. 3).

Whenever possible, we modified our descriptions and scorings so that they would be compatible or interchangeable with characters previously defined and used in the literature (e.g., Cifelli, 1993; Villarroel and Colwel Danis, 1997). This was done in an effort to approach a convention or standard so that analyses may be more universally applicable and comparable. In some cases we needed to modify some character definitions or scoring to better capture the morphological states in the taxa used in this study. Where we used or modified character states from the literature, we cite the source and give the character number (#) used in the original study. The work of Cifelli (1993) contained several notoungulate analyses, however, our discussion here is limited to his analysis of the “Advanced Toxodont Families” (Cifelli, 1993: appendix 4).

The characters selected and defined in this study were those that were most relevant for understanding relationships among the Leontiniidae and across the Toxodontia. Other notoungulate taxa were included as near outgroups; however, we defer any broader phylogenetic conclusions regarding the relationships among those outgroup taxa for other works we currently have in progress.

Morphological descriptors that include a taxon name were avoided. Specifically, we do not use “leontiniid basin,” which was defined by Colwell (1965) to note the region demarcated by the mesiolingual cingulum of the upper premolars. Although this convention has been followed by Villarroel and Colwell Danis (1997) and Shockey (2005), we regard it as undesirable since it implies a predetermined taxonomically diagnostic value (putting the cart before the horse) which, in this case, is inappropriate, since the complex of protoloph and mesiolingual cingulum manifests itself such that “leontiniid basins” also occur in some notohippids (e.g., Rhynchippus equinus [Ameghino, 1897] and Pascualihippus [Shockey, 1997]). Instead, we simply note the presence/absence of the mesiolingual cingulum (character #21), and code the various forms of the premolar protoloph that affect the appearance of the mesiolingual cingulum (chars. #23, #24, and #25).

For the purposes of exploring character states of basal notoungulates and for future higher-level phylogenetic studies, we did not exclude characters found to be universal among the notoungulates. Their ubiquity among notoungulates rendered them useless for the purposes of discovering phylogenetic relationships within notoungulates; however, such characters are not uninteresting, since they may be derived when compared to non-notoungulate outgroups. Fusion of the mandibular symphysis and absence of centrale, for example, were ubiquitous among our survey of notoungulates, but differ in outgroups. The relevance of these characters is noted in the discussion section.

1. Cheekteeth orientation: (0) strong anterior convergence of cheek teeth, resulting in the distance between left and right P1s being less than half that of the distance between left and right M2s, such that the upper dental arcade will appear V-shaped (fig. 9A); (1) no substantial anterior convergence of cheek teeth (the distance between left and right P1 is greater than half the distance between the M2s) with the anterior dental arcade U-shaped; (2) anterior dental arcade transverse and [-shaped. Modified from Cifelli (1993: #5 of the “advanced toxodont families” analysis) and Shockey (1997: # 6).

Cifelli's (1993) version of this character was defined as “(f)orm of anterior dental arcade” as “V-shaped” or “U-shaped.” With the exception of Colpodon, Cifelli (1993) scored all leontiniids as having V-shaped anterior dental arcades, whereas those of notohippids where characterized as “U-shaped.” Inexplicably, Cifelli (1993) scored all the toxodontids in his analysis as having V-shaped arcades, when in fact the anterior dental arcade of toxodontids is very blunt (not coming to a point), having a transverse dental arcade. Indeed, toxodont dental arcades deviate further from a V-shaped muzzle even more than do notohippids, which Cifelli (1993) characterized as U-shaped. The present character analysis thus includes three general forms of the anterior dental arcade: V-, U-, and [- shaped, the latter a shorthand designation of the transverse dental arcade of toxodontids and the notohippids Eomorphippus and Pascualihippus (in this convention the anterior muzzle is directed to the left in the [-shaped snout, whereas it is directed downward in the U-shape and V-shape). The U-shaped dental arcade is presumed to be the plesiomorphic form within Notoungulata, as it is common across most clades, and non-notoungulate ungulates.

2. Upper dental formula: (0) complete upper dentition, (1) C absent; (2) more than one upper tooth (the canine and another tooth). Modified from Villarroel and Colwell Danis (1997: #1).

3. Relative canine size: (0) caliber and height of canine greater than any of the incisors; (1) canine subequal in size to incisors, but none of the incisors are greatly enlarged; (2) one of the incisors is caniniform and larger than any canine.

4. I1 form: (0) incisiform (labiolingual compression, with transversely flat occlusal surface); (1) caniniform (ovoid or round in cross section with pointed cusp); (2) gliriform, much larger (broader and taller) than lateral incisors (if present).

5. I1 enlarged caniniform: (0) absent; (1) present.

6. I2 enlarged caniniform: (0) absent; (1) present.

7. Labial cingulum of upper incisors presence/absence: (0) present in all incisors except any that are hypertrophied; (1) absent.

Bond and López (1995) recognized the labial cingulum of the brachydont incisors as a diagnostic character for leontiniids. They included Colpodon in the Leontiniidae, based on the finding of incisive cingulae in a specimen they studied. The specimens we observed, however, lacked a cingulum, so Colpodon is scored as variable for this character (0/1).

8. Incisor root form: (0) incisors with determinate growth, roots form after crown formation; (1) incisors hypselodont (never form roots).

9. Diastema in anterior upper dentition: (0) absent, no significant gaps between anterior upper teeth; (1) present, large gap between anterior teeth and cheek teeth.

We ignored numerous and variable small gaps among teeth in scoring for the presence of diastemata. We regard a diastema as present only when a gap is larger than expected for a single tooth (typical of the region) to occupy.

10. Upper premolar root form: (0) upper premolars with determinate growth, roots form after crown formation; (1) premolars hypselodont, roots do not form at least one premolar is hypselodont.

11. Upper molar root form: (0) all upper molars with determinate growth, roots form after crown formation; (1) molars hypselodont, roots do not form.

12. Relative height of M1: (0) brachydont (hypsodonty index [HI] of M1 < 1); (1) mesodont (HI of M1 ≈ 1); (2) M1 hypsodont (HI > 1); (3) M1 euhypsodont (HI > 2) (adapted from Shockey, 1997: #11).

13. Cementum: (0) absent; (1) variably present or present in small amounts on one or more tooth; (2) thick coating of cementum is invariably present (adapted from Cifelli, 1993: #12).

As in other ungulates (e.g., horses and cows) the evolution of cementum-covered crowns appears to have occurred independently among several groups of notoungulates.

14. P1 orientation: (0) ectoloph in same plane as that of the other upper premolars; (1) ectoloph obliquely oriented, parastyle distinctly medial to the paracone.

15. Parastyle cingulum of P1–3 form: (0) may be separated from paracone ridge by a longitudinal valley, but does not form a sharp, distinctive mesiolabial “cingulum” that extends from the crown to the base of the tooth; (1) parastyle forms a sharp, distinctive mesiolabial “cingulum” that extends from the crown to the base of the tooth.

FIGURE9.

Selected dental characters of A, Anayatherium fortis, cast of holotype, alveoli of I3–P1 are exposed to show that A. fortis had a complete dental formula; B, m2 of Thomashuxleya externa (AMNH 28697) in occlusal (upper) and oblique (lower) views; and C, left m2 of Scarrittia canquelensis (AMNH 29592) in occlusal (upper) and lateral (lower) views. B and C illustrate the hypothetical evolution of the “paraconid” from the mesiolingual cingulid.

FIGURE10.

Hypothetical scapular transition series in Notoungulata. Not drawn to scale.

FIGURE11.

Hypothetical manual transition series in Notoungulata. All hands shown as left. Not drawn to scale. Abbreviations: Cun, cuneiform; Lu, lunate; Mg, magnum; Sc, scaphoid; Td, trapezoid; Tm, trapezium. Roman numerals indicate digit number.

FIGURE 12.

Hypothetical pedal transition series in Notoungulata showing left pes in dorsal view (upper row), left calcaneum in dorsal view (middle row), and left calcaneum in medial view (lower row). Illustrations are not to scale. Abbreviations are as follows: ast, astragalus; calc, calcaneum; cu, cuboid; ect, ectocuneiform; f, facet; fib, fibula; Mt, metatarsal; nav, navicular; sus, sustentacular; troch, trochlea.

16. Labial cingulum of P1–3 presence/absence: (0) present, lies at the base of the tooth between the paracone ridge and the metacone ridge; (1) absent.

17. Distolabial cingulum (“metastyle”) of P1–3 presence/absence: (0) absent; (1) present, distinctive longitudinal cingulum forms a sharp distal border of the ectoloph.

18. Cristae of upper premolar ectolophs: (0) separate cristae from ectoloph persist with modest wear, manifested by small loops in the labial side of the major fossette or enamel-lined pits between the fossette and ectoloph; (1) one or two cristae persist with moderate wear giving the central valley (or fossette) a branching appearance; (2) separation between cristae very shallow and quickly obscured by wear.

19. P2–4 protoloph: (0) robust, base inclined and extends to lingual extreme of tooth; (1) relatively small and recessed from lingual extreme, not inclined toward mesiolingual cingulum.

20. Upper premolar mesiolingual cingulae: (0) present; (1) absent. (Shockey, 1997: #16)

21. Upper premolar distolingual cingulum position: (0) distolingual cingulum far removed from the occlusal surface in little worn teeth, may be a little closer to crown than the mesiolingual cingulum; (1) distolingual cingulum close to crown (closer to crown than mesiolingual cingulum, if present) with tendency to come into wear early and produce mesiodistal bulge that extends more lingual than the protoloph; (2) tendency for distolingual cingulum to form distinctive “cup” (sensu Patterson, 1934.)

22. Upper premolar intermediate lingual cingulum extent: (0) no lingual cingulum, or if mesiolingual and distolingual cingulae are present they are not united by lingual (intermediate) cingulum; (1) mesiolingual and distolingual cingulae are united by intermediate lingual cingulum.

23. Upper premolar protocone lingual face groove: (0) absent, without vertical groove; (1) present, with vertical groove on lingual face (fig. 5; Patterson, 1934).

24. Mesiolingual face of upper premolar protoloph ridge: (0) absent, without ridge to mesiolingual cingulum; (1) present, with ridge to mesiolingual cingulum (separates the “leontiniid basin” in species having the ridge as well as mesio- and intermediate lingual cingulae [e.g., Leontinia gaudryi and Anayatherium fortis: fig. 9A]).

25. Labial cingulum of M1 presence/absence: (0) present; (1) absent.

26. Cristae of upper molar ectolophs, persistence with wear: (0) separate cristae from ectoloph persist with modest wear, manifested by variable numbers of small loops in the labial side of the major fossette or enamel-lined pits between the fossette and ectoloph; (1) one or two cristae persist with moderate wear giving the central valley (or fossette) a branching appearance; (2) separation between cristae very shallow and are quickly obscured by wear.

27. Dominant upper molar crista: (0) no single crista dominates; (1) first crista dominates, forming branch of anterior region of central valley.

28. M1 mesiolingual cingulae: (0) present, extends to “pseudohypocone”; (1) present, extends only to protoloph; (2) absent.

29. Posterior cingulum of upper molar form: (0) does not form distinct oval fossette with wear; (1) forms distinct oval fossette with wear.

30. Lower incisor orientation: (0) long axis of lower incisors directed relatively vertically, at an angle greater than 40° to long axis of mandible; (1) incisors procumbent, long axis of incisors more in line with long axis of mandible (angle to mandibular ramus <40°).

31. Third lower incisor (i3) size: (0) not enlarged relative to other incisors; (1) larger than other lower incisors.

32. Third lower incisor (i3) form: (0) incisiform; (1) caniniform or tusklike; (2) peglike, very small or absent.

In leontiniids and toxodontids, the i3 is larger than either of the mesial incisors and the canine. The cross section of the hypertrophied i3 is generally triangular in both leontiniids and toxodontids, but that of toxodontids tends to appear more spatulate (“incisiform”; Scott, 1912a) than the i3 of leontiniids.

33. Lower incisor labial cingulum presence/absence: (0) present; (1) absent.

34. Metaflexid of lower premolars shape: (0) simple without branching; (1) with bifurcations (adapted from Villarroel and Colwell Danis, 1997: #18).

35. Lingual cingulid of lower premolars: (0) distinct and continuous along the length of the tooth; (1) not distinct along metaconid, but appears anterior to metaconid and between metaconid and entoconid (or entolophid); (2) entirely absent (except for what may be manifest as a “paraconid” [see Discussion and figs. 9B, C]).

36. Labial cingulid of p2–p4: (0) present on p2–p4; (1) present on p2, but absent on p4; (2) absent on p2–p4. Modified from Villarroel and Colwell Danis (1997: #19) to include absence of premolar labial cingulid.

37. Lower molar “pseudoparaconid” presence/absence: (0) absent or minute; (1) presence of a substantial “pseudoparaconid” in a position similar to that of a paraconid, although it may be derived from the mesiolingual cingulum, rather than representing a true trigonid paraconid.

Cifelli (1993) noted the absence of the paraconid in early (Paleocene-Eocene) notoungulates and suggested that the lack (loss) of a paraconid is a synapomorphy for Notoungulata. Later notoungulates (e.g., leontiniids and many others) have a lophid that extends to the mesiolingual corner of the tooth suggestive of a paraconid-protoconid connection. Our observations of progressive occlusal wear in Thomashuxleya, however, suggest that this lophid originates from the mesial cingulid, which grades into the protoconid of Thomashuxleya and extends to the labial side and comes into occlusion as a functional lophid once occlusal wear has occurred (fig. 9B, C).

38. Lower molar metaconid morphology: (0) forms a simple, oblique crest; (1) possesses an accessory cusp.

Cifelli (1993) considered the presence of this accessory cusp a possible synapomorphy for the Isotemnidae, but noted its variable presence in isotemnids. We failed to find this cusp in our samples of Thomashuxleya externa.

39. Labial cingulid of m1 presence/absence: (0) present; (1) absent.

40. Labial cingulid of m2 presence/absence: (0) present; (1) absent.

41. Hypoconulid of m1–2 length: (0) short, extends little beyond the entolophid; (1) elongated, entolophid projects from near midpoint of talonid crest.

42. Molar entolophid fossettid presence/absence: (0) absent; (1) present. (Shockey, 1997: #21).

The entolophid fossettid develops between two transverse lophids that connect to the hypolophid (Loomis, 1914: fig. 52). Loomis (1914) noted these lophids and named the anterior of the two the “septum” and the posterior the “pillar.” These unite early in development to form the “entolophid,” but a persistent fossettid forms in notohippids (excepting Eurygenium and Morphippus), basal toxodontids (e.g., Proadinotherium, Adinotherium, and Nesodon), and all known leontiniids. This entolophid fossettid is also known as the “bay 3” fossettid (or “pit 3”) (Loomis, 1914) or the “posterior fossettid” (e.g., Soria and Alvarenga, 1989: fig. 1).

We avoid terms that designate relative position (e.g., number or “posterior”) since the absence of the related feature may be disorienting (e.g., a “second fossettid” appears absurd if there is no first, as does a fossettid posterior to the “posterior” fossettid). For example, the entolophid fossettid has been called the “posterior fossettid,” but a more posterior fossettid frequently forms via the uniting of the hypoconulid with the entolophid (e.g., Coresodon, [Loomis, 1914] as well as Pascualihippus and Rhynchippus brasiliensis [Shockey, 1997: fig. 2]).

43. Coracoid process shape: (0) distinctive, with recurved “beak”; (1) process present, but lacking a “beak”; (2) process indistinct or absent, scapular tuber small. (See fig. 10 regarding this and chars. 44–46.)

44. Acromion border extent: (0) extends distally to near the level of the glenohumeral joint or beyond; (1) limited to the region of the scapular body or scapular neck adjacent to the scapular body, but does not extend toward the region over the glenohumeral joint. Adapted from Meng et al. (2003).

Meng et al. (2003) regarded the condition of an “acromial plate” that overhangs the glenohumeral joint as being the primitive condition for eutherians. The acromial plate forms a fairly symmetric shield that suggests the contour of the proximal humerus. We regard the medial process of the shield as the acromion and the lateral one as the metacromion. (See #45.)

45. Metacromion extent: (0) extends distally to near the level of glenohumeral joint or beyond; (1) limited to the area above the region of the scapular body or the region of the scapular neck adjacent to the scapular body, but does not extend toward the region over the glenohumeral joint.

The spine of the scapulae of the isotemnids Thomashuxleya and Anisotemnus has paired acromial and metacromial processes that overlie the glenohumeral joint and appears to represent the plesiomorphic eutherian condition (Meng et al., 2003). The acromion and metacromion are located more proximally in most other notoungulates.

46. Secondary metacromion presence/absence: (0) absent; (1) present. The scapulae of the basal toxodontids (“nesodontine” toxodontids) Nesodon and Adinotherium have two metacromia (Scott, 1912a).

47. Greater tubercle height: (0) lower than the head of the humerus; (1) even with or higher than the head (modified from Rose and Lucas, 2000; and Meng et al., 2003).

48. Bicipital groove closure: (0) well enclosed by greater and lesser tubercles; (1) open, relatively wide space between greater and lesser tubercles.

49. Deltoid crest shape/position: (0) distinctive, extending to more than 2/3 the length of the humeral shaft; (1) distinctive, proximally placed; (2) no well defined crest or tuberocity for the deltoid muscles.

50. Pectoral crest shape/position: (0) distinctive, extending to more than 2/3 the length of the humeral shaft; (1) distinctive, proximally placed; (2) no well-defined crest or tuberocity for the deltoid muscles.

51. Tuberosity for m. teres major presence/absence: (0) present; (1) absent.

52. Supinator crest: (0) present, well developed, broad and bladelike; (1) inconspicuous (grades into shaft) or absent.

53. Entepicondylar process, medial extent: (0) extends medially with a dimension more than half that of the trochlear width; (1) medial extension short (dimension <40% that of the trochlea).

54. Entepicondylar foramen presence/absence: (0) present; (1) absent.

The presence of the entepicondylar foramen is generally regarded as the plesiomorphic condition for mammals (Landry, 1958). It is commonly absent (presumably lost) in terrestrial cursors, such as ungulates and canids. Its near ubiquity in notoungulates (present in isotemnids, and mesotheriids, hegetotheriids, interathertheriids), and presence in many eutherian outgroups, suggests that it, too, is plesiomorphic for Notoungulata. As far as known, it is absent only in the Toxodontia families Homalodotheriidae, Leontiniidae, Notohippidae, and Toxodontidae.-

55. Medial trochlear flange size: (0) large (Shockey and Flynn, 2007: fig. 2a); (1) small. (See Andersson, 2004, regarding function in felids.)

56. Brachial index (BI): (0) length of radius subequal to that of humerus (90 < BI < 100); (1) radius shorter than humerus (BI < 90); (2) radius longer than humerus (BI > 110).

57. Olecranon shape in lateral view: (0) major axis of olecranon the same as that of the ulnar shaft; (1) major axis of olecranon with upward curve (crouching stance morphology); (2) major axis of olecranon with downward curve (erect stance morphology).

Upward olecranon curvature is found in many extant climbing mammals (e.g., the binturong, Arctictis [see O'Leary and Rose, 1995: fig. 7], whereas downward curvature is associated with the erect stance of cursorial mammals (e.g., canids, Lama). This character varies among notoungulates, even among the few known Eocene forms (downward curvature in Thomashuxleya, upward curvature in Anisotemnus and Maxschlosseria; Shockey and Flynn, 2007).

58. Olecranon shape in dorsal (anterior) view: (0) major axis of olecranon in the same orientation as that of the ulnar shaft, (1) major axis of olecranon with distinct medial curvature (e.g., Trachytherus; Shockey et al., 2007).

59. Outline of proximal radial in proximal view: (0) oval; (1) subrectangular. Modified from Meng et al. (2003: #186).

The radial head of the arctocyonid Chriacus may represent a model of the plesiomorphic mammalian radial head (see Gebo and Rose, 1993; O'Leary and Rose, 1995). The radial head of Chriacus is oval and has a simple depression for the articulation with the humeral capitulum. The convex and ovoid ulnar facet of the proximal radius facilitates rotation of the radius, whereas such is restricted by the derived flat and subrectangular articulation (char. state 1). The derived form is exemplified by Phenacodus, in which the radial head has a transverse dimension more than 1.5 times that of the dorsoventral dimension. The fossa of the radial head of Phenacodus is undulating, rather than a simple concavity. (Compare drawings of the radial heads of Chriacus and Phenacodus in O'Leary and Rose, 1995: fig. 9).

60. Radial head fossa: (0) presence of simple fossa in proximal view; (1) depression of head is undulating. (See note above, #57.)

61. Capitular eminence (central process) of radial head presence/absence: (0) absent, the radial head is smooth or undulating, without distinctive capitular eminence; (1) present, with distinctive (sharp) capitular eminence of the proximal end on the dorsal side (pronated) of the radius.

A distinctive central process on (the “capitular eminence” of some workers (e.g., Gebo and Rose, 1993; Rose and Lucas, 2000) is generally regarded as being an indication of terrestrial locomotion and its absence an indication of climbing ability (Szalay and Sargis, 2001). However, Gebo and Rose (1993) noted that it is present in several extant scansorial and even arboreal mammals, thus discounting the notion that a capitular eminence inhibits rotary movements too seriously to permit climbing. It is more commonly present in terrestrial taxa (e.g., lagomorphs, Phenacodus, and Hyopsodus [Meng et al., 2003]).

62. Radial sesamoid presence/absence: (0) absent; (1) present.

The presence of the facet on the radius for the radial sesamoid may be used to score its presence/absence, thus this character (#60) may also be stated as the presence of a facet for radial sesamoid on the proximolateral surface of the radius in prone position: (0) absent or (1) present.

“Elbow” sesamoids are known to articulate with the dorsolateral surface of the radial head in several notoungulates, including the nesodontine toxodontid Nesodon (Scott, 1912a) and mesotheriid notoungulates, Trachytherus and Plesiotypotherium (Shockey et al., 2007). Scott (1930) deduced its presence also in Homalodotherium via the presence of the extra facet of the proximal radius, which was in the homologous position as that of Nesodon. This trait also is known in the pocket gopher Geomys (Shockey et al., 2007) and various didelphid and caenolestid marsupials (Szalay and Sargis, 2001, text and fig. 10b, c, f, g). The radial sesamoid (when present) in notoungulates differs from those of marsupials by being a more robust element that correlates with a large dorsolateral facet of the proximal radius, but without any evidence of contact with the humerus. Despite the relatively large size of the notoungulate elbow sesamoid, it may be easily separated from the joint postmortem or missed by collectors. Thus, its presence is most easily deduced by the sesamoid facet of the dorsolateral surface of the proximal radius.

63. Distal radius dorsal outline shape: (0) dorsal outline convex with no styloid process; (1) outline oblique, with no distinct styloid process; (2) outline with distinctive styloid process with facet for scaphoid and lunar on the ulnar side of the process.

A distinctive radial styloid process is absent in a variety of notoungulates, including isotemnids (Thomashuxleya, Anisotemnus), mesotheriids (Trachytherus, Mesotherium). Notohippids (e.g., Rhynchippus and Argyrohippus) have an oblique border of the distal radius, whereas toxodontids (e.g., Nesodon, Adinotherium) have a distinctive styloid process.

64: Distal ulna articulation with pisiform presence/absence: (0) present; (1) absent.

Chaffee (1952) noted the absence of ulnar-pisiform contact in Scarrittia and Homalodotherium, mentioning it as possible evidence of a close phylogenetic relationship between the two taxa.

65. Manus, digit number: (0) pentadactyl; (1) tetradactyl; (2) tridactyl (fig. 11).

66. Relative metacarpal/metatarsal lengths on digit III (Mc III/Mt III): (0) metacarpals subequal in length to metatarsals, with the Mc III-to-Mt III ratio not exceeding 1.5; (1) metacarpals relatively elongate compared to metatarsals, with Mc III/Mt III greater than 1.5.

Whereas small typothere notoungulates (e.g., the Santacrucian interatheriids and hegetotheriids) have longer metatarsals than metacarpals, a variety of larger notoungulates have longer metacarpals relative to their metatarsals. The Scarritt Pocket sample of Scarrittia metapodials yields a mean Mc III-to-Mt III ratio of 1.84 (n = 3) and suggests an allometry in which the relative metacarpal length increases with body size. The asymmetry of the metapodials is even greater in Homalodotherium, in which the metacarpals have greater than twice the length of the metatarsals (Mc III/Mt III = 2.26).

67. Radial facet of lunar extent: (0) extends over the dorsal surface of the lunar, to a region at least half way down the body of the lunar in dorsal view; (1) radial facet of lunar restricted to the proximal end of the element so that less than half of the dorsal surface of the lunar is devoted to the facet.

Since the time of Matthew (1937) the degree to which the radial facet covered the lunar has been used as an indication of the degree of wrist extension. In manual plantigrady, greater dorsal coverage of the radial facet is expected, whereas in digitigrade taxa the facet is more limited to the distal surface.

68: Unciform facet angle: (0) distal articulation transverse, with facet for Mc V (if present) at an angle less than 30° to Mc IV facet; (1) facet for Mc V at an angle nearly 45° to Mc IV facet.

69. Ungual phalanx of manus digits compression: (0) lateromedially compressed; (1) dorsopalmar compression, but distal region subequal in width to proximal; (2) hooflike phalanx, distal region wider than proximal.

70. Apex of ungual phalanx of manus, fissure presence/absence: (0) absent, distal apex pointed or rounded, without median fissure; (1) present, median fissure at distal apex.

71. Suprapatellar medial femoral ridge presence/absence: (0) absent; (1) present.

The femora of cf. Elmerriggsia fieldia and Anayatherium cf. A. fortis have distinctive medial ridges of the distal region of the femoral shaft, just proximal to the elongated medial ridge of the patellar groove (medial trochlear ridge). We have not observed this state in any other notoungulates, not even the leontiniid, Scarrittia, but the preservation of the region of the femora of this taxon is so poor that we cannot confidently score its presence or absence.

72. Number of digits of the pes: (0) pes pentadactyl, though Mt I may be reduced; (1) tetradactyl, Mt I absent; (2) tridactyl, Mt I absent and Mt V absent or reduced to tarsallike element. (See fig. 12 regarding chars. 72–83).

73. Dorsal prominence of calcaneum orientation: (0) oblique orientation; (1) orthogonal orientation.

74. Calcaneal fibular facet size: (0) small or absent; (1) large.

75. Calcaneal fibular facet form: (0) lenticular; (1) subquadrate; (2) wedge shaped, with proximal transverse dimension greater than the distal.

76. Ectal facet of calcaneum or astragalus orientation: (0) horizontal orientation; (1) steeply inclined articulation between calcaneum and astragalus.

77. Calcaneal-navicular articulation development: (0) absent; there is no evidence of a navicular facet of the calcaneum or a calcaneal facet of the navicular (there may be cuboastragalar contact [i.e., “alternating tarsus”]); (1) present; a small thin articular surface is present at the distal calcaneum that is not due to distal astragalar articulation but has evidence of being due to navicular contact; (2) well-developed navicular facet is present, indicating a strong “reverse alternating” tarsus.

A distinctive feature of the notoungulate tarsus is that it lacks cuboastragalar contact. In some taxa there is actually some to considerable overlap of the calcaneum with the navicular, a condition that Cifelli (1993) called a “reverse alternating tarsus.” This is in contrast to the cuboastragalar articulation of the typical “alternating tarsus.” Although it has been suggested that the “reverse alternating tarsus” may be a synapomorphy for advanced Toxodontia (Cifelli, 1993) we note that this state occurs in some typotheres (e.g., Protypotherium and Federicoanaya), as well as lagomorphs and arctostylopids (Shockey and Anaya, 2008).

78. Astragalar foramen (superior): (0) present, foramen is present within the fossa separating the astragalar trochlea with that of the flexor groove; (1) absent.

79. Groove for tendon of flexor hallicus longus (or other flexor tendon) development: (0) present; (1) united to the astragalar trochlea as a plantar component of the astragalar trochlea (Wang, 1993; Shockey and Flynn, 2007).

80. Neck of astragalus shape: (0) neck is well defined, being constricted behind the astragalar head and having a definable length between the head and astragalar body; (1) very short neck with little or no constriction behind the head, which lies very close to the body of the astragalus.

81. Astragalar head shape: (0) subspherical head that does not cover the lateral body in distal view (not expanded laterally); (1) ovoid, with head expanded laterally beyond the mid-point of the body in distal view; (2) teardrop shaped, due to lateral and plantar expansion of the head.

82. Proximomedial astragalar buttress of navicular presence/absence: (0) astragalar buttress present; (1) absent.

In archaic notoungulates and typotheres (except interatheriids) the navicular has a distinctive proximomedial process that articulates with the medial head of the astragalus. Its position is suggestive of the tibiale of rodents (Sinclair, 1909) and Arctocyon (Matthew, 1937), which may be the source of this distinctive navicular process (i.e., “tibiale” fused to navicular).

83. Symmetry of trochlea ridges of astragalus: (0) symmetrical, lateral and medial trochlear ridges subequal in height; (1) asymmetric, lateral trochlear ridge distinctly higher than medial.

PHYLOGENETIC ANALYSES

In order to generate phylogenetic hypotheses regarding the two new leontiniids described in this work and the placement of the Leontiniidae among the Notoungulata, a character-taxon matrix consisting of 27 notoungulates (including 12 leontiniids) and 83 craniodental and postcranial characters was analyzed in PAUP 4.0b10. Three subsets of this character-taxon matrices were derived from this larger matrix. The first two varied the taxonomic composition of the matrix, whereas postcranial characters were excluded from the third. The first analysis, the “Toxodontian” analysis, including only notoungulates traditionally referred to the Toxodontia (sensu Simpson, 1945; Cifelli, 1993) for the ingroup. The henricosborniid Colbertia magellanica served as the outgroup for all three analyses. The toxodontian ingroup taxa included leontiniids, toxodontids, notohippids, isotemnids, and a homalodotheriid. Otherwise, taxa traditionally referred to the Typotheria were excluded from this “Toxodontian” analysis. The second analysis, the “Typotheres included” analysis, included all the toxodontians plus five typotherian species (two interatheriids, a hegetothere, and two mesotheriids). All 83 characters were used in these first two analyses. The third analysis included all 27 notoungulates, but excluded all 41 of the postcranial characters. This was accomplished in order to estimate the influence of postcranial characters on the results of the first two analyses. A heuristic search was used in parsimony analyses in all analyses. Starting trees were obtained by stepwise addition with the tree bisection-reconnection branch-swapping algorithm.

RESULTS OF THE “TOXODONTIAN” ANALYSIS: The “toxodontian” analysis yielded 90 most parsimonious trees, each 156 steps long. The strict consensus tree is shown in figure 13. Indices (excluding uninformative characters) were as follows: Consistency index (CI) = 0.6054; homoplasy index (HI) = 0.3946; retention index (RI) = 0.7671; and rescaled consistency index (RC) = 0.4819. Synapomorphies of selected nodes in figure 13 are discussed below.

Colpodon Clade (Node A, fig. 13): The two species of Colpodon form a monophyletic clade nested within a monophyletic Leontiniidae, a group that in this analysis is shown to include only taxa traditionally regarded as leontiniids (Node C). The relatively thin, upper premolar (P2–4) protoloph, recessed from the mesiolingual cingulum (#19[1]) is an unequivocal synapomorphy for Colpodon. Equivocal synapomorphies include: absence (reduction: character state change 2→1) of caniniform incisor (#3[1]); mesiolingual face of upper premolar protoloph without connecting ridge (#24[0]); and absence of labial cingulum of lower incisors (33[1]).

“Tropical Clade” (Node B): The strict consensus resolves a “Tropical Clade” of leontiniids, sensu Villarroel and Colwell Danis (1997); however, the mid high-latitude species of Colpodon also are in this clade, more closely related to tropical Huilatherium than any are to the other tropical taxon (Taubatherium), whereas Colpodon grouped with Ancylocoelus in the Villarroel and Colwell Danis (1997) analysis. No unequivocal synapomorphies define this clade, but equivocal synapomorphies include: absence of canines (#2[1]), absence of cingulum of P1–3 (#16[1]), and absence of cingulid of p2, but its presence on p4 (#36[1]). No monophyletic “Patagonian” clade (sensu Villarroel and Colwell Danis, 1997) is resolved in the “Toxodontian” analysis.

FIGURE 13.

Cladogram of the “Toxodont” analysis. Strict consensus (left) and majority 50% consensus of the “Toxodont” phylogenetic analysis. Nodes: A, Colpodon; B, “Tropical clade”; C, Leontiniidae; D, “advanced Toxodontia”; E, Toxodontidae; F, “notohippid” + Toxodontidae. Numbers at nodes represent percent support when support is less than 100%.

Leontiniidae (Node C): This node includes only taxa that traditionally have been regarded as leontiniids (sensu Simpson, 1945; Villarroel and Colwell Danis, 1997). Unequivocal synapomorphies diagnosing this clade (and thus, the Leontiniidae) include: strong anterior convergence of the muzzle giving it a V-shaped appearance in palatal view (#1[0]); caniniform i3 (#32[1]); and medial ridge of the cranial face of the distal region of the femur (#71[1]); and a wedge-shaped fibular facet of the calcaneum (#75[2]). Equivocal synapomorphies for leontiniids include: caniniform 12 (#6[1]); enlarged i3 (#31[1]); and branched metaflexid of lower premolars (#34[1]).

The two Eocene taxa (Coquenia and Martinmiguelia) together with the late Oligocene Elmerriggsia appear as an unresolved polytomy of taxa basal to the remaining leontiniids. No unequivocal synapomorphies unite the remaining leontiniids, but three equivocal synapomorphies weakly suggest such a relationship. These are the tendency for the posterior upper premolar cingulum to rest high on the crown (#17[1]); formation of an intermediate lingual cingulum of the upper premolars (#22[1]); and the formation of a ridge along the mesiolingual face of the upper premolar protoloph (#24[1]).

FIGURE 14.

Cladogram of the “Typothere included” analysis. Strict consensus (left) and majority 50% consensus of the “Typothere included” phylogenetic analysis. Nodes: A, Colpodon; B, “Tropical clade”; C, Leontiniidae; D, “advanced Toxodontia”; E, Toxodontidae; F, “notohippid”+Toxodontidae (Eurygenium excluded); G, “advanced Toxodontia” plus interatheriine interatheriids. Numbers at nodes represent percent support when support is less than 100%.

The presence and placement of the enlarged caniniform upper incisor appears to have some diagnostic value. An enlarged caniniform 12 (#6[1]) appears in the basal leontiniid taxa, as well as Leontinia and Ancylocoelus. A Scarrittia-Anayatherium clade is evidenced by the I1 being the enlarged caniniform incisor (#5[1]). Colpodon may be diagnosed by the apomorphic reduction or absence of a large caniniform upper incisor (#3[1]), an apparent loss, changing from #3[2] as implied by its near outgroups having enlarged upper incisors.

“Advanced Toxodontia” (sensu Cifelli, 1993) (Node D): This analysis is consistent with that of Cifelli (1993) in which leontiniids, notohippids (including Eurygenium), and toxodontids formed a monophyletic group that excluded homalodotheriids and isotemnids, referred to by Cifelli as “advanced Toxodontia.” The numerous unequivocal synapomorphies for the “advanced Toxodontia” identified in this study include: elongated oval fossette between metaloph and distal cingulum of upper molars (#29[1]); absence of the entepicondylar foramen of the humerus (#53[1]); tetradactyl manus (#65[1]); subquadrate form of fibular facet of the calcaneum (#75[I]); “reverse alternating tarsus” with conspicuous articular facets between the navicular and calcaneum (#77[2]); absence of superior astragalar foramen (#78[1]); flexor hallicus longus groove continuous with astragalar trochlea (#79[2]); absence of constricted “neck” of astragalus (#80[1]); transversely elongated ovoid astragalar head (#81[1]); and absence of astragalar “buttress” of the navicular (#82[1]). Equivocal synapomorphies include: placement of the distal cingulum high on the crown of upper premolars (#21[I]); elongated hypoconulid (#41[1]); and presence of entolophid fossettid of lower molars (#41[1]); and supinator crest of humerus small or absent (#52[1]).

The “advanced Toxodontia” (Node D) in this analysis contains two subclades, one that includes the leontiniids (Node C) and another that includes toxodontids (Node E) plus taxa traditionally classified as notohippids. The toxodontids (Node E: Nesodon and Adinotherium) are recognized by three unequivocal synapomorphies and at least three equivocal derived traits (below). In this analysis, “notohippids” are nonmonophyletic (paraphyletic) as detailed below (fig. 13). Eurygenium lacks the derived characters that diagnose Node F, which includes the remaining notohippids (Rhynchippus spp. and Argyrohippus) the toxodontids. Thus, it is excluded from the unnamed clade at Node F.

Toxodontidae (Node E): The transverse muzzle (#1[2]), hypselodont incisors (#8[1]), and hypselodont molars (#11[1]) appear as unequivocal synapomorphies for the toxodontids Nesodon and Adinotherium. Equivocal synapomorphies include: procumbent lower incisors (#30[1); enlarged i3 (#31[1]); and the distal radius with a well-developed styloid process (#63[2]). The secondary metacromion of the scapulae of Nesodon and Adinotherium (#46[1]) may also be a synapomorphy for the Toxodontidae, however, the scapula is so poorly known in the related “notohippids” that the presence/absence of this character state cannot be assessed in these taxa.

Unnamed clade (Node F): The monophyletic Toxodontidae are nested among paraphyletic “notohippids,” with the latter in a polytomy as the nearest outgroup to the toxodontids. Unequivocal synapomorphies uniting the “notohippids” Rhynchippus spp. and Argyrohippus with toxodontids (to the exclusion of Eurygenium) include: first crista of upper molars distinct and persistent (#27[1]); articular surface of radial head subrectangular and undulating (#59[1] and #60[1]); tridactyl manus (#65[2]); Mc V facet of unciform at sharp angle to Mc IV facet (#68[1]); and tridactyl pes (#72[2]). Eurygenium lacks the half dozen unequivocal synapomorphies that characterize Node F, which includes the remaining taxa traditionally considered to be notohippids (Rhynchippus and Argyrohippus) plus the toxodontids; thus, it is excluded from the clade at Node F.

Other results from “Toxodontian Analysis”: The strict consensus analysis did not demonstrate any special relationship of Homalodotherium with the “advanced Toxodontia.” Instead, Homalodotherium and the isotemnids (Anisotemnus and Thomashuxleya) formed two branches of the tritomous toxodontian ingroup, with the “advanced Toxodontia” forming the third. The majority rule consensus analysis, however, placed Homalodotherium as sister to the “advanced Toxodontia” in 78% of the trees. When only dental characters were considered (postcranial characters deleted), Homalodotherium nested among the leontiniids.

The results of an Adams consensus analysis of the “Toxodontian matrix” were identical to those of the strict consensus (fig. 13). The only differences between these analyses and those of a 50% majority-rule consensus were that the majority-rule consensus suggested special relationships between Coquenia and Martinmiguelia as well as Leontinia and Ancylocoelus. The majority-rule consensus also resolved a monophyletic Rhynchippus (R. equinus and R. pumilus), and resolved the polytomy at Node F, placing Argyrohippus as the nearest outgroup to toxodontids.

RESULTS OF THE “TYPOTHERES INCLUDED” ANALYSIS: The “Typotheres included” analysis was based on the same character matrix as the “Toxodontian Analysis,” with five additional species included (two interatheriids, two mesotheriids, and a hegetothere). This analysis yielded 2635 most parsimonious trees, each 195 steps long. The strict consensus is shown in figure 14. Indices (excluding uninformative characters) were: CI = 0.5260; HI = 0.4740; RI = 0.7641; and RC = 0.4079.

Leontiniidae (Node C. fig. 14): The strict consensus of the “typotheres included” matrix places all the traditional leontiniid taxa within a monophyletic group, but otherwise shows little resolution of relationships among leontiniids (fig. 14). These taxa form a polytomy with the exception of the pairing of the two species of Colpodon, and the linkage of the two Anayatherium species in a polytomy with Scarrittia (fig. 14). When a 50% majority-rule consensus is applied, monophyletic Patagonian versus Tropical + Colpodon subclades are resolved. Unlike for the “toxodont” analysis, the 50% majority-rule consensus of this analysis identifies both groups as monophyletic (the Patagonian taxa were paraphyletic to a Tropical + Colpodon clade in the “toxodont” analysis of fig. 13).

The synapomorphies uniting the leontiniid taxa are the same as in the “Toxodontian analysis,” except that the mesodont condition of leontiniids equivocally appears as a reversal (#12[1], changed from the near outgroup condition of #12[2], hypsodont to mesodont).

Interatheriidae + “advanced Toxodontia” clade (Node G): The most unconventional result of this analysis is that the typothere taxa are paraphyletic to the “advanced Toxodontia.” This is due to several postcranial characters that the interatheriine interatheriids (Federicoanaya and Protypotherium) share with “advanced toxodontians” (Eurygenium, other “notohippids,” toxodontids, and leontiniids). Unequivocal synapomorphies uniting the interatheriids with Eurygenium, leontiniids, the “notohippids” Rhynchippus and Argyrohippus, and toxodontids include: tetradactyl manus (#65[1]); quadrate fibular facet of the calcaneum (#75[1]); calcaneal-navicular contact (#77[1]); and union of the groove for the tendon of the flexor hallucis longus with the astragalar trochlea (#79[1]). Steeply inclined ectal facet of the astragalus and calcaneum (#76[1]) is an equivocal synapomorphy of this group (the character state in leontiniids is #76[0]).

Inclusion of the typotheres yields Eurygenium as the earliest diverging advanced “toxodontian,” as the nearest outgroup to leontiniids plus the clade of the remaining “notohippids” + toxodontids. Unequivocal synapomorphies that unite Eurygenium with the other toxodontians include: absence of entepicondylar foramen of the humerus (#54[1]); short astragalar neck (#80[1]); transversely elongated astragalar head (#81[1]); absence of astragalar buttress of the navicular (#82[1]); and formation of fossette of the molars between the metaloph and the posterior cingulum (#29[1]).

In this analysis members of “advanced Toxodontia” are united by the unequivocal synapomorphies of oblique outline of distal radius (#62[1]) and well-developed articular facets for navicular-calcaneum articulation (#77[2]). Equivocal synapomorphies include enlarged i3 (#30[1]) and entolophid fossette of lower molars (#41[1]).

RESULTS OF “POSTCRANIALS EXCLUDED” ANALYSIS: The analysis that excluded the postcranial characters yielded 31 most parsimonious trees, each 105 steps long. Indices (excluding uninformative characters) were as follows: Consistency index (CI) = 0.5049; homoplasy index (HI) = 0.4951; retention index (RI) = 0.7927; and rescaled consistency index (RC) = 0.4077.

A strict consensus tree of this modified matrix resolves two major ingroup nodes, one of which contains a monophyletic Typotheria (mesotheriids, interatheriids, and a hegetothere) and a monophyletic Toxodontia composed of “notohippids” and toxodontids, but not the leontiniids. In this “postcranials excluded” analysis, leontiniids nested within isotemnids and the homalodotheriid.

COLPODON ANTUCOENSIS : BIOGEOGRAPHY AND AGE

The discovery of Colpodon antucoensis, new species, at Laguna del Laja, Chile, extends the geographic range of Colpodon. Species of Colpodon were previously known only from various localities in Chubut and Rio Negro provinces of central Patagonia, Argentina. This geographic range extension is modest (see map, fig. 1). At about 37.5° S and 71.3° W, Laguna del Laja is roughly 360 km northwest of the Colhuehuapian Paso Cordoba locality of Rio Negro, Argentina, which previously had been the northernmost record of Colpodon.

The presence of Colpodon at the Estero Correntoso locality of Laguna del Laja initially suggests a Colhuehuapian age for these lower horizons of the Cura-Mallín Formation (Tcm₁). Indeed, Colpodon has been used as the index fossil of the Colhuehuapian since Ameghino first recognized the beds as the “couches à Colpodon” (Ameghino, 1902). However, consideration of radioisotopic ages available for the lower Cura-Mallín and for fossil taxa coeval to the holo-type of C. antucoensis, as well as noting that this is a new species of Colpodon distinct from Patagonian forms, suggests that caution is warranted in assigning a faunal age to Tcm₁.

Horizons bracketing the type specimen of C. antucoensis yield ³⁹Ar/⁴⁰Ar ages of 19.25 ± 1.22 Ma above and 19.53 ± 0.60 Ma below (fig. 4), thus constraining its age at ∼19.5 Ma (corrected to ∼19.8 Ma for the revised Fish Canyon Tuff age). The age of the Colhuehuapian SALMA is best constrained by radioisotopic and paleomagnetic studies at its type locality at the Gran Barranca (Ré et al., 2010a, 2010b). Ré et al.'s preferred paleomagnetic interpretation (using the calibration of radioisotopic ages) from the Colhue-Huapi Member is that the “Lower Fossil Zone,” producing the Colhuehuapian SALMA fauna, is in Chron C6An.1n (Ré et al., 2010b). This suggests an age of 20.0 Ma-20.2 Ma for the Colhuehuapian SALMA, although its full temporal span remains unclear (tentatively estimated at 19–21 Ma by Flynn and Swisher, 1995). This ∼20 Ma age estimate is only slightly older than the age (∼19.8 Ma) of C. antucoensis in the Tcm₁ of Laguna del Laja.

Overlying the Colhuehuapian SALMA fauna, in the “Upper Fossil Zone” of the Colhue-Huapi Member at Gran Barranca, is a fauna recently recognized as being “Pinturan” (Kramarz et al., 2010), an informal biochron considered to be intermediate in age between Colhuehuapian and Santacrucian SALMA faunas (Vucetich et al., 2005; Carlini et al., 2005). Ré's preferred interpretation of radioisotopic-paleomagnetic results in that section is that the “Upper Fossil Zone” containing the “Pinturan” fauna lies within C6n, suggesting an age range from about 19.7 to 18.7 Ma (Ré, 2010b). Recognizing that “Pinturan” localities outside of Gran Barranca have yielded younger radiometric ages (∼17.5 Ma; Fleagle et al., [1995]), Kramarz et al. (2010) used the younger extreme of the age for Chron 6n as the older extreme of the suggested “Pinturan” age, proposing that the “Pinturan” ranged from 18.75 to 16.5 Ma. Croft et al. (2007) extended the beginning of the Santacrucian SALMA to be as old as ∼19 Ma, based on dates associated with the Chucal Fauna of northern Chile (correlated to the Santacrucian). This suggests either that the “Pinturan” is a very short time interval preceding the Santacrucian SALMA (consistent with the results of Ré, 2010b, but not with the younger estimates of Fleagle et al., 1995, and Kramarz et al., 2010), or that it falls within the Santacrucian SALMA.

The few other taxa identified to date from the same interval in the Estero Correntoso section (Tcm₁) at Laguna del Laja suggest a post-Colhuehuapian age for Colpodon antucoensis. For example, a lagostomine chinchillid is known from Tcm₁ of Estero Correntoso and is otherwise known only from Santacrucian or younger faunas (Flynn et al., 2008). A sloth from Estero Correntoso is referable to Nematherium (cf. N. angulatum or sp. nov; Flynn et al., 2008: fig. 4d). Nematherium is best known from the Santacrucian, but it also occurs in the “Pinturas” fauna of the Upper Fossil Zone of the Colhue-Huapi Member of Gran Barranca (Kramarz et al., 2010). The only taxon that occurs at both Estero Correntoso and in a known Colhuehuapian age fauna, aside from Colpodon, is Protypotherium sp. However, the age range of this interatheriid notoungulate is so great (extending from the Colhuehuapian and “Pinturan” up into the Chapadmalalan and possibly younger [Flynn et al., 2008]) that it has no utility in helping to discriminate time based on faunal similarities.

Thus, the radioisotopic age and taxa coeval with Colpodon antucoensis suggest that this species of Colpodon is possibly a little younger than the faunal interval that once carried its generic name—“couches à Colpodon.” Since Colpodon is the only known post-Deseadan leontiniid of southern South America, its temporal range extension also extends the temporal record of leontiniids in southern South America. Only the Laventan-aged Huilatherium of northern South America (Colombia) is younger (Villarroel and Colwell Danis, 1997).

PHYLOGENETIC CONSIDERATIONS

“TROPICAL” vs. “PATAGONIAN” CLADES: Our phylogenetic analyses provide weak support for the “Tropical” clade of leontiniids proposed by Villarroel and Colwell Danis (1997). The composition of that clade, however, differs in this study in that higher-latitude Colpodon is included in a clade with the “tropical” Taubatherium and Huilatherium (from Brazil and Colombia, respectively). All remaining leontiniids, including the low-latitude species of Anayatherium and the extra-Patagonian Mustersan (late Eocene) Coquenia and Martinmiguelia, are paraphyletic with respect to the “Tropical” clade in all our analyses, except in the 50% majority rule consensus of the “Typotheres included” analysis. In that latter analysis both a “Tropical” clade and a second monophyletic high-latitude clade share a similar composition to the “Patagonian” clade of Villarroel and Colwell Danis (1997). Three equivocal synapomorphies support the monophyly of the “Patagonian” clade in this 50% majority-rule consensus. Two of these appeared more frequently in other taxa than in the “Patagonian” clade. Thus, this lack of compelling evidence for a monophyletic “Patagonian” clade and the fact that half the taxa that form the group are not from Patagonia renders the concept unuseful. We consider it more likely that the “Tropical” taxa plus species of Colpodon form a natural group. Such a clade, with Colpodon nested among the lower-latitude near outgroups, Taubatherium and Huilatherium (Brazilian and Colombian taxa respectively), suggests an extra-Patagonian origin of Colpodon.

DENTAL PLASTICITY

It is of interest that an enlarged caniniform incisor is nearly ubiquitous in leontiniids, but the particular upper incisor that is enlarged is variable, although i3 is invariably the enlarged lower incisor. The enlarged, sometimes tusklike, upper incisor is the I2 in Coquenia, Martinmiguelia, Leontinia, Elmerriggsia, and Ancylocoelus, but I1 is enlarged instead in Scarrittia and Anayatherium.

Cifelli (1993: #1[1]) defined the character state of enlarged incisor, regardless of its particular locus (I1 or I2), in his phylogenetic analysis of “advanced Toxodontia.” Implicit in the use of such a character, where it does not matter which particular incisor has the derived state, is the assumption that the enlargement (of the anteriormost “incisiform”) is homologous even though the feature manifests itself at a different tooth locus. Two competing arguments can be made regarding such a definition of a derived state: (1) since the derived form occurs at two different tooth loci they should not be regarded as homologous, but rather as independent transformations from the plesiomorphic form, or alternatively, that (2) they should not be assumed to be independent, since the structures in question are close (adjacent) serial homologs, and the developmental influences that affect one locus do not need to be “reinvented” to affect another. In the latter scenario, a small change in the way in which the derived developmental process is executed may manifest itself at a slightly different location, and it is the inferred developmental change that is homologous.

Such “frame shifts” have been documented in development and evolution, (e.g., the putative shift in the manual digits of nonavian dinosaurs to those of birds [Wagner and Gauthier, 1999]). We regard it as both biologically plausible and most parsimonious to conclude that the gene expression for an enlarged upper incisor shifted from I2 to I1 in the common ancestor of Anayatherium and Scarrittia. The alternative hypothesis (that the two character states are independent) requires the independent reduction of I2 and a simultaneous (or nearly so) independent evolution of a large I1 in Anayatherium and Scarrittia. The nesting of Anayatherium-Scarrittia species among leontiniids having enlarged I2, when either means of scoring was used, supports this assertion.

POSTCRANIAL CHARACTERS IN THE PHYLOGENETIC ANALYSIS

The inclusion of postcranial data strongly influenced the tree topology. For example, when we excluded the postcranial characters, leontiniids and the homalodotheriid appear as sister taxa nested among isotemnids forming a clade, separate from the “advanced Toxodontia” and typotheres. Such a leontiniid, homalodotheriid, isotemnid group is reminiscent of the Entelonychia of Ameghino (1895). We note that, indeed, the dentitions of homalodotheriids and leontiniids are nearly identical. The relative height of the posterior cingulum is about the only difference between cheek teeth in the two groups (Patterson, 1934). The incisors, too, are nearly identical to each other, with both homalodotheriids and leontiniids having somewhat caniniform incisors with strong labial and lingual cingulae. Scott (1912b), however, expressed serious doubts regarding a close relationship between leontiniids and homalodotheriids, due to differences in the tarsus in each. He noted that a proximal tarsus referred to Leontinia gaudryi by Gaudry (1906; denoted as “Colpodon gaudryi”) was similar to that of the toxodontid Nesodon, but quite unlike that of Homalodotherium. The results of our analyses with postcranials strongly support Scott's rejection of a close homalodotheriid-leontiniid relationship.

Also, our analyses demonstrate that several characters of the postcranial skeleton are useful in recognizing taxonomic groups of notoungulates. Several of these of the scapula, manus, and pes are illustrated in figures 10, 11, and 12.

A hypothesis for a morphological transition series relevant to the leontiniids involves the pes (fig. 12). The “advanced Toxodontia” (sensu Cifelli, 1993; = Leontiniidae + “Notohippidae” + Toxodontidae) are characterized by orthogonal orientation of the dorsal prominence of the calcaneum (#73[1]), robust fibular facet (#74[1]) with quadrate form (#75[1]), and (excepting leontiniids) by steeply inclined ectal articulation of the proximal tarsals (#75-6[1]). These character states also occur in interatheriids (see also Shockey and Anaya, 2008) and are evident in the interatheriine interatheriids Federicoanaya and Protypotherium analyzed in this study. Additionally, the navicular and calcaneum make contact in all these taxa (reverse alternating tarsus, sensu Cifelli, 1993) and a distinctive facet develops in leontiniids, toxodontids, and some “notohippids” (e.g., Pascualihippus [Shockey and Anaya, 2008], but not Eurygenium [this study]). Leontiniids differ from interatheriids, “notohippids,” and toxodontids by their less inclined ectal facets of the calcaneum and astragalus; that is the astragalus more strongly overlies the calcaneum, where it is known in leontiniids. The contact between the two proximal tarsal elements of interatheriids, “notohippids,” and toxodontids may be characterized as side by side, rather than overlapping. The foot of the interatheriids Federicoanaya and Protypotherium is distinct from those of other “typotheres” (Shockey and Anaya, 2008). Like the traditional “advanced Toxodontia” their upper ankle joint has strong fibular-calcaneal contact resulting in a large, orthogonal fibular facet of the calcaneum. The upper ankle joint of the mesotheriids (Trachytherus, Mesotherium) and Hegetotherium have weaker fibular-calcaneal contact, oblique dorsal prominence of the calcaneum, and the astragalus overlies the calcaneum, with their ectal facets more horizontal than those of the interatheriids, “notohippids,” and toxodontids (fig. 12).

“UNINFORMATIVE” CHARACTERS

Some characters identified in this study were invariant in Notoungulata, and thus “uninformative” in terms of their utility in estimating phylogenetic relationships among notoungulates, so were not discussed in the final analysis. However, these characters may prove useful in resolving interrelationships among a broader suite of endemic South American “ungulates” and other eutherians. Examples of these characters include the fused mandibular symphysis, the absence of separate centrale of manus, and the absence of any “alternating” tarsus; that is, one in which the astragalus contact the cuboid. We also found a fused mandibular symphysis to be ubiquitous among notoungulates, occurring also in groups not included in our cladistic analysis (due to lack of postcranials) including notostylopids, henricosborniids, and oldfieldthomasiids. However, this derived trait may not be a synapomorphy for only Notoungulata, as we also found mandibular symphyses invariably fused in astrapotheres (Astrapotherium, Parastrapotherium, Trigonostylops) and litopterns (e.g., Theosodon, proterotheriids). Among South American native ungulates, it is unfused only in Pyrotheria (Pyrotherium) and Xenungulata (Carodnia). Likewise, where known, all notoungulates lack a centrale of the wrist, and every scaphoid we encountered had a magnum process that may represent the fusion of the centrale to the scaphoid. Further, we did not encounter any notoungulate having an “alternating” tarsus (i.e., none showed cuboastragalar contact). The tarsals were either serial (e.g., Trachytherus) or had the “reverse alternating” form (sensu Cifelli, 1993) in which the calcaneum and navicular make contact (e.g., toxodontids and leontiniids; fig. 12). Among the extinct, endemic South American ungulates, a robust alternating tarsus is found among pyrotheres (Shockey and Anaya, 2004).