Postcranial anatomy of the early notoungulate Allalmeia atalaensis from the Eocene of Argentina

We redescribe and interpret the postcranial elements of Allalmeia atalaensis Rusconi, found in the Divisadero Largo Formation, Mendoza, Argentina. Allalmeia is one of the few basal notoungulates with preserved postcranial remains. Rusconi initially documented a skull, several vertebrae and part of the appendicular skeleton but these specimens were subsequently lost. Recently though, distal components of the holotype humeri, the proximal portion of the right femur, several broken diaphyses and most of the left and right pes (MCNAM-PV 507), were relocated. Preparation of this material has revealed new details, especially from the plantar side of the feet. The calcaneum has a distal peroneal process with marked rugosity in the lateral end, a well-developed calcaneal plantar tubercle and a large sustentaculum tali. The naviculars exhibit a well-developed medial tuberosity and slightly contact the calcaneum on the dorsal side, a condition called a ‘reverse alternating tarsus’. The disposition and morphology of the cuneiforms are similar to those of more recent typotheres and the archaic ungulate Tetraclaenodon puercensis. The phalangeal rows are oblique to each other, as in some extant digitigrade mammals. The ungual phalanges are claw-like suggesting that they might have borne claws. Long bones are gracile in comparison with other notoungulates, but with well-developed muscle origins and insertions. The use of body-mass allometric equations and a comparison with extant analogues suggests that Allalmeia was a small, generalized, digitigrade animal. An estimated body mass of approximately 3 kg is consistent with oldfieldthomasiids and archaic ungulates.

NOTOUNGULATA was the most diverse group of South American native ungulates, incorporating at least 13 families, six of which were already differentiated by the early-middle Eocene. The notoungulate clade includes herbivores of various adaptive types, many of which are convergent to Holarctic mammals (Simpson 1978, López et al. in press). Hypotheses on the relationships within Notoungulata, and their placement relative to other mammals, have been derived largely on the basis of dental and cranial features (Billet 2010), or in conjunction with only a few incompletely documented postcranial traits (Cifelli 1993, Shockey et al. 2012. Basal notoungulates have rarely been found with dental and postcranial remains in close association. Examples include the notostylopid Notostylops (Bergqvist & Metello 2011, Lorente et al. 2013, the isotemnids Thomashuxleya, Anisotemnus and Pleurostylodon (Simpson 1967, Shockey & Flynn 2007 and the oldfieldthomasiid Colbertia (Bergqvist & Bastos 2009). Consequently, the appendicular morphology, locomotor adaptations and phylogenetic relationships of this group are poorly understood.
Allalmeia atalaensis is a basal notoungulate recovered from the Divisadero Largo Formation in Mendoza province (central western Argentina; López 2008). The faunal assemblage recovered from this unit was used to define the Divisaderan South American Land Mammal Age (SALMA; Pascual et al. 1965) and was originally characterized by taxa with a mix of generalized and derived morphologies. The particular nature of this association was considered to fill the gap between the Mustersan and Deseadan SALMAs during the late Eocene (Pascual et al. 1965). However, the Divisaderan is now no longer recognized and could be considered a local fauna probably equivalent in age to the Vacan or intercalated between the Vacan and Barrancan (López 2010).
The lower and middle levels of Divisadero Largo Formation have yielded a vertebrate fauna composed essentially of mammals and a few reptiles (i.e., turtles, crocodiles and boas). Mammals comprise more than 95% of the fossils collected from these levels and, remarkably, almost all of them can be referred to South American native ungulates (López 2010). Postcranial remains are uncommon in this assemblage and A. atalaensis is the only species recognized from both dental and postcranial remains.
Allalmeia atalaensis was named by Carlos Rusconi (1946a), who documented the skull, isolated vertebrae and part of the appendicular skeleton (MCNAM-PV 507). However, these fossils, and others housed in Museo de Historia Natural de Mendoza (now Museo de Materials and methods MCNAM-PV 507. Part of the holotype of Allalmeia atalaensis. The originally documented skull, vertebrae, a partial scapula, ulna, a patella, right tibia, left femur, left astragalus and left calcaneum are still missing. The relocated elements described here include: the distal epiphysis of both humeri, the proximal half of the right femur, the distal epiphysis of the left fibula, several broken diaphysis and most of the foot bones. The pedal components were cleaned and separated from the rock matrix. The tarsals, metatarsals and phalanges were retained in their articulated position. Most of the glue present in the upper ankle joint of the right foot was removed and cleaned by technical staff of Museo de La Plata. This procedure exposed the trochlea of the right astragalus and a broken bone, which might be a fragment from the calcaneum. Measurements were taken with a point digital caliper (Schwyz) and expressed in millimetres. Locomotor patterns were described following Samuels & Van Valkenburgh (2008). Bone topological names are written in Latin following the Nomina Anatomica Veterinaria (Schaller 1996) attached to their colloquial names. We compare A. atalaensis with extant and fossil remains housed in different museums mentioned below, and also with published data of other notoungulate skeletons. Orientation of the bones was based on the proposal of Sisson & Grossman (1982).
Body-mass estimations. The mass inferences for Allalmeia and extant mammals of similar size were inferred, depending of the available material, following allometric equations from Janis (1990) and Scarano et al. (2011) for dental measures, and Scott (1990) and Tsubamoto (in press) for postcranial parameters. Dental dimensions of the type material are from Rusconi (1946c), as are the length of the humerus, femur and the femoral diameter. Additional dental measures are from López (2008). For each equation we obtained a range of mass values, their mean and standard deviation. Following the analyses by Elissamburu (2012), the frequency distribution of the corporal mass was evaluated in order to avoid under and overestimation of mass.
Type locality. Divisadero hill, 8 km west of Mendoza city, Mendoza province, Argentina.
Age and stratigraphy. All specimens derive from the middle Eocene, Divisadero Largo Formation. This unit is exposed as a NNE-SSW-oriented ledge extending from the Divisadero Largo hill up to the Papagallos locality, approximately 2200 m long and 60-250 m wide. The Divisaderan SALMA was originally proposed as an Eocene biochronological unit (Pascual et al. 1965), older than the Tinguirirican and younger than the Mustersan. This unit was originally characterized by the coexistence of taxa with non-specialized morphology (e.g., Allalmeia atalaensis, Brachystephanus postremus, Xenostephanus chiottii, Acamana ambiguus Simpson, Minoprio & Patterson, 1962) together with taxa with more specialized characters (e.g., Ethegotherium carettei Simpson, Minoprio & Patterson, 1962, Trachytherus? mendocensis Simpson & Minoprio, 1949 comparable with closely related taxa in Deseadan and post-Deseadan faunas (Oligocene). Cerdeño (2007) noted the similarity between the Divisadero Largo and the Mariño mesotheriids. Subsequently, López (2008) showed that the two last species do not come from beds in the Divisadero Largo Formation but rather come from the overlying Mariño Formation. A detailed study of the Divisaderan fauna revealed that it has more affinities with faunas of Casamayoran Age (Vacan and Barrancan) than with those of Mustersan Age (Cerdeño et al. 2008, López 2008, López & Manassero 2008. Consequently, the Divisaderan SALMA is no longer considered to be a valid unit (López 2008(López , 2010.

Description
Humerus. Only the distal epiphyses of both right and left humeri are preserved. The epiphysis of the right humerus is the most complete (Fig. 1). It has a supratrochlear foramen (Foramen supratrochleare) and an epicondylar foramen (Foramen supracondylare). These characters are variable in more recent typotheres, as Protypotherium lacks the supratrochlear foramen but instead has a deep oleocraneal fossa, and both foramina are present in Hegethotherium but absent in Interatherium (Sinclair 1908). The olecraneal fossa (Fossa olecrani) or supratrochlear foramen (if it is perforated) is part of the elbow articulation capsule. Animals with a supratrochlear foramen are able to hyperextend the elbow joint (Lamb 1890, De Wilde et al. 2004, Voisin 2011). The epicondylar foramen, passage for the medial nerve and the branchial artery, was obliterated during the original preparation, but has now been re-exposed during our work on the material. The presence of an epicondylar foramen is considered a primitive condition within Mammalia and is retained in most notoungulates (Landry 1958, Shockey et al. 2012. Allalmeia has a wide medial epicondyle that is almost the same size as the trochlea (including the capitulum); this resembles the condition in Colbertia magellanica, Notostylops murinus Ameghino, 1897 and in interatheriids (Bergqvist et al. 2007, Lorente et al. 2013. The capitulum (Capitulum humeri) is round and continuous with the trochlea. The lateral trochlear crest is visible only on the palmar side and is indistinguishable from the capitulum on the dorsal side. As in Anisotemnus distentus (Ameghino, 1901), the medial crest is high, pronounced and marks the distal end of the trochlea (Shockey & Flynn 2007: AMNH 28906). Conversely, the medial crest is not continuous with the deltoid crest (Crista humeris). The lateral epicondyle is so poorly developed that the capitulum marks the lateral border in dorsal view. The relation between the trochlea and the medial epicondyle is similar to those present in N. murinus (e.g., MLP 82-V-1-8). In contrast, Hegetotheriids (e.g., Pachyrukhos, MLP 99-X-2-1; Paedoetherium, MLP 50-X-5-29), leontiniids (e.g., Scarritia, AMNH 29585) and toxodontids (e.g., Toxodon, MLP 96-III-10-103; Nesodon, MLP 55-XII-13-46), have a more reduced medial epicondyle. The presence of a lateral supracondylar ridge (Crista supracondylaris lateralis) could be tenta-tively inferred from the sharp and angular edge on the proximal side of the right humerus, but this side of the bone is broken. A small part of the proximal ulna is attached to the trochlea ventrally.
Femur. Only the proximal part of the right femur has been relocated (Fig. 2). The femoral head (Caput ossis femoris) is broken, but the greater trochanter (Trochanter major) and most of the proximal diaphysis are preserved. Rusconi (1946c) described a lower position of the femoral head than the greater trochanter, but in these remains it is not possible to determine if it was lower or higher, because the head is broken from the area of the lesser trochanter (Trochanter minor). The lesser trochanter seems poorly developed and there was a third trochanter (Trochanter tertius), but only its broken base is evident. The fossa trochanterica, insertion for the Obturatorius externus, Obturatorius internus and the Gemelli muscles, is well marked, deep and oval in outline. The diaphyseal lateral border forms a sharp ridge; judging from its position, it is a probable equivalent to a primate Linea aspera (White et al. 2011) or to a marked Labium laterale of Facies aspera of extant ungulates (Sisson & Grossman 1982). In both cases, this sharp ridge is the insertion area for several muscles, especially for the Glutaeus superficialis, that abducts the leg (Sisson & Grossman 1982).
The femur of Allalmeia closely resembles those of Colbertia lumbrerense (Berqgvist et al. 2007) and interatheriids in its distinguishable trochanters, which are more prominent than in hegethotherids or toxodonts but comparatively less than in Notostylops. The femur of Allalmeia is longer than the tibia (Rusconi 1946c) like the isotemnid Pleurostylodon (e.g., AMNH 29604) but in contrast to interatheriids and hegetotheriids.
Fibula. Only the distal epiphysis of the left fibula is preserved. It is well developed and covers part of the lateral crest of the astragalus. The facet for the calcaneum is as large as that for the talus. Situated between these structures is a fossa for the talofibular ligament as occurs elsewhere in Protypotherium (MLP 90-XII-26-10). The lateral malleolus (Malleolus lateralis) forms a distinct tuberosity with a sulcus for the peroneal tendons. The fibula is not fused with the tibia (Rusconi 1946c) as it is in interatheriids, isotemnids, mesotheriids ) and toxodontids; the tibiofibula of hegetotheriids is fully fused (Croft et al. 2004, fig. 14).
Calcaneum. The distal half of the right calcaneum is preserved in articulation with the rest of the pes. It has a distal peroneal process with a marked rugosity at its lateral end, and a cuboid facet that is oblique to the long axis of the bone (most clearly observable in plantar view). The calcaneum also has a well-developed plantar tubercle and a large sustentaculum (Sustentaculum tali) of the same width as the calcaneal tuber (Tuber calcanei). The peroneal process, sustentacular groove and plantar tubercle of Allalmeia are similar to those of Protypotherium (e.g., MLP 90-XII-26-10) in morphology and disposition, although the peroneal process is more expanded laterally, and the plantar tubercle is slightly more distally positioned. This is similar to the associated calcanei of Paulacoutoia protocenica (Paula Couto, 1952; e.g., AMNH 55390), Lamegoia conodonta Paula Couto, 1952(e.g., MCN-PV 1271, C. magellanica (e.g., DNPM LE446B), and in particular, those those referred to the kollpaniines Tiuclaenus minutus Muizon & Marshall, 1987 (e.g., MHNC 8322) and Molinodus (e.g., MHNC 8318) from the early Paleocene of Tiupampa, Bolivia (Cifelli 1983, Bergqvist 1996, Muizon et al. 1998).
Astragalus. The trochlea (Trochlea tali) of the right astragalus is preserved. It has asymmetrical crests that diverge proximodistally (Fig. 3). The lateral trochlear crest is higher and larger than the medial. The astragalar foramen is small. The flexor groove is well developed; the medial facet is small and crescentic, and has a central projection; there is a broken medial plantar tuberosity, and the lateral crest was sharp with laterally projecting plantar border. The base of the astragalar neck projects medially (Collum tali), and overall the astragalus appears to have been broad. The astragalus of Allalmeia with its shallow crests and crescent medial facet with a central projection is reminiscent of that in C. magellanica (e.g., AMNH 55373), Notostylops? (e.g., MLP 56-XII-18-92 to 201), Hegetotherium (e.g., MACN-A 9866:80), Thomashuxleya? (e.g., MLP 75-II-1-10), as well as the kollpaniine Tiuclaenus (e.g., MHCN 8308), the archaic North American ungulate Tetraclaenodon puercensis (e.g., AMNH 3878), and also more basal eutherians such as Protungulatum and Procerberus (see Szalay & Decker 1974). Leontiniidae, Notohippidae and Toxodontidae are characterized by elongate astragali with a reduced neck and flexor groove; the medial malleolar facet also extends onto the neck. Later interatheriids and hegetotheriids, excluding Hegethotherium, have an astragalus with a deep trochlea. Mesotheriids, such as Trachytherus (e.g., MLP 61-IV-11-6), have a similar medial facet to Allalmeia, but with a markedly asymmetrical trochlea.
Navicular. The naviculars have a 'W-shaped' outline in plantar view, with a lateral projecting rim (Figs 4,5). The medial tuberosity (which provides insertion for the Tibialis posterior) is well developed and extends proximally. Both naviculars have a narrow contact with the calcaneum on the dorsal side, as in Protypotherium (e.g., MLP 90-XII-26-10) termed a 'reverse alternating tarsus'. Cifelli (1993) considered this feature a synapomorphy of Leontiniidae, Notohippidae and Toxodontidae, but Shockey et al. (2012, p. 40) alternatively regarded it as a non-exclusive trait of 'advanced toxodontia' (being elsewhere in some Typotheria interatheriids such as Protypotherium and Federicoanaya). A reverse alternating tarsus is also present in some litopterns, where the calcaneocuboid articulation is placed more distally than that of the astragalonavicular. We identified that the medial tuberosity is well developed as in typotherians, basal toxodontians and the notoungulates of São José de Itaboraí.
Cuneiforms. The three cuneiforms are preserved. The ectocuneiform is cubic and the largest of the three, with a sesamoid in the plantar side of the left foot. The mesocuneiform is the smallest. The endocuneiform is longest proximodistally. Protypotherium, Hegetotherium and Interatherium and Tetraclaenodon have a similar cuneiform disposition and morphology (Matthew 1897, fig. 12;Sinclair 1908, pls II, V and VIII). This is unlike toxodontians and Phenacodus, which have a shorter endocuneiform.
Metatarsus. Allalmeia was pentadactyl, and although the first and fifth metatarsals were reduced, they were sufficiently developed to articulate with complete distal digits. The metatarsal of the hallux is less than half the Fig. 3. Allalmeia atalaensis right astragalus. Grey surfaces indicate attached bone remains. A, dorsal; B, lateral; C, medial; D, plantar; E, proximal; af, astragalar foramen; at, astragalar trochlea; lc, lateral astragalar trochlear crest; lf, lateral or fibular facet; fg, flexor groove; mc, medial astragalar trochlear crest; mf, medial malleolar facet. Scale bar = 10 mm. Fig. 4. Allalmeia atalaensis right foot. Metatarsals identified by roman numbers. A, plantar; B, lateral; C, dorsal; ast, astragalus; cal, calcaneum; cub, cuboid; ecc, ectocuneiform; edc, endocuneiform; ft, tuberosity for Flexor brevis minimi digiti; iph, intermediate phalanges; mc, mesocuneiform; nav, navicular; nmt, navicular medial tuberosity; plt, tuberosity for Peroneuss longus; pp, peroneal process; pph, proximal phalanges; pt, plantar tuberosity; sus, sustentaculum; uph, ungual phalanx. Scale bar = 10 mm. length of the second metatarsal. The second metatarsal is positioned more proximally than the others; it articulates not only with the mesocuneiform but with the ectocuneiform and endocuneiform. The third and fourth metatarsals are parallel to each other; the fourth metatarsal is only slightly larger than the third. The fifth metatarsal is less reduced than the first, and it has a well-developed tuberosity for origin of the Flexor brevis minimi digiti (a flexor of the fifth digit). The metatarsals are wider in the distal epiphysis, with diameter constriction between the proximal epiphysis and the half of the diaphysis; they are also compressed dorsoplantarly. We observed that the metatarsals are thin and long, as in typotherians in contrast to toxodonts. In Protypotherium and Interatherium, as in Allalmeia, the third and fourth metatarsals are the longest, although in Allalmeia, the fifth is more reduced than the second, whereas the opposite is the case in Protypotherium (Sinclair 1908).
Phalanges. Six proximal phalanges are preserved. Only one proximal phalanx in the left foot is complete and, judging from its in situ position and comparison with phalanges from the right foot, was probably a proximal phalanx of the second digit. In the right foot, the first and fourth digits have complete proximal phalanges; those of the second, third and fifth digits are represented only by their proximal ends. The proximal phalanges are less than half the size of the metatarsals, except for the first metatarsal, which is greatly reduced but the first digit phalanges are not reduced, so the proximal first phalanx is almost the size of the first metatarsal. The distal articulation in the proximal phalanges is plantarly oriented.
There are two intermediate phalanges, the fourth of the right foot and probably the first of the left foot; there is also an impression of the third right intermediate phalanx in the matrix. The fourth intermediate phalanx of the right foot is like a triangular prism, with the distal end more constricted and smaller than the proximal one. This articulates with the plantar side of the proximal phalanx such that the intermediate phalangeal row is positioned oblique in relation to the proximal one (Fig. 4). The distal articulation of intermediate phalanges is also plantar, making the entire ungual row oblique.
There are two ungual phalanges on the right foot: an isolated proximal portion of an ungual phalanx and a fourth digit ungual phalanx. They are claw-like; the fourth is flat dorsoplantary and convex but it is broken on the lateral border. The isolated ungual is eroded so the outline is difficult to determine. The bone is larger dorsoplantarly as expected for a claw-like phalanx, the Tuberculum flexorium is flat and barely marked on the ventral side of the phalanx.
The preserved unguals are claw-like but different to those of Homalodotherium (e.g., MACN-A 11721), which are bifurcated and claw-like but as those present in the cimolestan Stylinodon (Coombs 1983) or the creodont Oxyaena (Rose & Emry, 1993).
Lost postcranial elements. Rusconi (1946c) described five cervical, some sacral and caudal vertebrae that he considered similar to those of Paedotherium and Pachyrukhos in their general aspect. The ulna was apparently robust, with an arched diaphysis and triangular cross-section. The left femur was complete, and the lesser and third trochanter were enlarged into 'wings' with the third trochanter extending below the mid-diaphysis; the femoral trochlea was thin, long and well excavated, articulating with a thin patella. The tibiae were reportedly with robust proximal tibial epiphysis, shallowly excavated articulations for the astragalus and a robust medial malleolus. The almost complete right fibula was completely separated from the tibia. Rusconi (1946a) described the left astragalus as having a longer neck than those of Pachyrukhos or Paedotherium; the plantar aspect was not observed because it was articulated with the calcaneum. The left calcaneum was not discussed in detail by Rusconi (1946a), and the scapula, radii and manual bones were all fragmentary. Rusconi (1946a) depicted the pelvis with a large obturator foramen; there were one or two sacral vertebrae as in N. murinus (Lorente et al., 2013). The cleaning of several portions of the hindlimb, which were not originally prepared, allowed us to reinterpret characters of several bones and provide new anatomical information about this taxon. Rusconi's (1946a) figures of the pes are inverted with left representing right (Rusconi 1946c, figs 20, 21). In contrast to the original observations of Rusconi (1946c), there is an aperture that we identified as the supratrochlear foramen in the right humerus; based on its smooth border it does not seem to be a taphonomic artefact.

Morphological comparisons and lifestyle
Among notoungulates, phylogenetic analyses and palaeobiological inferences have been based mostly on dental and cranial features. Knowledge of the postcranial characters in this group is important for testing these inferences and to add new information about the evolutionary trends in the locomotor system and lifestyle.
Allalmeia was first considered by Rusconi (1946a, b, c) as a member of Notostylopidae. In Notostylops, from the middle Eocene of Patagonia, the humerus (i.e., MLP 82-V-1-8) is similar to Allalmeia in its wide proximal portion and the poorly developed lateral epicondyle. However, Notostylops is more robust with a proportionately smaller and less spherical capitulum, a less prominent medial crest, the supratrochlear foramen is absent and a smooth oleocraneal fossa is present instead, and the medial epicondyle is slightly narrower and more proximally oriented. The femur of Notostylops has a large but thinner and shallower trochanteric fossa, and more robust trochanters.
As observed by Simpson et al. (1962), the postcranial character states of Allalmeia are mostly non-diagnostic, but there are some traits that differentiate Allalmeia from other notoungulates: the low crural index and the shape of the Peroneus longus groove on the cuboid (although the cuboid morphology is unknown in most other taxa). In general, Allalmeia is morphologically comparable with the oldfieldthomasiid Colbertia magellanica (Bergqvist et al. 2007), but differs in the medial epicondyle of the humerus being larger, the epicondylar foramen being smaller and more proximally positioned, and the femur possessing a deeper trochanteric fossa. The astragalus of C. magellanica has more parallel trochlear crests, a more distally situated medial facet, a more medial and larger medial plantar tuberosity, and a shallower flexor groove. Both, C. magellanica and Allalmeia share a well-developed navicular medial tuberosity that articulates with the medial extension of the navicular facet on the astragalus. In contrast, the spring or collateral ligaments contact the astragalus in North American archaic ungulates (e.g., Arctocyon). This well-developed navicular medial tuberosity is present in all typotheres and basal toxodontians (Shockey & Flynn 2007), but is absent in the more recent toxodontids such as Nesodon (e.g., MLP 55-XII-13-46) and Toxodon (e.g., MLP 12-1125), as well as Deseadan leontiniids, such as Scarritia (e.g., AMNH 29585). These forms also lack a medial extension on the astragalus. Litopterns and astrapotheres lack a medial navicular tuberosity, and its condition is unknown in xenungulates and pyrotheres. Litopterns lack a medial tuberosity and have a plantar, instead of a medial, extension of the navicular facet in the astragalus. Astrapotherians have a medial extension of the navicular facet in the astragalus, but an almost flat navicular. If the proximally expanded navicular medial tuberosity can be confirmed in more basal notoungulates, it could be a derived character for Notoungulata. Astragali of both Allalmeia and Colbertia are similar to those attributed to the early Paleocene kollpaniines Tiuclaenus and Pucanodus (morph 1 and 2), and the calcanei of Tiuclaenus and Molinodus (morph 1 and 3, Muizon et al. 1998): the lateral process, astragalar medial facet and medial plantar tuberosity of Tiuclaenus are more similar to those of Allalmeia than to C. magellanica, but the trochlea and flexor groove bear greater resemblance to those of Colbertia. Tiuclaenus also possesses a crest in the astragalar neck, called 'tibial stop', which is similar although not identical to that of Colbertia. We observed that in Colbertia and other ungulates, this crest appears to be related to the trochlea, whereas in Tiuclaenus it develops from the head, as in the protoliptern Miguelsoria parayirunhor (MCT PV 1353), which otherwise possesses a 'litoptern-like' astragalus. At least for notoungulates, this crest has been considered a possible derived character (Shockey & Flynn 2007). The calcaneum of Tiuclaenus has a well-marked plantar tuberosity and a distal peroneal process like those of Allalmeia. The ectal facet is also oval in outline, and the fibular facet is half the size of the ectal; there is also an inverted lacriform sustentacular facet as expressed in the notoungulate calcanei from the Eocene of Paso del Sapo, Patagonia (Lorente et al. 2012). In the Pucanodus astragalus, the proximal half of the trochlea is broken, but the distal part shows a morphology compatible with that of Allalmeia. The plantar side is more similar, and head and neck more different to notoungulates than Tiuclaenus. The Molinodus calcaneum resembles that of Tiuclaenus Table 2 Measurements and inferred body mass of Allalmeia atalaensis.  (Muizon et al. 1998), but has a marked rugosity in the lateral end of the peroneal process as in Allalmeia. The lateral side of the calcaneum of Molinodus has equivalent structures to the notoungulate Notostylops murinus (Lorente et al. 2013), but the fibular facet is smaller. The sustentacular facet is also more rounded than that of notoungulates. Several palaeobiological features, such as Allalmeia body mass and their locomotor posture can be inferred from the postcranium. Rusconi (1946c) considered Allalmeia to be a digitigrade runner that could have employed jumping at high speeds, and suggested close comparisons in lifestyle with Pachyrukhos moyani and Paedotherium imperforatum. Structurally, however, Allalmeia differs from these taxa (e.g., Pachyrukhos, MLP 99-X-2-1; Paedoetherium, MLP 50-X-5-29) in its wider humeri, unfused zeugopods, more prominent trochanters, the higher position of the greater trochanter (positioned below the head in Paedotherium and Pachyrukhos), and the wider and shallower astragalar trochlea. Allalmeia also had a shorter tibia relative to its femur (Rusconi 1946c), a feature thought to be characteristic of fossorial, terrestrial, arboreal and some cursorial extant mammals, as opposed to ricochetal (Samuels & Van Valkenburgh 2008). Furthermore, the epicondyle is broadened, suggesting substantial Pronator teres and Flexor digitorum musculature as in fossorial, terrestrial or arboreal taxa (Salton & Sargis 2008). Rusconi (1946c) calculated the pes length index (III metatarsal length/femur length) and crural index (tibia length/ femur length) for Allalmeia as 0.33 and 0.809 respectively, which are additionally compatible with arboreal, fossorial and especially semifossorial habits (Samuels & Van Valkenburgh, 2008), such as the American badger Taxidea taxus; some cursorial/terrestrial mammals such as the South American tapir (Tapirus terrestris, MLP 1) and the guinea baboon, Papio papio (e.g., MACN 23,383); and some arboreal marsupials (Jungers 1980) such as the koala Phascolarctos cinereus. Nevertheless, the long, thin metatarsal and the low astragalar trochlea would argue against a fossorial or cursorial lifestyle (Van Valkenburgh 1987, Carrano 1997, and instead we propose that Allalmeia was probably a terrestrial (noncursorial) or arboreal animal.
The foot posture of Allalmeia can be inferred from its metatarsal length-width ratio (sensu Carrano 1997), which falls between those of plantigrades and digitigrades, but lower than subunguligrades. Oblique positioning of the intermediate versus proximal phalangeal rows implies a digitigrade posture, as extant animals with this disposition are digitigrades.

Body mass estimations
Several measures were considered in order to infer body mass. Considering a normal distribution of the data (Table 2), the confidence interval for the mean at 99% was calculated between 1.140 kg and 6.336 kg. The Allometric equation from Janis (1990), Scarano et al. (2011), Scott (1990) and Tsubamoto (in press (2012), based on allometric equations for dental and postcranial remains, estimated the body mass for several oldfieldthomasiids, and derived a mean of 5.567 kg for Oldfieldthomasia, 5.73 kg for Maxschlosseria, 3.98 kg for Colbertia and 5.98 kg for Ultrapitecus. These corporal masses agree approximately with the expected values if a comparison with a living analogues of the same size is undertaken. We compared several astragalar measures of Allalmeia with those of a sample of extant mammals with known body mass (Table 3). We also applied to this sample the procedure of Tsubamoto (in press) in order to estimate their body mass from astragalar dimensions. The greatest difference between weighted and estimated body mass was 0.816 kg for Conepatus chinga, with the least (0.035 kg) being for Dasyprocta azarae. Allalmeia is closer to Lepus callotis in astragalar measures with a weighted body mass of 2.7 kg, but if the inferred body mass of these extant mammals is considered, it is closer to the Kinkajou Potos flavus at 2.970 kg.

Palaeoenvironmental implications
Sedimentological analysis by Kokogian et al. (1988) concluded that the Divisadero Largo Formation was deposited on an extensive alluvial plain with semi-permanent water bodies associated with ephemeral rivers in interdune areas. However, this reconstruction seems to be inconsistent with the faunal assemblage recovered from the unit. Certainly, microscopic, palaeosol and facies assessments by López (2008), indicated at least five distinct lithofacies reflecting different and contrasting depositional systems. A predominance of ephemeral and episodically high-energy river systems confined to channels was distinguished within the lower and middle stratigraphic levels. In contrast, shallow lacustrine facies incorporating epilimnic open environments predominated up-sequence. The presence of palaeosols indicates subaerial exposure. Climatic conditions are reconstructed as warm and humid, with abundant availability of water, which probably promoted the development of trees and abundant vegetation cover (López 2008). This complies with the abundant small mammals, such as Allalmeia, with low-crowned teeth and a non-specialized or even an arboreal morphology; and is different to the specialized high-crowned notoungulates of Mariño Formation that correspond to a time of drier conditions correlated with the initial uplift of the Andes range (López 2008).

Summary
Allalmeia was a small, digitigrade generalist notoungulate that could have been either terrestrial or arboreal, but with a capacity to hyperextend the elbow as in modern canids and suids. Like its relative Colbertia, the generally gracile postcranial skeleton of Allalmeia is more closely compatible with archaic ungulates and Typotheria (e.g., Protypotherium) rather than notostylopids such as Notostylops, or the basal toxondontian isotemnids Thomashuxleya, Anisotemnus and Pleurostylodon. This resemblance might have been both morphofunctional and phylogenetic. The latitudinal provenance (32°53′S) of Allalmeia, suggests that it could have been a possible link between Patagonian notoungulates and those recorded from the northwest of Argentina. Linear dimensions in mm and body mass in kg. Extant animals are ranked from the most similar to the most different based on linear values. Abbreviations: dpdt, mid disto proximal diameter of the astragalar throclea; lltc, length of the lateral trochlear crest of the astragali; lmtc, length of the medial astragalar trochlear crest; tdn, transversal diameter of the neck of the astragali; tdt, transversal diameter of the astragalar trochlea. Mean of weighted body mass from 1 Feldhamer et al. (2003), 2 Janis (1984), 3 Walter & Cant (1977), 4 Simmen et al. (2010), 5 McWilliams (2009) and 6 Kasper et al. (2012). Inferred body mass using allometric equations of: a Janis (1990), Scarano et al. (2011) and Scott (1990); b Tsubamoto (in press).