A new set of ancestors for hippos

Figure 4. Hippopotamus. This stout, wide-faced, fanged mammal does not nest with deer.

The following clears up
another fine mess traditional paleontologists have provided based on taxon exclusion and professional bias. Today, let’s talk about the ancestry of hippos (Fig. 1).

Figure 1. Living hippopotamus. Now I ask you, does this look like a relative to deer and giraffes? Or to mesonychids?

Wikipedia, representing traditonal paleontology, reports
here, “The earliest known hippopotamus fossils, belonging to the genus Kenyapotamus in Africa, date to around 16 million years ago. Hippopotamidae are classified along with other even-toed ungulates in the order Artiodactyla. Other artiodactyls include camels, cattle, deer and pigs, although hippopotamuses are not closely related to these groups.” 

Do you sense the underlying problem here?
They say, hippos are in the Artiodactyla, but not close. This is what is known as a ‘red flag’.

But wait, there’s more! “The most recent theory of the origins of Hippopotamidae suggests that hippos and whales shared a common semiaquatic ancestor that branched off from other artiodactyls around 60 million years ago.” 

This is all getting to be completely bogus (based on LRT results)
And very disturbing. Tenrecs are the sisters to whales, as we learned earlier here. But wait, there’s more. “A rough evolutionary lineage can be traced from Eocene and Oligocene species: Anthracotherium and Elomeryx to the Miocene species Merycopotamus and Libycosaurus and the very latest anthracotheres in the Pliocene. Look those taxa up. They’re all skinny, long-legged terrestrial grazers. Not hippo-like at all ~

As an example, here’s Anthracotherium….
a big artiodactyl and an anthracothere (Fig. 2). Does it have huge fangs on a wide gape like a hippo? No. It’s just a big old prehistoric deer. Superficially, it does look like a giant tenrec with that long rostrum, but it does not nest with Andrewsarchus, a real giant tenrec.

Figure 2. Anthracotherium and the anthracotheres nest with their long narrow tiny tooth snouts with Ancodus, deer and other gracile artiodactyls, not hippos. Whale ancestors do have a long rostrum, but they are hoof-less like tenrecs, not hoofed like ungulates.

Figure 3. From the Paleocene, Ocepeia is the current sister to Hippopotamus in the LRT. Already the wide rostrum and high orbit are apparent.

Here’s how the LRT sees it
Apparently hippos have an ancestry that extends at least into the Paleocene.

Ocepeia daouiensis (Gheerbrant et al 2001, 2014; Paleocene, 60 mya; 9 cm skull length) was considered the oldest known of the ‘Afrotherians’ known from skulls, but the ‘Afrotheria’ is an invalid clade. Gheerbrant et al. 2016 nested Ocepeia with aardvarks and Potamogale, an extant aquatic tenrec. These were all derived from a sister to Arctocyon nesting within the Arctocyonidae. The LRT nested Arctocyon as an omnivorous marsupial.

Here (Fig. 3) Ocepeia is a Paleocene hippo sister, derived from a sister to Mesonyx (Fig. 5; Cope 1872) and Harpagolestes (Fig. 4; Matthew 1909). With Ocepeia in the Paleocene, just imagine how far back those mesonychids must go back into the Cretaceous.  On Ocepeia note the hippo-like relatively longer rostrum, elevated orbits and strong retro process on the dentary. The jaw joint is aligned with the maxillary tooth row. Those little canine tusks on Ocepeia are reduced from a larger sister via neotony according to the LRT.

The pneumatized skull of Ocepeia
contains many air spaces, even though it is relatively small. Slightly larger skulls have larger canines and so are considered male.

Wikipedia reports, “Mammals are extremely rare in the Ouled Abdoun in contrast to the associated marine vertebrate fauna which includes sea birds, sharks, bony fish, and marine reptiles (including crocodilians, sea turtles, and the sea snake Palaeophis). Terrestrial species were probably transported off shore into the Moroccan sea before fossilization.” Apparently ignored by traditional paleontologists as possible candidates in this formation, hippos and their ancestors, like Ocepeia, could have been aquatic even at that early stage.

Figure 4 Robust Harpagolestes nests between the hippos and Mesonyx. Now, doesn’t this look more like a hippo? The LRT agrees with you.

Long known as a robust mesonychid
Harpagolestes makes a good interim taxon between Mesonyx and hippos. Look at hose tusks! Yet it had not yet developed the low, wide skull and elevated orbits that characterize surface dwelling hippos, like Hippopotamus and Ocepeia.

Figure 5. Mesonyx, the first known mesonychid was a sister to Hippopotamus in the large reptile tree. So maybe it was a plant eater. Now with the addition of Ocepeia and Harpagolestes, it moves a little further away.

We should someday find
mesonychids, like Mesonyx, in Cretaceous strata based on its phylogenetic nesting and the Paleocene placement of Ocepeia.

Figure 6. Subset of the LRT, focusing on herbivorous mammals, including Hippopotamus and its ancestors. This cladogram needs a bit of updating based on recent additions to the LRT. Note the wrong placement of Astrapotherium, which we looked at yesterday with the inclusion of a better sister taxon. 

References
Cope ED 1872. Descriptions of some new Vertebrata from the Bridger Group of the Eocene. Proceedings of the American Philosophical Society 12:460-465.
Gheerbrant E, Sudre J,Iarochene M, Moumni A 2001. First ascertained African “Condylarth” mammals (primitive ungulates: cf. Bulbulodentata and cf. Phenacodonta) from the earliest Ypresian of the Ouled Abdoun Basin, Morocco. Journal of Vertebrate Paleontology. 21(1):107–118.
Gheerbrant E, Amaghzaz M, Bouya B, Goussard F and Letenneur C 2014. Ocepeia (Middle Paleocene of Morocco): The Oldest Skull of an Afrotherian Mammal”. PLoS ONE. 9 (2): e89739.
Gheerbrant E, Filippo A and Schmitt A 2016. Convergence of Afrotherian and Laurasiatherian Ungulate-Like Mammals: First Morphological Evidence from the Paleocene of Morocco”. PLOS ONE. 11 (7): e0157556.
Matthew WD 1909. The Carnivora and Insectivora of the Bridger Basin, middle Eocene. Memoirs of the American Museum of Natural History 9:289-567.
wiki/Harpagolestes
wiki/Ocepeia
wiki/Hippopotamus
wiki/Anthracotherium

A better sister for Astrapotherium: Meniscotherium

This starts with a confession:
Earlier I, like everyone else before today, interpreted the tusks of Astrapotherium (Fig. 1; Burmeister 1879; Hatcher 1902) as canines. They’re not canines. They’re incisor tusks, like those of elephants. The original drawing (complete skeleton in Fig. 1) including some dotted lines indicating missing rostral material. That was conjectural and mistaken. The mandible of Astrapotherium really does stick out quite a bit further than the rostrum. It’s important to remember that not all drawings have to be interpreted the way they look. Mistakes can be made by anyone anywhere.

Figure 1. New interpretation of Astrapotherium skull with premaxilla and large incisor tusks replacing old canine tusks. The canines are absent. That manual digit 1 that Hatcher 1902 did not like in his drawing (Fig. 1 toned beige) is actually a good fit and works in phylogenetic bracketing. 

And that’s not the only error
Like so many others before me, I also fell prey to errors that arise via taxon exclusion. I didn’t have one really good enough sister taxon for Astrapotherium. Now I do.

Meet Meniscotherium
(Figs. 2, 3; Cope 1874; Williamson and Lucas 1992; Middle Eocene 54-38 mya; 25-50 cm long), which Wikipedia describes as a dog-sized herbivore with hooves found as a pack of individuals.

Cooper et al. 2014 nested Meniscotherium with Phenacodus as a condylarth, a possible member of Afrotheria, perissodactyl. They did not test Astrapotherium.

Wible et al. 2007 nested Meniscotherium close to early cetioartiodactyls (an invalid clade) and close to early Carnivora. They, likewise, did not test Astrapotherium.

The LRT nests the clade of Astrapotherium + Meniscotherium between the clade of Edentates and the clade of Proboscideans + Sirenians. I will add those and several other taxa to the LRT when I have a little more time.

Figure 2. Meniscotherium skull. Note the genesis of premaxillary tusks here along with the diminishing canine (orange). This is a smaller predecessor to Astrapotherium.

The retention of five fingers and five toes
is key to the phylogenetic nesting of these taxa. More derived taxa start losing digit 1. We can see the genesis of incisor tusks in this genus.

Figure 3. Meniscortherium skeleton. The fingers and toes are not known. This reconstruction differs from the original in that the pelvis is rotated more vertically. Some specimens were 25 cm long. Others were 50 cm long estimated.

Meniscotherium is the smaller and more plesiomorphic
of the two and is found in earlier strata (Eocene, 50-38 mya) than Astrapotherium (late Oligocene, Middle Miocene, 28-15 mya).

Figure 4. Astrapotherium to scale with two specimens of Meniscotherium.

References
Burmeister 1879. Description physique de al République Agentine, T. III 1879:517.
Cooper LN, Seiffert ER, Clementz M, Madar SI, Bajpai S, Hussain ST, Thewissen JGM 2014-10-08. Anthracobunids from the Middle Eocene of India and Pakistan Are Stem Perissodactyls. PLoS ONE. 9 (10): e109232. doi:10.1371/journal.pone.0109232. PMID 25295875.
Hatcher JB 1901. Report of the Princeton University Expeditions to Patagonia 1869-1899. Mammalia of the Santa Cruz Beds. IV. Astrapotheria. Scott WB ed. Vol. 6, Paleontology 3. Princeton, NJ Stuttgart 1909-1928.
Wible JR, Rougier GW, Novacek MJ, Asher RJ 2007. Cretaceous eutherians and Laurasian origin for placental mammals near the K/T boundary. Nature 447: 1003–1006. doi: 10.1038/nature05854
Williamson TE, Lucas SG 1992. Meniscotherium (Mammalia, “Condylarthra”) from the Paleocene-Eocene of western North America. Bulletin of the New Mexico Museum of Natural History and Science 1: 1–54.

Former ‘notoungulate’ Periphrangis is really a wombat

Periphrangis harmeri (Roth 1899; Fig. 1; Oligocene, 48-28 mya) has long been considered a notoungulate. Earlier the LRT nested two former notoungulates as wombats. Periphranigis also shares several wombat traits, including a jugal that contacts the jaw glenoid, procumbent incisors and a septomaxilla.

Figure 1. Periphrangis was considered a notoungulate, but it is clearly a wombat with four molars and a jugal that contacts the jaw glenoid, among several other identifying traits.

When we first looked at Haramiyavia
(Jenkins et al. 1997, Luo et al. 2005) here, this small Late Triassic mammal was considered a basal multituberculate. Now that several wombats have been added to the LRT Haramiyava could be another wombat. Wombats share procumbent incisors and a convex ventral mandible. Hard to tell with present data. In either case, both wombats and multituberculates are rather derived taxa for the Late Triassic.

Figure 2. Haramiyavia reconstructed and restored. Missing parts are ghosted. Three slightly different originals are used for the base here.

Arctocyon
(Fig. 3; Blainville 1841, Gould and Rose 2014; YPM VP 021233; 60 mya) was long and widely considered (see Wikipedia page) a primitive plantigrade ungulate condylarth procreodi placental.  In the LRT Arctocyon nests with basal carnivorous/omnivorous marsupials. Essentially it is a giant opossum, like Didelphis, but with a few derived traits, more like Thylacinus, a taxon that reduces the epipubes and molar count, hence the earlier traditional confusion. Just look at these taxa side-by-side. It’s obvious, but it’s also in the matrix scores.

Figure 3. Arctocyon mumak is no longer an ungulate placental, but a carnivorous marsupial, close to Thylacinus.

Small brains and long jugals extending to the jaw glenoid
also give them away as metatherians. Not sure why even recent authors (Gould and Rose 2014) are not seeing this. They must be counting molars.

References
Blainville HM 1841. Osteographie et description iconographique des Mammiferes récentes et fossiles (Carnivores) 1, 2 Paris.
Gould FDH and Rose KD 2014. Gnathic and postcranial skeleton of the largest known arctocyonid ‘condylarth’ Arctocyon mumak (Mammalia, Procreodi) and ecomorphological diversity in Procreodi. Journal of Vertebrate Paleontology 34(5):1180-1202.
Jenkins FA, Jr, Gatesy SM, Shubin NH and Amaral WW 1997. Haramiyids and Triassic mammalian evolution. Nature 385(6618):715–718.
Luo Z-X, Gatesy SM, Jenkins FA, Jr, Amaralc WW and Shubin NH 2015. Mandibular and dental characteristics of Late Triassic mammaliaform Haramiyavia and their ramifications for basal mammal evolution. PNAS 112 (51) E7101–E7109.
Roth S 1899. Aviso preliminar sobre mamíferos mesozóicos encontrados en Patagonia [Preliminary notice on Mesozoic mammals found in Patagonia]. Revista del Museo de La Plata 9:381-388

 

What is Triopticus? (It’s not what they think it is…)

riIt’s been a long time
since an interesting ‘reptile’ showed up in the literature. Especially an enigma like this one.

A recent paper by Stocker et al. 2016
reports on a domed and expanded Late Triassic cranium that they identify as an archosaur, but it’s unlike that of any other archosaur. Triopticus primus was named for its three eyes, with a big one on top (Fig. 1). The authors compared the domed appearance of the cranium in Triopticus with a Cretaceous dome-headed ornithischian dinosaur, Stegoceras.  They also discussed convergence in general and provided a CT scan brain endocast of Triopticus.

Unfortunately 
the authors employed a prior cladogram (Fig. 2) by Nesbit et al. 2015, expanded from Pritchard et al. 2015. that was shown to not recover sister taxa that looked alike and did not provide a gradual accumulation of derived traits at several nodes. In their cladogram Triopticus nested without resolution among basal archosauriforms like Proterosuchus, which looks nothing like it. By contrast, the LRT was able nest and fully resolve Triopticus elsewhere.

Figure 1. In round 1 I added characters shown here to the LRT in two passes. One recovered a sisterhood with mesosaurs. The other nested with Tanytrachelos, among the tanystropheid tritosaur lepidosaurs. Both shown here for comparison. Triopticus would be 2x the size of the giant Tanyrachelos from New Mexico.

From the Stocker et al. abstract:  “Exemplifying this extreme morphological convergence, we present here a new dome-headed taxon from the assemblage, which further illustrates the extraordinary range of morphological disparity present early in the Late Triassic.” That ‘extraordinary range’ should be — and will be — chopped down substantially with the right sister taxa.

A few problems with the archosauriform hypothesis include:

1. No other archosauriforms, until you get to pachycelphalosaurs in the Cretaceous, expand the cranium deleting the upper temporal fenestra.
2. The entire rostrum and mandible is absent, so no naris, antorbital fenestra or teeth are known, even in part.
3. They dubiously identified an antorbital fenestra and fossa at the edge of the fossil.
4. And they were not aware that Tanytrachelos and kin, including pterosaurs within the – Lepidosauria -, also have an antorbital fenestra, but without a fossa.
5. A large pineal opening is present, but never present at such a size in archosauriforms.
6. Large and expanded supratemporals (overlooked originally) are present.
7. The extreme angle of the rostrum coupled with the large orbit are traits not found in basal archosauriforms that typically have a long boxy rostrum.

Figure 2. Stocker et al. 2106 cladogram nesting Triopticus uncertainly within a set of unresolved basal archosauriforms and far from the Tanystropheidae. The LRT completely resolves all nodes. Note how this cladogram mixes Lepidosauromorpha with Archosauromorpha and separates the protorosaurus, Protorosaurus and Prolacerta.

This is a perfect problem
for the large reptile tree (LRT) which now provides 820 (chart needs to be updated from 812) opportunities for Triopticus to nest in. With that large number of taxa, unfortunately I had to split the matrix in two, even for a simple Heuristic Search. By contrast, the Stocker et al. matrix included 30 taxa and 247 characters.

Stocker et al. report,
“We chose this dataset because the following combination of character states in Triopticus are also present in some archosauromorph taxa:

1. presence of a single occipital condyle;
2. ossified laterosphenoid;
3. presence of a metotic strut of the otoccipital;
4. presence of upper and lower temporal fenestrae;
5. presence of an antorbital fenestra and fossa formed by the lacrimal.”

They provided no reconstructions of included taxa.

First,
I tested Triopticus against basal tetrapods and the new Lepidosaurmorpha and found that Triopticus nested with the aquatic, long-necked tritosaur Tanytrachelos (Fig. 1), large specimens of which were recently found in New Mexico (Fig. 3). Like Triopticus the rostrum descends at a high angle from a tall cranium in Tanytrachelos, which also shares a large orbit and a large pineal foramen (at present known only from sister taxa). Like related fenestrasaurs and langobardisaurs, Tanytrachelos also had a small antorbital fenestra without a fossa, but that would have been beyond the rim of the broken skull in Triopticus (Fig. 5).

Figure 3. A large incomplete Tanytrachelos from New Mexico compared to the smaller more complete East Coast specimen. Triopticus would be twice as large as the New Mexico specimen.

Second,
I tested Triopticus with the rest of the matrix, the new Archosauromorpha, and found that Triopticus nested with the mesosaurs (Fig. 4), an aquatic enaliosaur clade close to thalattosaurs and ichthyosaurs, all derived from basal pachypleurosaurs. It did not nest with archosauriforms. While basal mesosaurs have typical diapsid temporal regions, Mesosaurus, like Triopticus, closes up the upper temporal fenestra, then the lateral temporal fenestra with bone expansion.  Mesosaurs also retain a relatively large pineal foramen and have large eyes, but they don’t have a sharply descending preorbital region.

Figure 4. Mesosaurus, like Triopticus, has a large pineal foramen and expands the skull bones to obliterate former temporal fenestrae.

Digital Graphic Segregation
was applied to the cranial lump that is Triopticus (Fig. 5) and the skull suture patterns, perhaps overlooked by those with firsthand access due to the expansion of the cranial bones, revealed a Tanytrachelos-like morphology (Fig. 2). I illustrate this interpretation here with the hope that this hypothesis can be either confirmed or falsified. This is a tough assignment.

Figure 5. Triopticus reconstructed along the bauplan of Tanytrachelos. Note the zone of weakness where the fossil split appears to be at the top of the narial/oral opening. The lateral temporal fenestra is the top half of a divided LTF. The UTF appears to be sealed over by expansion of the frontal over the postorbital. As in Tanytrachelos, the lacrimal has a dorsal process over the orbit separating the maxilla from the prefrontal. At 72 dpi this is 90 percent of actual size.

Tanystropheids
have been reported from the Hayden Quarry of northern New Mexico (Chinle Formation) far from the west central Texas location of Otis Chalk. Stocker et al. included Tanytrachelos in their study, even though they have not provided a reconstruction of it, so it is difficult to imagine how they interpreted it. Tanystropheids, in general, have widely varying skull shapes. Triopticus appears to have expanded the morphospace just a little, not a lot.

The loss of ventral material in the Triopticus fossil
appears to have occurred at the roof the narial/oral opening.

So what other long-necked animal
expands the cranium like Triopticus? Giraffa, the giraffe. Maybe it will turn out to be a better analogy than short-necked Stegoceras?

References
Nesbitt SJ, Flynn JJ, Pritchard AC, Parrish JM, Ranivoharimanana L and Wyss AR 2015. Postcranial osteology of Azendohsaurus madagaskarensis (?Middle to Upper Triassic, Isalo Group, Madagascar) and its systematic position among stem archosaur reptiles. Bulletin of the American Museum of Natural History 899, 1-125.
Pritchard AC, Turner AH, Nesbitt SJ, Irmis RB and Smith ND 2015. Late Triassic tanystropheid (Reptilia, Archosauromorpha) remains from northern New Mexico (Petrified Forest Member, Chinle Formation): insights into distribution, morphology, and paleoecology of Tanystropheidae. Journal of Vertebrate Paleontology, 10.1080/02724634.02722014.02911186.
Stocker MR, NesbittSJ, Criswell KE, Parker WG, Witmer LM, Rowe TB, Ridgely R  Brown MA 2016. A Dome-Headed Stem Archosaur Exemplifies Convergence among Dinosaurs and Their Distant Relatives. Current Biology (advance online publication)DOI: http://dx.doi.org/10.1016/j.cub.2016.07.066   pdf

Hyopsodus: an Eocene pre-dog, not an archaic ungulate

Hyopsodus lepidus (H. paulus type specimen, Leady 1870; AMNH 143783; Eocene; Fig. 1) is traditionally considered an odd-toed ungulate, despite having a five-clawed manus and a four-clawed pes. Wikipedia also promotes this nesting, but unlike most ungulates, they report, “It is believed to have been swift and nimble, living in burrows, and perhaps able to use echolocation,” and it is shown climbing on a tree trunk. Another website lists Hyopsodus as a condyarth.

Here, in the LRT, 
Hyopsodus nests with Canis and more closely with Miacis (Fig. 2), two members of the Carnivora in the large reptile tree (LRT). Shifting Hyopsodus over to the Condylarthra adds nearly 30 steps. Those tiny feet beneath that long and wide body do not look to me like they could be used to excavate burrows or climb trees. Plus that short tail and long torso are not typical of climbing animals.

Figure 1. Hyopsodus as originally reconstructed (below) and as reconstructed here above in two views. This former condylarth now nests with dogs. Hyopsodus has more teeth than Canis and Miacis, and the canine is small and there is no carnassial, similar to the outgroup taxon, Protictis.

From the Orliac et al. 2012 abstract:
“Hyopsodus presents one of the highest encephalization quotients of archaic ungulates and shows an “advanced version” of the basal ungulate brain pattern, with a mosaic of archaic characters such as large olfactory bulbs, weak ventral expansion of the neopallium, and absence of neopallium fissuration, as well as more specialized ones such as the relative reduction of the cerebellum compared to cerebrum or the enlargement of the inferior colliculus [hearing]. The detailed analysis of the overall morphology of the postcranial skeleton of Hyopsodus indicates a nimble, fast moving animal that likely lived in burrows.” Sounds like a member of the Carnivora…

Figure 2. Miacis, an Eocene ancestor to extant dogs, such as Canis. Note the transverse premaxilla and tiny premaxillary teeth as in Hyopsodus.

Miacis has long been known as a dog ancestor.
And here it also nests with Canis (Fig. 3). So, compare Miacis to Hyopsodus (Fig. 2) and you’ll find very few differences.

Figure 3. Canis lupus, the wolf that gave rise to extant dogs through selective breeding. Note the five phalanges on the manus and four on the pes.

Several specimens and species are known
of Hyopsodus, most from just teeth and jaws.

References
Orliac MJ, Argot C and Gilissen E 2012. Digital Cranial Endocast of Hyopsodus (Mammalia, “Condylarthra”): A Case of Paleogene Terrestrial Echolocation? PlosOne v.7(2); 2012PMC3277592

Thomashuxleya: another former notoungulate, joins the Condylarthra

Yet another
former notoungulate, dog-bear-like Thomashuxleya rostrata (Ameghino 1901; Eocene, 48-40 mya; 0.33m skull length, 1.3 m overall length; Fig. 1). Named after Thomas Huxley, Darwin’s ‘bulldog’ (his strong supporter), this peccary-mimic had a generalized 11 teeth (per quadrant), four toes (per foot, but look at the odd distribution! and meanwhile several descendants retain five toes so this is a red flag on the data). Since all the data for this taxon currently comes from this drawing, based on several specimens and still incomplete, the manus data might be a little untrustworthy… to say the least. The pedal data is conjectural from the start.

Figure 1. Thomashuxleya using the only available data. Here the dental formula is 3.1.4.3. Skeleton modified from the chimaera created by Simpson 1967. Based on this data is not a notoungulate nor a toxodont. That diverging thumb suggests a climbing ability, lost in all later taxa. Plantigrade pes and digitigrade manus.

Tred carefully here
|the reconstruction offered (Fig. 1) is based on an old line drawing and that is based on a chimaera of specimens. Back in the day, as we’ve seen with other museum mounts, that was common practice. Thomashuxleya has been referred to the Isotemnidae family of Notoungulata. Evidently other taxa in this family are known from isolated teeth or jaw fragments, but not enough to warrant their own Wikipedia page. Dr. Darin Croft includes Toxodon as a sister to this clade.

According to Croft 2016
Isotemnids have low-crowned teeth that gradually decrease in size from back to front without a diastema. That also sounds like Arsinoitherium, doesn’t it? So, it’s not a unique trait.

Several new basal condylarths
have been added to the LRT. We’ll look at them soon.

References
Ameghino F 1901. Notices préliminaires sur des ongulés nouveaux des terrains crétacés de Patagonie [Preliminary notes on new ungulates from the Cretaceous terrains of Patagonia]. Boletin de la Academia Nacional de Ciencias de Córdoba 16:349-429
Croft D 2016. Horned armadillos and rafting monkeys: the fascinating fossil mammals of South America. Indiana University Press 320 pp.
Simpson GG 1967. The beginning of the age of mammals in South America. Part II. Bulletin of the American Museum of Natural History 137, 1-260.
Simpson GG 1980. Splendid Isolation, the Curious History of South American Mammals. Yale University Press, New Haven, CT.

tetrapod-zoology/isotemnid-toxodonts-2012/
wiki/Thomashuxleya

Nesting twin-horned Arsinoitherium within the Condylarthra

Figure 1. Famous and enigmatic, Arsinoitherium has been known for over a century, and traditional paleontologists still do not know what it is.

FIgure 1. Subset of the large reptile tree, the Condylarthra, featuring Astrapotherium. Note the phylogenetic proximity of Astrapotherium and Tapirus.

Traditionally Arsinoitherium zitteli has been hard to classify.
Wikipedia reports, “Arsinoitherium (Beadnell 1902; Eocene-Oligocene, 36-30mya; 3 m in length; Fig. 1) is related to elephants, sirenians, hyraxes and the extinct desmostylians.” That’s a pretty broad gamut of taxa.

And they’re all wrong according to the large reptile tree (now 812 taxa, subset Fig. 2).

And this came as a surprise to me, too
among 811 other taxa, Arsinoitherium nests with Gobiatherium mirificum (Fig. 3; Osborn and Granger 1932; Middle Eocene), which Wikipedia considers, “one of the last uintatheres” of which Uintatherium is the titular and most famous member. Wikipedia goes on to report, “Gobiatherium lacked knob-like horns, or even fang-like tusks. Instead, it had enlarged cheekbones and an almost spherical snout. Because of the noticeable lack of many diagnostic uintathere features (the horns and tusks), the genus is placed within its own subfamily.” Here’s where tradition and the LRT agree… but let’s push this a little further to see where it takes us within the friendly confines of the current LRT taxon list.

Figure 3. Gobiatherium skull (A. M. 26624) in three views. Though not immediately apparent, Gobiatherium is closest to Arsinoitherium in the LRT. Image from Osborn and Granger 1932.

Among all tested placental taxa, and despite distinct overall appearances
only Arsinoitherium and Gobiatherium:

1. redevelop the ascending process of the premaxilla, completely enclosing the naris;
2. produce a wide, elevated set of nasals, further expanding into horns in Arsinoitherium;
3. only two molars, rare among placentals;
4. and no other condylarths have a wide flat cranium, usually a crest or a convex cranium is present.

That premaxillary ascending process
looks so normal. But among marsupial and placental mammals it is very rare indeed! Of course, the LRT does not depend on one or several traits, several dozen nest Arsinoitherium with Gobiatherium and their sisters.

Even without Gobiatherium
Arsinoitherium nests with Uintatherium. Coryphodon nests closer to Uintatherium. All descend from a sister to Thomashuxleya (Fig. 4), which we’ll look at soon in greater detail.

Figure 4. Thomashuxleya is basal to uintatheries and arsionoitheres. It is not a notoungulate, an invalid taxon.

We hold as an ideal
a gradual accumulation of derived traits in derived taxa, like Gobiatherium and Asinoitherium. In this clade, unfortunately we don’t have enough taxa to make that gradual accumulation of traits any more gradual than it currently is. This is the best we can do, at present, with available data and the present taxon list.

But it’s a good start!
And closer than anyone figured out before.

References
Beadnell HGC 1902. A preliminary note on Arsinoitherium zitteli, Beadnell, from the Upper Eocene strata of Egypt. Public Works Ministry, National Printing Department. Cairo: 1–4.
Lucas SG 2001. Gobiatherium (Mammalia: Dinocerata) from the Middle Eocene of Asia: Taxonomy and biochronological Significance. Paläontologische Zeitschrift 74 (4): 591–600.
Osborn HF and Granger W 1932. Coryphodonts and uintatheres from the Mongolian expedition of 1930. American Museum Novitates 552:1-16.

wiki/Arsinoitherium
wiki/Gobiatherium