Description

AMNH 21608 is well preserved except for the anterior portion of premaxilla and nasal, and the zygomatic arch (figs. 2–4). In lateral view (fig. 2), the dorsal profile is fairly flat, characteristic of a dolichocephalic cranium. The premaxilla is nearly completely damaged, and the alveoli of the upper canines represent the anterior tip of the preserved cranium. The rostral portion of the nasal is not preserved, and in anterior view, the nasal is dorsoventrally thin with the cross section platelike. The lateral border of the nasal smoothly curves lateroventrally to meet the maxilla with the facial exposure of the nasal narrow. The suture of the nasal and maxilla is almost horizontal. The maxilla is large and high, and trapezoidal in outline in lateral view. Above the maxillary diastema (between the alveoli of the upper canines and upper premolars), the lateral surface of the maxilla is slightly concave. The preorbital fossa is present, shallow and smooth on the lateral surface. The infraorbital foramen is a distinctive and large opening, situated above P3. The maxilla posteriorly contacts the zygomatic and lacrimal by forming a nearly vertical suture on the lateral surface. The postorbital process of the frontal is well developed, situated above the posterior border of M3. The dorsal roof of the orbital region is formed by the nasal extended posterolaterally and the wide frontal. The lacrimal has a rough surface with some damages and its anterior and ventral sutures are identified, nearly forming a right angle. The zygomatic contacts the lacrimal nearly at the middle point of the anterior orbital margin. The zygomatic is flat on the lateral surface, rough on the ventral surface, and shows a well-developed ridge along the ventrolateral margin. The zygomatic arch is not preserved, and rises slightly posteriorly. The poor preservation of the orbital region makes it hard to accurately identify particular foramina and fissures, except the one regarded as the optic foramen on the left side (fig. 2). Ventral to the optic foramen is the foramen orbitale and the foramen rotundum, but the boundary of the two openings is not preserved. Laterally, the zygomatic process of the squamosal is dorsoventrally thin. The zygomatic arch is badly damaged on the both sides. The glenoid fossa is widely open. The squamosal ventrally bears the postglenoid process, which possesses two articular facets divided by a ridge, the medial facet facing anteriorly and the lateral facet facing anterolaterally. Posteriorly, the posttympanic process is widely separated from the postglenoid process. The posttympanic process is partially fused with the paroccipital process posteriorly, and the former is robust and anteroposteriorly thickened, whereas the latter is more slender, slightly longer and nearly vertically directed. Lateral to the posttympanic process and the paroccipital process is the posttympanic-paroccipital foramen with a large opening at the dorsal extremity (equivalent to mastoid foramen) (fig. 2) (Holbrook, 2001). The occipital extends posteriorly beyond the level of the occipital condyle.

TABLE 3.

Measurements of mandible and lower dentition of Forstercooperia ulanshirehensis (mm). H = height; L = length; W = width.

FIG. 2.

Cranium of F. ulanshirehensis (AMNH 21608) in lateral view. Abbreviations: for, foramen orbitale; fr, frontal; fro, foramen rotundum; iof, infraorbital foramen; ju, jugal; lc, lacrimal; mr, median ridge of postglenoid process; mx, maxilla; na, nasal; oc, occipital condyle; opf, optic foramen; pa, parietal; pf, preorbital fossa; pgp, postglenoid process; pop, postorbital process; pocp, paroccipital process; ppf, posttympanic-paroccipital foramen; ptp, posttympanic process; sc, sagittal crest; sq, squamosal; tc, temporal crest.

FIG. 3.

Cranium of F. ulanshirehensis (AMNH 21608) in dorsal view. Abbreviations: fr, frontal; iof, infraorbital foramen; ju, jugal; mx, maxilla; na, nasal; nf, nutrient foramen; pa, parietal; sc, sagittal crest; sq, squamosal.

In dorsal view (fig. 3), the tip of the nasal might extend anteriorly more than the level of the upper canine, judging from the width of the posterior part of the nasal. The nasal increases in width posteriorly, and extends lateroposteriorly to the anterior part of the postorbital process, contacting the frontal at the level of the anterior border of the orbit. The frontal contributes to the widest part of the cranium at the level of the postorbital process, and decreases in width posteriorly. The frontoparietal crests gradually rise from the postorbital process posteriorly, and fully merge together slightly anterior to the suture of the frontal and parietal (∼10 cm anterior to the posterior tip of the cranium) forming the well-developed sagittal crest. The sagittal crest slightly elevates posteriorly. The dorsal roof decreases in width from the postorbital process posteriorly. The posterior part of the cranium is clearly narrower, with the dorsal profile steep and slightly higher than the anterior, especially the parietal and occipital regions. There are at least seven nutrient foramina present on the dorsal surface of the parietal and squamosal with some variation in size, of which a pair of nutrient foramina are nearly symmetrically positioned along the sagittal crest and distinctively larger than the others.

FIG. 4.

Cranium of F. ulanshirehensis (AMNH 21608) in ventral view. Abbreviations: bo, basioccipital; bs, basisphenoid; gch, partial great cornu of hyoid; hf, hypoglossal foramen; ju, jugal; mr, median ridge of postglenoid process; mx, maxilla; oc, occipital condyle; of, oval foramen; pac, posterior opening of the alisphenoid canal; pal, palatine; pgp, postglenoid process; pocp, paroccipital process; ppa, pterygoid process of alisphenoid; ptp, posttympanic process.

In ventral view (fig. 4), the anteriormost part of the cranium is the large upper canine alveoli. Posteriorly, the diastema between the upper canines and upper premolars is short (∼3 cm), and the ventral margin of the diastema is slightly concave. The horizontal part of the palatine between the cheek-teeth rows is slightly concave rather than flat with a smooth ventral surface. The anterior process of the palatine is not preserved, and the number of the incisive foramina is unknown. The posterior border of the palatine is located at the level of the posterior border of M2. The suture between the maxilla and the jugal is distinct, nearly parallel to the long axis of the cranium. A short ridge that is similar to the facial crest rises from the anterior suture between the maxilla and jugal, and anteriorly shrinks at the level of the M2. Ventrally, the pterygoid process of the alisphenoid and the pterygoid process of the palatine form a small concave plate with an uneven ventral surface potentially for muscular attachment of the m. pterygoideus. Posteriorly, the pterygoid process of the alisphenoid bears a sharp ventral margin. Lateral to the alisphenoid, there are two large openings, of which the anterior is the posterior opening of the alar canal that opens posteriorly, while the posterior is the oval foramen that opens anteriorly. The cavity of the posterior choana has been fully filled by layers of plaster during preparation, and so it is not possible to identify many details in this region (e.g., the vomer and sphenoid). The posterior part of the cranium is elongate (the length from the postglenoid process to the occipital condyle ∼96.5 mm). The postglenoid process is massive and transversely narrow. The glenoid fossa is extensive with a distinct component on the ventral surface of the zygomatic process of the squamosal. There is no sign of the postglenoid foramen posterior to the postglenoid process. On the left side, the posttympanic process is distinctly inflated and anteroposteriorly thickened despite the damage to the ventral part. Posteriorly, the paroccipital process is slender, transversely wider and anteroposteriorly thinner than the posttympanic process, and the two processes are partially fused. The basioccipital is well developed and ventrally bears a distinctive basilar process. Posterior to the basilar process is a distinct median ridge on the ventral part of the basioccipital. The hypoglossal foramen is large and medial to the paroccipital process. A fragmentary bone that is connected with the base of the temporal bone on the right side is the posterodorsal part of the great cornu of the hyoid bone. The foramen magnum has an oval outline in posterior view, formed by a pair of large occipital condyles. The occipital profile is nearly vertical. Dorsally, the occipital crest is damaged.

The whole mandible of IVPP V20155 is well preserved, but slightly distorted nearly in the middle and slightly damaged at the right condyle (figs. 5, 6). The mandibular symphysis is short, narrow, and dorsoventrally deep, making the anterior dentition tightly arranged. The posterior border of the mandibular symphysis is situated at the boundary between p1 and p2. The diastema between the lower canine and p1 is rather short (34.5 mm) and transversely constricted, forming a distinctive lateral depression. The horizontal ramus is slender with a slightly convex ventral border. On the left side, there is a large mental foramen below the anterior border of p1. On the right side, there are four mental foramina: the anterior one situated below i3, the middle two close to each other lying below the diastema, and the posterior one situated below p2. The lateral surface of the horizontal ramus is smooth, but slightly concave below p1–p4. The mandibular angle is rounded with the ventral border of the horizontal ramus slightly convex, and the posterior border of the ascending ramus nearly vertical. Both the lateral and medial surfaces of the mandibular angle are considerably rough, of which the latter is more concave than the former. Laterally, ventral to the coronoid process is the masseteric fossa, represented by a shallow depression with a nearly triangular outline. Medially, the mandibular foramen is large, situated slightly lower than the alveolar border and vertically below the coronoid process. The coronoid process is inclined posteriorly at the top, and is higher than the mandibular condyle. The condyle is quite high above the occlusal plate of the cheek teeth. The surface of the condyle is rough and fairly flat dorsally. Posterior to the condyle, the postcotyloid process is rather weak, showing a convex posterior margin of the mandible in lateral view. The mandibular incision is broad and opens posteriorly.

FIG. 5.

Mandibles of F. ulanshirehensis (IVPP V20155) in occlusal view. Abbreviations: cdp, condylar process; crp, coronoid process; pcp, postcotyloid process.

The dental formula is ?·1·4·3/3·1·4·3. The dentition of AMNH 21608 (fig. 4) and AMNH 22101 (fig. 8) are moderately worn, whereas the dentitions of IVPP V20154 (figs. 7A, 7B), IVPP V20155 (figs. 5, 6), IVPP V20156 (figs. 7C, 7D) are less worn.

The P1 is preserved only on the left side of AMNH 21608. This tooth is double rooted and heavily worn. Its outline is nearly trapezoidal with the length larger than the width in occlusal view. The paracone is the largest and highest portion of the tooth. The cingulum is distinct and continuous in the lingual and labial side. P2 is larger than P1, nearly quadrate in occlusal view with the protoloph well developed and the metaloph (or metaconule) weak. The protoloph of P2 is oriented anterolaterally to contact the ectoloph with a notch at the base of the paracone. The metaloph is short and extends posterolingually to contact the protoloph. From the wear surface, the protocone and hypocone of P2 are slightly separate, because there is a small isolated wear facet immediately posterior to the posterolingual margin of the protoloph, where the small hypocone is located but that facet becomes invisible after moderate or heavy wear. Despite the damage to the ectoloph, the paracone and metacone of P2 is nearly equal in height, and the parastyle fold is shallow. The cingulum of P2 is well developed at the lingual, anterior, and posterior bases. The P3 is similar to P2 in basic structure, but larger in size. The ribs of the paracone and metacone are distinct on the lateral side of the ectoloph. The protoloph of P3 is more robust than that of P2, occupying most of the lingual space of tooth crown. The metaloph of P3 is weak and short, and extends nearly transversely. In IVPP V20154, the metaloph of P3 contacts the protoloph, but it is not the case in AMNH 21608, in which there remains a tiny gap between the protoloph and the short metaloph. The postfossette of P3 (space posterior to the metacone) is broad compared to those of P1–P2. The cingulum of P3 is similar to that of P2, but absent on the buccal side. The P4 is rectangular in occlusal outline with its width greater than its length. The parastyle is more distinct than in P1–P3, and transversely pinched, forming a deep parastyle fold with the paracone. A weak crista is present in the lingual side of the ectoloph of P3–P4. The protoloph of P4 is massive on the lingual side, and sharply decreases in width buccally to contact the ectoloph. The metaloph of P4 is slightly longer and more robust than that of P3, and transversely extends to contact the protoloph. The postfossette of P4 is relatively transversely wider than that of P3. The cingulum of P4 is similar to that of P3.

FIG. 6.

Mandibles of F. ulanshirehensis (IVPP V20155) in lateral view. Abbreviations: cdp, condylar process; crp, coronoid process; mf, mandibular foramen; pcp, postcotyloid process.

The M1 is heavily worn and nearly quadrate in occlusal outline with distinct parastyle and parastyle folds on the buccal wall. The paracone rib is well developed, while the metacone rib is faint. The M2 is similar to M1 in basic structure, but distinct from the latter in having a larger size, relatively more elongated and lingually oriented ectoloph, and shorter metaloph. The M3 is distinguished from M1–2 with a nearly triangular outline. The parastyle of M3 is slightly weak, situated at the anterolateral base and slightly warped toward the lingual side. The ectoloph of M3 is completely fused with the metaloph. The metacone of M3 can be traced by the presence of a slightly concave posterior surface of the metaloph in AMNH 21608, but it is completely absent in IVPP V20154. The cingulum of the upper molars is distinct at the anterior, lingual, and posterior bases, and absent on the buccal side.

The lower incisors are preserved only in IVPP V20155. The left i3 is not preserved and the right i3 is damaged at the mesial side. Three incisors are tightly arranged nearly longitudinally. The lower incisors are generally subequal in size, but the i2 is slightly larger than the i1 and i3. The occlusal outlines of the lower incisors are similar in having somewhat conical and nearly symmetric crowns with distinct mesial and distal flanges. Each incisor bears a blunt ridge on the distal (lingual) wall of the tooth crown. The first lower incisor is relatively procumbent with its tooth crown more asymmetric than in i2 and i3. The crown of i2 extends much more laterally (transversely) than anteroposteriorly, and possesses a triangular outline in posterior view. The distal flange is well developed in i2, while slightly weaker in i1 and narrower in i3. The right i3, represented by only the distal portion, is similar to i2 in outline, except it aligns nearly anteroposteriorly.

The lower canine is well preserved on both sides of IVPP V20155 and V20156. There is no distinct diastema between the i3 and lower canine. This tooth is conical, rather stubby and much larger than lower incisors with its base more rounded, especially in IVPP V20156. Anteriorly, a distinct flange smoothly extends from the top to the mesial (lingual) base of the tooth crown. Posteriorly, this tooth bears two flanges, which descend from the tip to the base of the tooth crown, forming a triangular facet on the distal side. The ridges are evident in IVPP V20155, but slightly weaker in IVPP V20156. The cingulid of the lower canines at the base is weak, and present only at the anterior base.

FIG. 7.

Dentition of F. ulanshirehensis in lateral and occlusal view. A, IVPP V20154 in lateral view; B, IVPP V20154 in occlusal view; C, IVPP V20156 in occlusal view; D, IVPP V20156 in lateral view.

The p1 is double rooted. It is roughly bladelike in occlusal view, but wide with a convex lateral wall. The trigonid of p1 is simply structured and mainly comprises a weak paraconid and a distinct protoconid, which are nearly anteroposteriorly arranged, making the trigonid basin shallow. The paralophid is represented by a very short ridge. The talonid is short, and its short hypolophid forms a shallow talonid basin. The cingulid is weak and present only at the anterolingual and posterolingual base. In IVPP V20155 the anterior ridge extends more anterolingually, deepening the trigonid basin more than in IVPP V20156.

The p2 has a nearly rectangular base in occlusal view, and is larger than p1. The trigonid of p2 is wider, and possesses a larger protoconid than that of p1. The protolophid of p2 anterolingually extending makes the trigonid basin a little deeper than that of p1. The paralophid of p2 is short and the metaconid is not distinctly separated from the protoconid. The trigonid of p2 is higher, longer, and narrower than the talonid. The talonid is primitive with a short and lingually oriented hypolophid and absent entoconid. The talonid is surrounded by a short ridge around the posterior base of the talonid. The cingulid is distinct anteriorly and posteriorly, and absent laterally and lingually in p2. In IVPP V20156, the p2s are slightly damaged on the lingual side. In IVPP V20155, the metalophid is oriented relatively posterolingually.

The p3 is basically similar to p2, but larger. It also differs from p2 in having a longer paralophid and a higher hypoconid in the talonid. The metalophid of p3 is oriented more transversely than that in p1 and p2. The trigonid of p3 is slightly shorter than the talonid, in which the hypoconid is much higher than that of p2. Both the trigonid and the talonid of p3 are distinctly deeper and wider than those of p2. The cingulid of p3 is well developed at the anterior and posterior bases, continuous at the buccal base but absent on the lingual side.

The p4 is larger than p3, but with blunter cusps and lophs. The paraconid of p4 is located more lingually and the metalophid of p4 is oriented more transversely than in p1–3, making the trigonid U-shaped and the trigonid basin deep. The talonid of p4 is almost equal in length with the trigonid. Compared with p1–p3, the remarkable difference is the presence of the entoconid in p4. The entoconid of p4 is still weak and isolated from the hypoconid, which is much higher in the talonid. The cingulid of p4 is the same as that of p3.

The m1 is well preserved with intermediate abrasion in IVPP V20155 and heavy abrasion in AMNH 22101. This tooth is distinctly larger than lower premolars. The trigonid of m1 is wide with the paralophid long and laterally extended, and the protolophid of m1 more anterolingually oriented. The metalophid of m1 is strong and extends slightly posterolingually. The talonid of m1 is wide with a well-developed hypolophid (“cristid obliqua”) and entolophid (“hypolophid”). The entoconid of m1 is a little more posteriorly displaced than the hypoconid of m1. The hypolophid contacts the entolophid with an oblique angle. The cingulid of m1 is distinct anteriorly and posteriorly, and absent lingually, whereas the buccal cingulid is continuous, but slightly reduced at the base of the protoconid and hypoconid. The m2 is basically similar to m1, but larger than the latter. The m2 is less worn in occlusal view than m1. Its trigonid is longer than that of m1. The trigonid of m2 as well as that of m1 slightly increase in width posteriorly. The m3 is similar to m2 in crown outline. The m3s have not completely erupted in IVPP V20155. The cingulid of m2–3 is roughly the same as that of m1, except that the posterior cingulid is more distinctive in m3.

FIG. 8.

Dentition of F. ulanshirehensis in occlusal and lateral view (AMNH 22101). A, occlusal view; B, lateral view.

Comparison within Forstercooperiiines

The new specimens display characteristic features of Forstercooperiinae, such as the conical and nearly equal-sized lower incisors, large and stubby canines, primitive upper and lower premolars, and large size (Radinsky, 1967; Lucas et al., 1981; Lucas and Sobus, 1989). Specifically, the new specimens are unequivocally referable to Forstercooperia rather than Pappaceras, given the derived upper and lower premolars, particularly P2 and p4, larger body size (fig. 9) and younger age (Irdinmanhan). But they are distinguishable from the specimens of Forstercooperia totadentata and Forstercooperia mazhuangensis primarily based on the dental morphology, such as smaller size (fig. 9), primitive P2, short metalophs of P3–P4, and weak entoconid of p4. Consequently, we establish a new species, Forstercooperia ulanshire-hensis, on the basis of these specimens, which is the third and best-known species of Forstercooperia. Besides the new species, Forstercooperia includes two species: the type species F. totadentata Wood 1938, which is represented by a considerably large, fragmentary cranium (AMNH 20116), and Forstercooperia mazhuangensis Qiu and Wang 2007, which is represented by a pair of fragmentary maxillae and mandibles (IVPP V5072). The new species, F. ulanshirehensis, is distinguished from F. totadentata and F. mazhuangensis by dental morphology (mainly the upper and lower premolars) and its smaller size. It is similar to F. totadentata (AMNH 20116) with the presence of large upper canines (based on the large alveoli of upper canines), complete upper premolars (P1–4), short diastemata, a shallow preorbital fossa, transversely wide P4, distinct crista, and straight metaconules of P3–4. But F. ulanshirehensis differs from F. totadentata in having transversely slightly narrower P2–3, a weak hypocone on P2, less-developed ribs of the upper premolars, and short metalophs of P3–4. In terms of size, F. ulanshirehensis is about 25% smaller than F. totadentata with the length of upper premolars about 81.4 mm for F. ulanshirehensis and about 108 mm for F. totadentata (fig. 9, table 2). F. ulanshirehensis is distinguishable from F. mazhuangensis with the short and posterolingually oriented metaloph of P2, transversely straight metaloph of P4, presence of entoconid of p4, and a smaller size (the length of P1–4 ∼103 mm for F. mazhuangensis, ∼20% larger than F. ulanshirehensis) (fig. 9, table 2).

The cranial features of F. ulanshirehensis display a derived condition within forstercooperiines. F. ulanshirehensis (AMNH 21608) shares primitive cranial structures with Pappaceras, such as a shallow preorbital fossa, high maxilla, wide frontal, greater ratio of the length of basal cranium to skull length, enlargement of posttympanic process, and the presence of posttympanic-paroccipital foramen and large occipital condyles. Compared with the well-preserved specimen of Pappaceras meiomenus (IVPP V20254) (Wang et al., 2016), F. ulanshirehensis is distinctively different in its (1) larger size, (2) enlarged occipital condyles, and (3) longer distance between the postglenoid process and the occipital condyle, which is about 26% of the distance between P1 and the occipital condyle (table 1). These three features of F. ulanshirehensis are also characteristics of paraceratheriines (giant rhinos) (Qiu and Wang, 2007). The posttympanic process of F. ulanshirehensis is thicker than that of P. meiomenus. The infraorbital foramen and oval foramen appear to be considerably larger openings in F. ulanshirehensis than in P. meiomenus. In F. ulanshirehensis, the postglenoid process is transversely narrow, and divided by a more developed median ridge into two facets, facing anteriorly and anterolaterally respectively, whereas it is transversely broad and relatively flat with a weak median ridge in P. meiomenus. The basicranium is broken in the holotype of P. confluens (AMNH 26660). F. ulanshirehensis (IVPP V20155) has five large mental foramina, one from the left and four from the right, whereas only one mental foramen is present below the p1 in P. confluens. The ventral border of the mandible in F. ulanshirehensis is relatively horizontal, while P. confluens and P. minuta have ventral borders that are slightly convex along with the presence of a distinct vascular impression (Wood, 1963; Lucas et al., 1981). In addition, the mandible of F. ulanshirehensis differs from that of Pappaceras in having a nearly vertical posterior border of the ascending ramus and a more rounded mandibular angle.

FIG. 9.

Scatter plots of length vs. width across P4/p4 and M1/m1 of forstercooperiines. A, scatter plots of P4; B, scatter plots of M1; C, scatter plots of p4; D, scatter plots of m1.

F. ulanshirehensis provides the complete morphology of upper check teeth of Forstercooperia. The presence of P1 is consistent in Forstercooperia, including F. ulanshirehensis, and is different from that of Pappaceras meiomenus. The P2s of Forstercooperia show the rudimentary distinction of the protocone and hypocone in the lingual side, in contrast to the absence of a hypocone of P2 in Pappaceras. The structure of M1–2 in F. ulanshirehensis is similar to those of P. confluens and P. minuta in having relatively elongated ectolophs and lingually oriented metastyles. Another evident morphological difference in upper cheek teeth between F. ulanshirehensis and Pappaceras is that the metacone of M3 in F. ulanshirehensis is nearly absent, similar to that of F. mazhuangensis, forming a triangular occlusal outline. This is distinguished from Pappaceras, in which the M3 is either trapezoid in outline with the metacone lingually oriented and the ectoloph confluent with the metaloph in P. confluens, or almost quadrate with distinct metacones and metastyles for P. minuta and P. meiomenus. In addition, the antecrochets of M2–3 of F. ulanshirehensis are absent, more similar to F. mazhuangensis, P. minuta, and P. meiomenus, but different from P. confluens.

The morphology of the lower dentition of Forstercooperia, especially the anterior dentition, is comprehensively represented by the specimens of F. ulanshirehensis (IVPP V20155, V20156; AMNH 22101). It is confirmed from F. ulanshirehensis that the status of i1 of Forstercooperia is as follows: (1) i1 is almost equal to i2 in size, only slightly smaller than the latter; (2) the i1 is less conical and less symmetric in outline with distinct mesial and distal flanges than i2 and i3, similar to those of most primitive perissodactyls; (3) the i1s are closely situated medially and the roots rise much more vertically rather than anteriorly compared with those of paraceratheriines. The crowns of i2s in F. ulanshirehensis are more laterally extended, which is distinct from those of P. confluens and P. minuta. The lower canine of F. ulanshirehensis is immediately next to i3, similar to that of P. confluens, and it is considerably stubbier and larger than those of Pappaceras, especially at the base. Meanwhile, compared with the inferred characters of lower dentition of F. totadentata by Wood (1938), several differences that are revealed by the new mandibular specimens are in contrast to Wood's inference. F. ulanshirehensis has large lower canines, which are nearly equal in size to the upper canine based on the alveolus of upper canines, rather than smaller than the upper canines as inferred by Wood (1938). The p1 is primitive and large on both sides rather than the inferred reduced or absent condition. Within forstercooperiines, the reduction of p1 is present only in P. meiomenus. The p3 is relatively shorter (>30 mm) in F. ulanshirehensis than the inferred length in F. totadentata, and it is probably larger than 30 mm for F. totadentata. In contrast to Wood's inference that the p4 of Forstercooperia was shorter than p2–3, which corresponded to the transversely wide and longitudinally short P4, the p4 of F. ulanshirehensis is clearly longer than p2 and p3. Furthermore, the p4 bears an entoconid in F. ulanshirehensis, and this is a derived characteristic of Forstercooperia within forstercooperiines. On the basis of these characters, it is highly possible that the p4 (even p3) of F. totadentata may possess a distinctive entoconid, considering that the P2 in F. totadentata is more molarized than in F. ulanshirehensis with the protocone and the hypocone partially separate. The length of the lower premolar series is about 92 mm in F. ulanshirehensis, longer than Wood's inference (83 mm) for F. totadentata, which should be much larger.

Species-Level Taxonomy of Forstercooperiines

Excluding the new species, there are 12 species of forstercooperiines (“Forstercooperia”) formally reported, and only six species are regarded as valid in the present study after examining all specimens formerly assigned to Forstercooperiinae. All species that were formally created in previous studies are listed here (invalid names marked by quotation marks): Forstercooperia totadentata Wood, 1938, Pappaceras confluens Wood, 1963, Hyrachyus grandis Peteson, 1919 (“Forstercooperia grandis Radinsky, 1967″), Eotrigonias borissiaki Beliayeva, 1959 (“Forstercooperia borissiaki Lucas et al., 1981″), “Forstercooperia ergiliinensis Gabounia and Dashzeveg, 1974,” “Forstercooperia shiwopuensis Chow et al., 1974,” “Forstercooperia jigniensis Sahni and Khare, 1973,” “Forstercooperia crudus Gabounia, 1977,” “Forstercooperia minuta Lucas et al., 1981,” “Forstercooperia huhebulakensis Qi, 1987,” Imequincisoria mazhuangensis Wang, 1976 (Forstercooperia mazhuangensis Qiu and Wang, 2007), and Pappaceras meiomenus Wang et al., 2016. Currently, three species of Forstercooperia, F. totadentata, F. mazhuangensis, and F. ulanshirehensis, are regarded as valid, while Pappaceras confluens (“Forstercooperia confluens”) has been resumed and Pappaceras minuta (“Forstercooperia minuta”) has been regarded as a composite on the basis of the validity of Pappaceras. “F. shiwopuensis” has been accepted as a junior synonym of F. totadentata based on previous works (Lucas et al., 1981; Lucas and Sobus, 1989), with which we agree in the present paper. Another three species, “F. borissiaki,” “F. ergiliinensis,” and “F. jigniensis,” have been correctly excluded from Forstercooperia (Lucas et al., 1981; Lucas and Sobus, 1989; Qiu and Wang, 2007). The type specimens of “F. grandis” have been assigned to Uintaceras by Holbrook and Lucas (1997); as a result, F. grandis is no longer a valid name.

The latest study of forstercooperiines erects Pappaceras meiomenus based on a complete cranium (IVPP V20254) (Wang et al., 2016). This specimen is distinguished by the loss of P1. It should be noted that the specimens assigned to Pappaceras minuta (“F. minuta”) by Lucas et al. (1981) show a mixed condition on the reduction of P1/p1, which was interpreted as interspecies variation at the time. In the two type specimens of P. minuta (AMNH 26672), despite the lack of P1, a distinct alveolus for P1 on the fragmentary maxilla confirms its presence, while the p1 is hardly visible from the damaged mandible of a young juvenile individual due to poor preservation. Considering the completeness of premolar series in most primitive perissodactyls, the interpretation that regarded the absence of P1/p1 in P. minuta as intraspecific variation can be seriously challenged. Alternatively, it is proposed that the questionable specimens formerly assigned to P. minuta with the absence of P1/p1 are temporarily referred to P. meiomenus. These specimens include AMNH 26677, 107851, 107852, 107856 (see appendix). In other cases that cannot directly estimate the presence or absence of P1/p1 in a particular specimen, we are temporarily inclined to refer these specimens tentatively as Pappaceras sp. (including AMNH 26669, 26671, 107848, 107849). In addition, a juvenile left fragmentary dentary with deciduous p2–3 (AMNH F: AM 99662) has been regarded as an indeterminate species (Holbrook and Lucas, 1997). The specimens that involve taxonomic status changes are listed in the appendix. Accordingly, the content of three species of Pappaceras has been changed after the current revision.

The confusion of “F. crudus” was discussed and tentatively considered a nomen nudum in the revision work for forstercooperiines (Lucas et al., 1981). The specimens assigned to “F. crudus” are from the Sargamys Formation of the Zaissan Basin in Kazakhstan (Gabounia, 1977). Gabounia (1977) failed to provide a formal and detailed description of involved specimens in the original text, in which an isolated ?M2 and a fragmentary mandible with p2–m3 were referred to in the text as “figure 17″ and “figure 19″ respectively (Gabounia, 1977: 31). The plate on which the images appeared, however, were not labeled but figures, but rather were labeled alphabetically (so fig. 17 would correspond to q and fig. 19 to s); however, the caption describing the isolated ?M2 (“q” on pl. 1) was omitted for some reason (along with its scale) (Gabounia, 1977: 36), whereas the mandibular specimen (“s”) was confusingly described as “Forstercooperia sp.” in the caption (Gabounia, 1977: 36, pl. 1). Without a standard for ?M2 in the plate, it is hard to measure the size of this tooth accurately and assign it to any specific genus, because it is similar to those of Forstercooperia and some hyracodontids in having a relatively elongated ectoloph, such as in Prohyracodon and Epitriplopus. As a result, it is not possible to confirm its affinity based on an incomplete and less characteristic M2, even if the size of this tooth could be determined in the future. In terms of the mandible specimen, Lucas et al. (1981) mentioned that the mandible is similar to P. minuta in size, but has derived p3–4 with distinct entoconids and relatively strong entolophids. In general, the derived features of lower premolars in the Zaissan specimen are distinctively in contrast to Forstercooperia and even to most of the Middle Eocene rhinocerotoids, which usually lack entolophids on p3–4. In addition, the cheek teeth of the Zaissan specimens were measured according to the plate, and the length of m1–3 is about 68.0 mm, which is remarkably smaller than that of Forstercooperia. Therefore, it seems unreasonable to assign this specimen to Forstercooperia under any circumstance. These specimens might be related to some derived hyracodontids based on dental morphology, and therefore, we suggest considering “F. crudus” as an indeterminate rhinocerotoid.

The species “F. huhebulakensis” is not a valid species of Forstercooperia. It was proposed that F. huhebulakensis occurred stratigraphically lower than “F. confluens” (Qi, 1987), but recent studies have suggested the argument is incorrect (Meng et al., 2007; Wang et al., 2010). Meng et al. (2007) and Wang et al. (2012) confirmed that the holotype (IVPP V5770) of “F. huhebulakensis” was unearthed from the “upper gray clays” of the Arshanto Formation in the Huheboerhe area, which is stratigraphically equivalent to that of P. confluens. Meanwhile, Qi (1987) failed to compare “F. huhebulakensis” with P. minuta, and he suggested that “F. huhebulakensis” was smaller than F. totadentata, and differed from P. confluens in having a “flatter and straighter” zygomatic arch, “flatter frontal and nasal bones” and M2–3 without anticrochets. Actually, the holotype of “F. huhebulakensis” has a characteristic M3 of P. confluens with the metacone lingually oriented and the ectoloph distinctively reduced. The length of M1–3 in “F. huhebulakensis” (79.7 mm) is slightly smaller (8%) than that of P. confluens (87.1 mm) and considerably smaller than that of F. totadentata (123.0 mm). The zygomatic and frontal of “F. huhebulakensis” are also similar to those of Pappaceras (P. minuta and P. meiomenus). Consequently, the holotype of “F. huhebulakensis” should be assigned to P. confluens as a relatively small individual, making “F. huhebulakensis” a junior synonym of P. confluens.

Furthermore, we agree with Qiu and Wang (2007) that “Imequincisoria mazhuangensis Wang, 1976″ should be combined with Forstercooperia mazhuangensis. “Imequincisoria mazhuangensis” was created based on a pair of fragmentary maxillae, mandibles, and some isolated anterior teeth (IVPP V5072) (Wang, 1976). Among these materials, the maxillae and mandibles have been assigned to F. mazhuangensis, whereas the anterior teeth have been put into Juxia micracis Qiu and Wang, 2007 (Qiu and Wang, 2007). The morphology of upper premolars in F. mazhuangensis is more similar to that of Forstercooperia than Juxia; in particular, the metalophs of P2–3, though relatively well developed, are distinctively transversely straight, the protocone and metacone of P2 are barely separate and the P4 is transversely wide in F. mazhuangensis. With regard to the fragmentary mandibles, the entoconid is absent in the lower premolars of F. mazhuangensis, primitively similar to Pappaceras, but distinct from Juxia in which the entoconid is usually present in p3–4. The anterior dentition that has been assigned to J. micracis is, though generally conical in outline, clearly smaller and narrower than those of F. ulanshirehensis, and considerably varies in size, which links this specimen to Juxia rather than forstercooperiines. In addition, Qiu and Wang (2007) considered that the large-sized postcranial specimens (carpals and metacarpals) (IVPP V5074.1-8) from the Wucheng basin formerly assigned to Imequincisoria (?) sp. would be referable to F. mazhuangensis, mainly because of its larger size and tridactyl manus. Given that there remains no known unequivocal postcranial specimens of forstercooperiines, the taxonomy of the mentioned specimens (IVPP V5074.1-8) is in need of further discussion and revision. Namely, these specimens are tentatively regarded as indeterminate rhinocerotoid, and not used as coding resources of forstercooperiines in phylogenetic analyses.

The specimens from the Lunan Basin that were referred to as Forstercooperia sp. (PGM V799, PGM V806, and PGM V 07, probably from one individual) are slightly larger than P. confluens in size (Chow et al., 1974). The complete M3 (PGM V799) resembles that of P. confluens, but differs from that of Forstercooperia by having a trace of metacone, straight protoloph, short ectoloph, and well-developed lingual cingulum. In addition, the fragmentary mandibles (PGMV 806, PGM V807) have a shallow mandibular ramus (Chow et al., 1974), which is also similar to that of P. confluens and distinguished from F. ulanshirehensis. These specimens are likely referable to P. confluens.

A cranial fragment with complete cheek teeth and canine alveoli (IVPP V5075) was assigned to Pappaceras sp. (Wang, 1976). This specimen is typically characteristic of forstercooperiines in having large upper canines (represented by large alveoli), and transversely wide outlines of upper premolars. Despite its larger size (length of M1–3 ∼105.5 mm), it uniquely resembles P. confluens in sharing a trapezoidal outline of M3 with a distinct trace of metacone. There seems no doubt that this specimen has an affinity with Pappaceras, probably P. confluens, but the hypothesis that variation in body size is related to sexual dimorphism in Pappaceras requires further testing.

Zhai (1977) assigned a fragmentary maxilla with M1–3 and some isolated check teeth (IVPP V5241) to Forstercooperia sp. (“Imequincisoria sp.”), this assignment appears inappropriate as suggested by Qiu and Wang (2007), because it is distinct from forstercooperiines with the hypocone of P2 lingually positioned, the metaloph of P3 posteriorly extended, the ectoloph of M1–2 straight, and the well-developed entoconid of p4. Meanwhile, the evidence for the age of the Changxindian Formation is apparently contradictory based on the records of fossil mammals, fossil invertebrates, and plants (including gastropods and pollens) in Zhai's proposal. These rhinocerotoid specimens might belong to later derived rhinocerotoids (possibly late Eocene taxa) rather than forstercooperiines.

Ding et al. (1977) referred a complete p3 (IVPP V5108.1) and p4 (IVPP V5108), and incomplete left m1–2 (IVPP V5109) as Forstercooperia sp. The tooth crown of IVPP V5109 is badly damaged and less characteristic for detailed comparisons. The p3 and p4 (IVPP V5108.1 and V5108) are at least 20% larger than those of F. ulanshirehensis (IVPP V20155 and V20156), furthermore they have well-developed entoconids and entolophids, which are rarely present in Forstercooperia. Therefore, the two specimens are unlikely to belong to Forstercooperia, and might be referable to paraceratheriines.

Comparisons with Other Early Rhinocerotoids

Forstercooperia is one of the most poorly known rhinocerotoid genera due to its poor preservation. Previously, forstercooperiines were usually assigned to the broad hyracodontid group together with paraceratheriines, mainly considering the divergence of anterior dentition within rhinocerotoids (Radinsky, 1967; Prothero et al., 1986). In the latest study, the craniodental morphology of Pappaceras has been briefly compared with representatives of other early rhinocerotoids (Wang et al., 2016), Here comparisons between forstercooperiines and other early rhinocerotoids are further expanded by adding complete craniodental morphology of Forstercooperia, which also lays the grounds for improved exploration of the phylogeny of early rhinocerotoids.

Hyrachyus is commonly considered as ancestral or sister taxon to rhinocerotoids, though its taxonomy is in need of revision (Wang et al., 2016; Bai et al., 2017). Generally, forstercooperiines share several primitive features with Hyrachyus that are characteristic of primitive tapiromorphs, such as a dolichocephalic cranium structure, long nasal, shallow narial notch, premaxilla-nasal contact, slender zygomatic, complete dentition (except for P. meiomenus), and primitive premolars. However, the new specimens of F. ulanshirehensis demonstrate there remains a distinct morphological gap between Hyrachyus and Forstercooperia. Hyrachyus lacks a distinctive preorbital fossa, which is shallow in Forstercooperia. The distance between the postglenoid process and the occipital condyle is short in Hyrachyus (Hyrachyus eximius) (13% of the skull length) (Wood, 1934), whereas it is much longer in Forstercooperia (AMNH 21608) (>25% of the skull length). Similar to most primitive tapiromorphs and different from forstercooperiines, the posttympanic process of Hyrachyus is considerably shorter than the paroccipital process, which is slender, oriented slightly posteriorly, and completely separated from the posttympanic process. Forstercooperia differs from Hyrachyus in several derived dental features: well-developed transverse lophs of cheek teeth, small parastyles of cheek teeth, slightly separate protocone and hypocone of P2, clearly reduced metacone and metastyle of M3, relatively elongated talonids of p2–4, and the presence of an entoconid in p4. The similarity of cranial morphology between Forstercooperia and Hyrachyus is indicative of the primitive status of the former, but the craniodental differences between them reveal that Forstercooperia is an unequivocal rhinocerotoid, more derived than Pappaceras, with several aforementioned derived dental features, such as triangular M3, which are shared by most of the later rhinocerotoids.

Triplopus is regarded as the most primitive and typical hyracodontid with its earliest record in Early Uintan (NALMA) North America (Prothero, 1998; Prothero et al., 1986). It is contemporary with Forstercooperia (F. totadentata and F. ulanshirehensis). Triplopus cubitalis, which was erected, described, and measured based on an incomplete skeleton (AMNH 5095) by Cope (1884), is considerably smaller than Forstercooperia (with the length of the skull ∼128 mm). The facial region of T. cubitalis is distinctly shorter than that of Forstercooperia, with the anterior border of orbital situated at about one-third of the cranial length (though its incompleteness of the anterior tip). Compared with Forstercooperia, T. cubitalis has a much broader frontal that distinctively overlaps the orbital region with a ratio of the zygomatic width to the skull length about 1:2 (zygomatic width 64 mm, length of skull 128 mm in AMNH 5095), whereas the ratio is probably less than 40% in Forstercooperia (zygomatic width slightly more than 182 mm, cranial length more than 455 mm in AMNH 21608). The glenoid fossa is broad both in Triplopus and Forstercooperia, but the postglenoid process of the latter is more massive than that of the former. The postglenoid process is further separated posteriorly from the posttympanic process in Triplopus, forming a large external meatus, while in Forstercooperia the posttympanic process is ventrally expanded to be close to the postglenoid process. In addition, the posttympanic process and paroccipital process are slender and separated from each other in T. cubitalis, while these two processes are partially fused in Forstercooperia. The sagittal crest of T. cubitalis, as well as the weak parietal crest, is less developed than that of Forstercooperia, which is comparatively high and strong. The anterior dentition of T. cubitalis remains primitive, comprised of spatulate incisors and slender canines, which are different from the conical specialization of upper incisors and stubby canines of Forstercooperia. T. cubitalis shares several features with Forstercooperia in the upper cheek teeth, such as primitive upper premolars, distinct paracone and metacone ribs of upper premolars, and a elongated metastyle of M2. They are easily distinguished from each other by following aspects: (1) the metalophs (metaconules) of upper premolars are more developed in T. cubitalis than in Forstercooperia; (2) the protolophs of upper premolars are more curved to contact the metaconules in T. cubitalis, nearly forming a lingual loop, than the straight form of Forstercooperia; (3) the premolars of T. cubitalis are nonmolarized, but the P2s display the rudimentary distinction of protocone and hypocone and the p4s have entoconids in Forstercooperia; (3) the parastyle is relatively larger and located far from the paracone in the cheek teeth of T. cubitalis, while Forstercooperia has smaller parastyles on the upper cheek teeth, which are closely situated with the paracone. The molars of T. cubitalis usually are distinguished from derived hyracodontids by their lack of ornaments, such as crochet and antecrochet, whereas an antecrochet is present in the M1 of Forstercooperia. The M3 of T. cubitalis differs from that of Forstercooperia with a distinct metacone and metastyle. Despite the contemporaneity of T. cubitalis and Forstercooperia (both F. totadentata and F. ulanshirehensis), the craniodental difference between Triplopus and Forstercooperia reflects the distinct morphological gap between hyracodontids and forstercooperiines, suggesting that these two groups are unlikely to be closely related.

Hyracodon is a derived hyracodontid from North America (Radinsky, 1967) that also shares several primitive craniodental features with Forstercooperia, such as the large premaxilla with a distinct ascending branch, large facial exposure of the lacrimal, well-developed sagittal crest, and complete dentition. Compared with well-preserved crania of Hyracodon (H. nebraskensis) (AMNH 12296, 12460), Forstercooperia (length of cranium over 455 mm) is considerably larger than Hyracodon (length of cranium ∼290 mm). The nasal of Forstercooperia bears a thin profile in cross section, whereas Hyracodon shows the derived condition of reduced nasals with a triangular cross section that medially contact and curve slightly, forming a distinct groove along the suture. The facial region, relative to the skull length, is longer in Forstercooperia than in Hyracodon. The reason is that the maxilla of Forstercooperia is more anteroposteriorly extended than that of Hyracodon. Particularly, the anterior border of the orbit is situated almost at the middle point of the cranium (the boundary of M2/3) in Forstercooperia, while it is much more anteriorly located (the boundary of M1/2) in Hyracodon. The lacrimal tubercle is rather weak and the lacrimal fossa is more medially situated in Forstercooperia, while there is a well-developed lacrimal tubercle and two distinct lacrimal fossae on the posteroventral border of the orbit in Hyracodon. The ventral surface of the palatines in Forstercooperia is nearly flat (or slightly concave), and probably possesses a pair of incisive foramina (assumedly similar to Pappaceras). This is distinguished from that of Hyracodon, in which a single incisive foramen is present, and a distinctive crest rises posteriorly at the level of P2 along the suture of the palatines. This crest appears to be unique to Hyracodon among all early rhinocerotoid groups. The pterygoid fossa is rather weak in Forstercooperia, Pappaceras, and most primitive rhinocerotoids, but it is clearly well developed in Hyracodon. Similar to Triplopus, Hyracodon lacks the unique characteristics of the basicranium in Forstercooperia and Pappaceras. The postglenoid foramen is present in Hyracodon, but absent in Forstercooperia. The anterior surface of the postglenoid process lacks ridges in Hyracodon and other typical hyracodontids, and this is different from forstercooperiines, amynodontids, and rhinocerotids, in which it is divided into two surfaces by a median ridge (Wang et al., 2017). Correspondingly, the postcotyloid process of the mandibles is absent in hyracodontids, but present in other rhinocerotoids despite the variation of shape and orientation (Wang et al., 2017). In particular, the postcotyloid process is characteristic of rhinocerotids, but clearly weak in paraceratheriids and amynodontids. The short posttympanic process of Hyracodon has a sharp lateral margin, and is posteriorly separated from the paroccipital process by a distinctive groove without the posttympanic-paroccipital foramen in lateral view. This is not the case in Forstercooperia, for which the posttympanic process is ventrally thickened and partially fused with the paroccipital process and the posttympanic-paroccipital foramen is present. The mandibular symphysis of Hyracodon is transversely more constricted in width than that of Forstercooperia. The posterior border of the ascending ramus in Hyracodon is more anteriorly inclined than in Forstercooperia.

Forstercooperia is also clearly distinguishable from Hyracodon by a number of dental characteristics. Despite the similarity of their generally conical incisors, Forstercooperia has subequal and stubby incisors (I2 slightly larger than I1 and I3), while all incisors of Hyracodon are distinctively slender and conical (I1/i1 slightly larger than other incisors) accompanied by a reduced mandibular symphysis. The upper canine of Forstercooperia is much stubbier and larger than the incisors, similar to that of amynodontids and eggysodontids. This is different from the upper canine of Hyracodon, which has a slender, conical, and incisorlike outline. Forstercooperia possesses primitive upper premolars, of which only P2 has barely separate protocone and hypocone, whereas the upper premolars of Hyracodon show a higher molarization and the hypocone and metaloph are well developed. The ribs of the protoloph and metaloph on the buccal wall are distinct in Forstercooperia, but weak in Hyracodon. The antecrochets of M1–2 are weak in Forstercooperia, whereas they are well developed in Hyracodon. The metastyle of M2 is elongated and lingually extended in Forstercooperia, while relatively straight in Hyracodon. The M3 of Forstercooperia lacks the metastyle, unlike that of Hyracodon, in which there remains a short but evident metastyle on the posterior side of the ecto-metaloph. The derived features of the lower cheek teeth of Hyracodon are consistent with the upper. The p1 is reduced in Hyracodon, different from all forstercooperiines except for Pappaceras meiomenus. Also, the lower premolars of Hyracodon are distinctly molarized with well-developed entoconids and entolophids. In contrast to Hyracodon, all lower premolars of forstercooperiines remain primitive without the entolophid, and the entoconid in the lower premolars only appears in the p4 of F. ulanshirehensis.

Another subgroup of derived hyracodontids is represented by the Asian Ardynia (A. praecox and A. kazachstanensis) (Matthew and Granger, 1923; Beliajeva, 1952; Radinsky, 1967). Forstercooperia is similar to Ardynia (AMNH 26039 and AMNH 26183) in several features: premaxilla-nasal contact, slender zygomatic arch, and complete anterior dentition. Similar to other hyracodontids, Ardynia praecox differs from Forstercooperia in several aspects. Compared with conical and equal-sized upper incisors of Forstercooperia, Ardynia retains the primitive spatulate, loosely spaced upper incisors, of which I1 is larger than I2 and I3. Similar to other hyracodontids, the upper canine of Ardynia is reduced and smaller than the upper incisors, unlike the stubby canines of forstercooperiines. Ardynia also exhibits the reduction of premolars by the absence of P1/p1 and the presence of single-rooted p2s, whereas within forstercooperiines the reduction of P1/p1 is present only in P. meiomenus. On the other hand, the diastema between canines and cheek teeth is longer in Ardynia than that in Forstercooperia. The upper premolars of Ardynia are derived with distinct protolophs and metalophs, whereas Forstercooperia possesses weak metalophs or metaconules only in the upper premolars. Differing from Hyracodon, the M3 of Ardynia is more derived than Forstercooperia and early hyracodontids by the absence of metacone and metastyle, and the presence of the convex posterior wall of ecto-metaloph. The lower cheek teeth of Ardynia are distinguished from those of Forstercooperia and most other rhinocerotoids by a uniquely V-shaped trigonid, resulting from a short paralophid and considerably lingually directed protolophid. In addition, Ardynia bears a ridge on the ventral surface of the palatines, which is probably unique for derived hyracodontids (like Hyracodon). Although the ascending part of the premaxilla in Ardynia is relatively extensive, the narial notch is more posteriorly located at the level of P2, deeper than those of forstercooperiines. The paroccipital process of Ardynia is considerably more slender and narrower than those of forstercooperiines. The mandibular symphysis of Ardynia is relatively elongated and more anteriorly extended (rather than constricted as in forstercooperiines), and similar to those of most primitive tapiromorphs.

Rostriamynodon grangeri (AMNH 107635) is the earliest representative of amynodontids with an age of Arshantan (Wall and Manning, 1986). It is another rhinocerotoid genus contemporary with Forstercooperia (F. totadentata and F. ulanshirehensis). Rostriamynodon also shares a few primitive features with Forstercooperia, including premaxilla-nasal contact, nasallacrimal contact, a distinct preorbital fossa, and a slightly concave ventral surface of the palatines and large canines. However, Forstercooperia obviously differs from Rostriamynodon in numerous craniodental features. The skull of F. ulanshirehensis is smaller than that of Rostriamynodon; the latter is about 560 mm in length and presumably close to that of F. totadentata (Wall and Manning, 1986). The anterior tip of the nasal in Forstercooperia is slightly narrower, similar to that of Hyrachyus or Triplopus, whereas it is generally rounded in Rostriamynodon and close to other amynodontids (like Amynodon) and derived rhinocerotids. The preorbital fossa of Forstercooperia is slightly shallower than that of Rostriamynodon and later amynodontids. The lower incisors of Forstercooperia are larger and more conical than those of Rostriamynodon, which are primitively spatulate, tightly spaced between lower canines, and relatively small. The lower canine of Forstercooperia is stubby but not tusklike, whereas Rostriamynodon bears more tusklike lower canines with sharp mesial and distal ridges. There remains a complete premolar series in Forstercooperia (neither the reduction of P1/p1, nor is there the anteroposterior constriction of the premolar series), whereas Rostriamynodon shows the absence of the first premolar and shortened premolar series, which are characteristic of amynodontids (like Amynodon) and derived hyracondontids (like Triplopides and Ardynia). The diastema of Forstercooperia is slightly shorter than Rostriamynodon. In addition, the upper cheek teeth of Forstercooperia are less quadrate in outline than those of Rostriamynodon, despite the poor preservation of the upper molars of Rostriamynodon. Specifically, the M3 of Forstercooperia has a triangular outline, distinct from the quadrate form of Rostriamynodon, which is attributed to the completeness of the metacone and ectoloph of M3. Also, the upper premolars of Forstercooperia, especially P4, are slightly transversely narrower than those of Rostriamynodon, and this is reflected by the ratio of width to length of P4, which is less than 1.5 in Forstercooperia (∼1.3 for F. ulanshirehensis), but more than 1.5 in Rostriamynodon. Corresponding to the upper cheek teeth, the lower cheek teeth of Rostriamynodon also display characteristics of amynodontids; for instance, the metalophids of lower molars are considerably extended anterolaterally rather than laterally as in non-amynodontid rhinocerotoids. In addition, the mandibular angle of Forstercooperia is more rounded than that of Rostriamynodon.

The new occurrence of Forstercooperia is definitely significant for comprehensive comparative study of paraceratheriids (paraceratheriines and forstercooperiines) after Qiu and Wang's work in 2007, further narrowing the morphological gap between forstercooperiines and paraceratheriines. Juxia sharamurenensis (IVPP V2891) is widely regarded as the direct ancestor or sister taxon of later giant rhinos, and it occurs in the Sharamurun Formation and the “Ulan Gochu Formation” of the Shara Murun and Urtyn Obo areas, Nei Mongol, China, which is commonly regarded as stratigraphically younger than the Irdin Manha Formation (Chow and Chiu, 1964; Qiu and Wang, 2007). Qiu and Wang (2007) proposed that paraceratheriids possess an elongated posterior cranial portion, which is reflected by the percentage of the length from the postglenoid process to the condyle to the length from P1 (or P2) to the condyle. It is greater than or equal to 25% for Juxia and later paraceratheriids, which is partially attributed to the enlargement of the posttympanic process, paroccipital process, and occipital condyle. Forstercooperiines have a similar ratio to paraceratheriids, about 0.22 for P. meiomenus (IVPP V 20254) (Wang et al., 2016), and 0.26 for F. ulanshirehensis (AMNH 21608) (table 1). The nasalpremaxilla contact is primitively present in Forstercooperia, but already lost in J. sharamurenensis. The narial notch of Forstercooperia is shallow with its posterior margin at the upper canine level, whereas in J. sharamurenensis it is more posteriorly located above P2. The anterior border of the orbit is more posteriorly located above the anterior end of M3 in Forstercooperia than it is in J. sharamurenensis, on which it is situated above the anterior end of M2. Furthermore, the new specimens of Forstercooperia reveal that morphological gaps still remain between Forstercooperia and Juxia in the basicranium. Compared with other early rhinocerotoids, the posttympanic process of Forstercooperia is clearly anteroposteriorly thickened, similar but still rudimentary compared to that of Juxia. The posttympanic process and paroccipital process have different heights, relatively slender outlines (especially the paroccipital process), and are closely situated but incompletely fused in F. ulanshirehensis, whereas they are roughly equal in height, nearly vertically situated, and completely fused with a transversely blunt outline in J. sharamurenensis. The ratio of the height of the occipital condyle to its width is about 1.2 in both F. ulanshirehensis and J. sharamurenensis (table 1), but the size of the occipital condyle in the former is smaller than that in the former (height:width: 42.96:35.95 mm vs. 60:50 mm). Otherwise, Juxia is the earliest record of distinct enlargement of I1/i1 and procumbent i1, which are regarded as significant characteristics of Paraceratheriinae (Lucas and Sobus, 1989; Qiu and Wang, 1999, 2007). Forstercooperia is easily distinguished from Juxia by the absence of unique specialization of I1/i1. Additionally, the stubby canines in Forstercooperia are also different from the incisorlike canines of Juxia. The earlier stratigraphic occurrence of Forstercooperia relative to Juxia seems consistent with the morphological distinction of upper premolars between the two genera. Specifically, Forstercooperia possesses more primitive premolars than Juxia, showing that the separation of protocone and hypocone is present only in P2 of Forstercooperia, but that separation occurs in P2–3 of J. shramurenensis along with the presence of well-developed metaloph of P2. Meanwhile, the protolophs of P3–4 are transversely straight in Forstercooperia, but more posteriorly extended at the lingual side in J. shramurenensis. Similarly, among lower premolars, the entoconid is present in p3–4 of J. shramurenensis, whereas only p4 has a weak entoconid in F. ulanshirehensis.

Teletaceras is considered the most primitive rhinocerotid (Hanson, 1989; Prothero, 2005). It displays craniodental features that are generally shared by primitive tapiromorphs (including forstercooperiines), such as a small to medium size, dolichocephalic structure, complete dentition, and primitive premolars. In addition, Forstercooperia has derived features that are characteristic of rhinocerotids, such as a medial ridge on the ventral surface of the glenoid fossa, a weak postcotyloid process, and a distinct median crest on the ventral surface of basioccipital. However, Forstercooperia is clearly distinguished from Teletaceras in the following aspects: (1) the nasal is transversely wider and relatively thinner than that of Teletaceras, which has a triangular cross section and tapers gently; (2) the narial notch is above the upper canine and shallower than that of Teletaceras, which is above the postcanine diastema; (3) the frontal is much wider than in Teletaceras; (4) the facial region is more elongated with the anterior border of orbital above the boundary of M2–3, rather than above the boundary of M1–2 as in Teletaceras; (5) the sagittal crest is relatively high and distinct, more developed than that of Teletaceras, in which the supratemporal crests converge at the midline to form a low and narrow sagittal crest; (6) the postglenoid foramen is absent in Forstercooperia, but present in Teletaceras; and (7) the posttympanic process is thickened and partially fused with and the paroccipital in Forstercooperia, while the two processes are considerably slender and completely separated in Teletaceras. Meanwhile, the differences in dental characters between Forstercooperia and Teletaceras demonstrate that they are not closely related. The most diagnostic character of Teletaceras lies in the anterior dentition. Despite the incompleteness of the anterior dentition, Teletaceras (UCMP 129009) differs from Forstercooperia with distinctively enlarged and tusklike I1 and i2, which are the earliest records of the diagnostic rhinocerotid features. The I1 of Teletaceras has a single wear facet compatible with the i2, whereas the I1 in forstercooperiines has two wear facets compatible with lower incisors (i1 and i2). According to the original reconstruction (Hanson, 1989: fig. 20), the I1 and i2 are probably larger than other incisors, as in other rhinocerotids. This is distinct from the conical and subequal incisors of Forstercooperia. The canines are reduced in size and conical in Teletaceras, but rather stubby and large in Forstercooperia. The protolophs of upper premolars that are posteriorly extended at lingual side in Teletaceras differ from those of Forstercooperia. The metaconules of P3–4 are extended more posterolingually in Teletaceras, whereas they are relatively transversely straight in Forstercooperia. The M3 of Teletaceras still retains a small trace of metacone and metastyle, but Forstercooperia lacks the metacone and metastyle of M3, possessing a distinctly triangular outline.

Eggysodontids are one of the most poorly known rhinocerotoid groups and include four genera to date: Proeggysodon Bai and Wang, 2012; Guangnanodon Wang et al., 2013; Eggysodon Roman, 1911; and Allacerops Wood, 1932. The former two are represented by fragmentary mandibles from the Upper Eocene of northern China. The earliest-known eggysodontid, Proeggysodon, shows some similarity to Pappaceras confluens in its smaller size, short diastema, and large lower canines (Bai and Wang, 2012). Compared with Forstercooperia, the Eocene genus Proeggysodon is distinguished by its smaller size, broad mandibular symphysis (rather than transversely narrow or constricted as in forstercooperiines), primitively spatulate lower incisors, and reduction of lower incisor number. Guangnanodon differs from Forstercooperia in the presence of entoconids and weak entolophids of p3–4 and the short paralophids of the lower molars (Wang et al., 2013). The Oligocene eggysodontids Eggysodon and Allacerops are, despite their relatively large size (length of m1–3 ∼80–96.7 mm), still smaller than those of Forstercooperia (Radinsky, 1967; Reshetov et al., 1993; Qiu and Wang, 1999; Uhlig, 1999; Bai and Wang, 2012; Wang et al., 2013). Although the Oligocene eggysodontids show some variation in the number of incisors (2 or 3) and premolars (3 or 4), they are distinct from Forstercooperia, with more molarized upper and lower premolars, of which P3–4 have fully separated hypocones and strong metalophs, and their p3–4s have well-developed entoconids and entolophids (Reshetov et al., 1993; Qiu and Wang, 1999; Uhlig, 1999). Based on the recently reported fragmentary cranium of Eggysodon, Antoine et al. (2011) have listed a few apomorphic traits of eggysodontids, for instance, a short nasal and weak posttympanic and paroccipital processes, which are considerably distinct from those of forstercooperiines and roughly similar to rhinocerotids. Compared with Forstercooperia, the mandibular symphysis of Eggysodon is more upright and the posterior margin of the ascending ramus is distinctively inclined forward. Although eggysodontids are not comprehensively known from the current collection, the evidence is indicative of a morphological gap between forstercooperiines and eggysodontids.

Holbrook and Lucas (1997) established Uintaceras to encompass most of the North American “forstercooperiines” that were formerly assigned to Forstercooperia by Radinsky (1967) and Lucas et al. (1981) and proposed a novel affinity between Uintaceras and rhinocerotids. This genus is somewhat unique within rhinocerotoids mainly because of its anterior dentition. In the phylogenetic results suggested by Holbrook and Lucas (1997), the Uintaceras-rhinocerotid clade is moderately supported by the synapomorphy of the buccolingually compressed upper incisors. This feature seems uniquely distinctive across all rhinocerotoids, given the morphological gap of anterior dentition between Uintaceras and Teletaceras, the earliest representative of rhinocerotids. The similarities of postcranial features between Uintaceras and amynodontids and rhinocerotids were also already discussed (Holbrook and Lucas, 1997), including the rhomboidal radial facet of the scaphoid, the distal process of the posterior ectocuneiform, the prominent, laterally hooking deltopectoral crest, and robust, noncursorial limbs; these are as significant as the features of the anterior dentition, although this resemblance was regarded as a scaling deviation in their paper. Furthermore, the similarities among Uintaceras, forstercooperiines (Pappaceras), and amynodontids have been broadly discussed (Wang et al., 2016), supporting a close affinity among these three groups. More importantly, the new specimens of F. ulanshirehensis reinforce the similarity between forstercooperiines and Uintaceras, demonstrating the close relationship between Uintaceras, amynodontids, and paraceratheriids. The postglenoid process of Uintaceras is massive and transversely broad, similar to forstercooperiines. The posttympanic and paroccipital processes in Uintaceras, which are relatively enlarged and extended ventrally rather than posteroventrally, are most comparable to those of paraceratheriids within early rhinocerotoids. The diagnostic posttympanic-paroccipital foramen is present both in Uintaceras and paraceratheriids. The resemblance of dental features between Uintaceras and forstercooperiines has been discussed by several authors (Radinsky, 1967; Lucas et al., 1981; Holbrook and Lucas, 1997), and the cheek teeth of Uintaceras are most comparable to those of forstercooperiines with the presence of strong ribs on upper premolars, protolophs of upper premolars posteriorly extended (similar to Pappaceras confluens), antecrochet of M1–2, and relatively distinct metacone and short metastyle of M3 (similar to Pappaceras minuta and Pappaceras meiomenus). On the other hand, Uintaceras can be distinguished from forstercooperiines by its narial incision deeper above P1 and the anterior border of orbital located above M1–2 (Holbrook and Lucas, 1997). Consequently, we conclude that Uintaceras has uniquely mosaic features of early rhinocerotoids, suggesting that this former “forstercooperiine” genus should be closer to forstercooperiines rather than to rhinocerotids.

Phylogenetic Analysis

Only one most-parsimonious tree is returned from the phylogenetic analysis as below (fig. 10). Similar to the recent study by Wang et al. (2016), the primary groups of early rhinocerotoids (Hyracodontidae, Amynodontidae, Paraceratheriidae, Eggysodontidae, and Rhinocerotidae) are monophyletic. The monophyly of hyracodontids is well supported from the results of the phylogenetic analysis, showing relatively high Bremer support values for the nodes within the hyracodontid clade, despite the exclusion of some taxa con ventionally included in the hyracodontids (e.g., Radinsky, 1967; Prothero et al., 1986; Wang et al., 2016). The amynodontid clade is the most stable group in the phylogenetic analysis with a high Bremer support value. This clade is closely related to the paraceratheriids and Uintaceras, but the monophyly of these three clades is moderately supported due to character conflicts. For instance, the dentition of amynodontids exhibit numerous unique features distinct from those of other rhinocerotoids. After adding F. ulanshirehensis and F. mazhuangensis to the data matrix, the phylogeny of the monophyletic forstercooperiine clade has better resolution than the previous study (e.g., Wang et al., 2016), demonstrating the monophyly of Pappaceras and Forstercooperia in a phylogenetic context for the first time. This is in contrast to the previous argument that all species of forstercooperiines were congeneric (Forstercooperia) (e.g., Radinsky, 1967; Lucas et al., 1981). Contrary to the latest study (Wang et al., 2016), the interrelationship of three species of Pappaceras is well resolved, showing that P. confluens and P. minuta are closer to each other than to the newly erected species P. meiomenus. However, the new species F. ulanshirehensis is clustered with F. totadentata and F. mazhuangensis, and the interrelationships of these three species of Forstercooperia are unresolved. This result is not unexpected because F. totadentata and F. mazhuangensis are represented only by fragmentary crania and incomplete dentition, and as such the character matrix lacks many character-state codings for the two species. Furthermore, the monophyletic clade comprising paraceratheriids and Uintaceras is moderately supported, which is probably attributed to conflicting phylogenetic signals from a sizable part of autapomorphies of Uintaceras (Holbrook and Lucas, 1997; Wang et al., 2016). The enigmatic eggysodontid clade, which comprises three genera represented by incomplete fossils, is placed as the sister group of rhinocerotids. This novel result is in contrast to the previous views regarding either hyracodontid or paracerotheriid affinity for eggysodontids (Radinsky, 1967; Reshetov et al., 1993; Qiu and Wang, 1999; Uhlig, 1999; Bai and Wang, 2012; Wang et al., 2013). However, this question is in need of further investigation based on more morphological data added to the present character matrix, in particular, from more complete Eocene specimens.

FIG. 10.

Phylogenetic relationships of early rhinocerotoids based on the single most-parsimonious tree returned from phylogenetic analyses. The number by each node denotes the Bremer Support value. Tree length = 433, CI = 0.536, RI = 0.714. The forstercooperiine clade is marked by the pink rectangle.

Biostratigraphic Implication

Generally, stratigraphic correlations are the basis for the correlative relationships of biostratigraphic units or zones within particular areas. The Paleogene stratigraphic framework of the Erlian Basin has already been established based on numerous investigations since the 1890s as summarized by Wang et al. (2012) and Meng et al. (2007), such that four faunas (Arshantan, Irdinmahan, Sharamurunian, Ulangochuan) from this area have laid the ground for the Paleogene Asian Land Mammal Ages in the geologic time scale (Gradstein et al., 2012). But there still remains some confusion regarding fossil locality names and lithostratigraphic units, some of which have been discussed and corrected in a recent study after detailed field investigation over the last decade (Wang et al., 2012). The correction of fossil locality names and lithostratigraphic correlations inevitably affects biostratigraphic correlations. In terms of the locality names involved in the present study, the localities from which the new specimens (AMNH 21608 and 22101) were collected were originally described as “8 miles north of Tukhum Lamasery” and “4 miles north of Tukhum Lamasery,” respectively (Granger, 1925, 1928). Based on the detailed fieldwork and comprehensive comparisons by an international team composed of staff from IVPP, AMNH, and CMNH (Meng et al., 2007; Wang et al., 2010, 2012), the two localities should be regarded as the Wulanhuxiu site and Wulantaolegai site of the western part of Erlian Basin respectively. With regard to the lithostratigraphic units, the mentioned specimens (AMNH 21608 and 22101) were unearthed from the Ulan Shireh Formation, although the correlation between the sediments and the eastern part of the Erlian Basin is still unclear. Conventionally, the Ulan Shireh Formation was correlated to the Irdin Manha Formation (Granger and Gregory, 1934; Granger, 1938; Ye, 1983; Wang et al., 2012), but recent investigation demonstrates that the mammalian fauna from the Ulan Shireh Formation is probably composed of both Arshantan and Irdinmanhan taxa (Wang et al., 2012). The latest study of the mammal fauna from the Ulan Shireh Formation at the Wulanhuxiu site provided a detailed description of the recently discovered small mammals, summarized the mammal records from this area, and discussed faunal turnover in the Chinese Eocene record (Li et al., 2016). That work suggests that the faunal assemblage is much more complex than previously thought, proposing that the lower fossiliferous bed (horizon 4) is indicative of an Irdinmanhan age chiefly with a typical assemblage of perissodactyls (Li et al., 2016: table 1), while the upper horizon (horizon 2) is probably Sharamurunian (or post-Irdinmanhan) in age with abundant remains of Sharamurunian Gobiolagus (Meng et al., 2005; Li et al., 2016). Combining the field notes recorded by Granger and recent investigations, the lower part of the Ulan Shireh Formation (e.g., horizon 4 at the Wulanhuxiu site, see Li et al., 2016) yields a considerable number of fossils (mainly dominated by perissodactyls and Glires), whereas the upper part, containing less abundant fossil material, may be equivalent to the Tukhum Formation, which is another lithostratigraphic unit in the western part of the Erlian Basin (Wang et al., 2012). The stratigraphic information for the two specimens in the collection database is extremely limited, and it is impossible to ensure which lithostratigraphic bed of the Ulan Shireh Formation contained them (AMNH 21608 and AMNH 22101). It is highly possible that the two specimens were unearthed from the lower part of the Ulan Shireh Formation, judging from the preservation condition (H.-B.W., personal obs.) and the dominant component of perissodactyls in this faunal assemblage. If this hypothesis is correct, their occurrence would lend support to the notion that the lower part of the Ulan Shireh Formation is equivalent to the Irdin Manha Formation in the eastern part of the Erlian Basin, where the other three specimens of F. unlanshirehensis (IVPP V20154, V20155, V20156) were recovered. In that case, forstercooperiines are Arshantan-Irdinmanhan in age, with the smaller, Arshantan-aged Pappaceras having a primitive dentition and the larger, Irdinmanhan-aged Forstercooperia having a derived craniodental morphology.