A new specimen of Paleoparadoxia found from a marine lower Miocene deposit in the Chikubetsu area, Hokkaido, Japan, is described. The material consists of a distal part of the scapula, proximal end of the humerus from the right side, as well as a fragmentary rib, preserved in a float of calcareous fine sandstone. The specimen is referred to the order Desmostylia and subsequently to the genus Paleoparadoxia sp. The well preserved shoulder girdle of this specimen provides the first detailed morphology of this anatomical region in Paleoparadoxia. We compared the specimen with a wide range of desmostylid samples to reveal new diagnostic characters for the genus, such as the greater tubercle extending toward the proximal side above the head and the distinct lesser tubercle located on the medial side, projected medially. The lower Miocene Sankebetsu Formation outcrops in the area where the float was found, and the lithology and associated fossil fauna of the float indicate that it was derived from there. The fossil pollen assemblages and molluscs found in the formation indicate that it was deposited under cool to temperate conditions. The new specimen thus suggests that Paleoparadoxia had inhabited a cool environment in addition to warm areas, as suggested in previous studies. The published age estimate of the formation places the present specimen between 23.8±1.5 and 20.6±1.0 Ma. This represents the oldest record of the genus Paleoparadoxia, exceeding the previous record of ca. 19 Ma for a specimen found from the Chichibu Basin, and nearly matches the oldest record of Paleoparadoxiinae in the Northwestern Pacific. This indicates that the ages of the oldest occurrences of basal and derived paleoparadoxiines overlapped (i.e., Archaeoparadoxia from the Northeastern Pacific Region and Paleoparadoxia from the Northwestern Pacific, respectively). It is also likely that the geographic range of Paleoparadoxiinae had already expanded from the Northeast Pacific to the Northwest Pacific in the early stage of their evolution.
Introduction
Paleoparadoxia is a genus of extinct marine mammals belonging to the order Desmostylia, one of the constituents of the Tethyteria, which also includes the extant orders Proboscidea and Sirenia (Domning et al., 1986; Thewissen and Domning, 1992; Inuzuka, 2000a). Desmostylian fossils are found from strata ranging from the lowest Oligocene (Domning, 1996; Barnes, 2013) to lowest lower Miocene in the coastal areas of the North Pacific Ocean (e.g. Inuzuka et al., 1994; Domning, 1996; Inuzuka, 2000a, b; Beatty, 2009; Barnes, 2013). Previously, the order Desmostylia had been divided into two families, Desmostylidae and Paleoparadoxiidae (Reinhart, 1959; Domning et al., 1986; Clark, 1991; Inuzuka et al., 1994; Inuzuka, 2000a, b, 2005). Beatty (2009) excluded Behemotops from the Paleoparadoxiidae based on cladistic analysis. Paleoparadoxiidae has been considered to include only one genus, Paleoparadoxia (e.g. Inuzuka et al., 1994; Domning, 1996; Inuzuka, 2005). Barnes (2013) revised this view by dividing the genus Paleoparadoxia into three genera based on the study of new specimens: Archaeoparadoxia Barnes, 2013; Paleoparadoxia Reinhart, 1959; and Neoparadoxia, Barnes, 2013; and erected a subfamily Paleoparadoxiinae comprising these three genera. He presented Archaeoparadoxia as a basal paleoparadoxiine, followed by Paleoparadoxia, and subsequently Neoparadoxia (Figure 1; Barnes, 2013). We follow the taxonomic scheme proposed by Barnes (2013) in this study.
In 2010, one of us (S. K.) found a float containing a superbly preserved mammalian shoulder girdle from the lower Miocene marine deposit in the Chikubetsu area of Hokkaido, Japan (Figures 2 and 3). The shoulder girdle was identified as belonging to the genus Paleoparadoxia, as discussed below. Although a large number of paleoparadoxiine remains have been reported (e.g. Reinhert, 1959; Inuzuka et al., 1994; Domning, 1996; Inuzuka, 2000a; Inuzuka, 2005), well preserved shoulder girdles are rare, probably because forelimbs of tetrapods tend to disconnect from the trunk early in the decomposition process of a postmortem body (Lyman, 1994; Inuzuka, 2005). Although Shikama (1966) and Saegusa (2002) reported paleoparadoxiine humeri (NMNS PV-5601 and SMNH VeF-61, respectively), their proximal portions were poorly preserved. Among Paleoparadoxiinae, only the forelimb of Neoparadoxia (UCMP 81302, Inuzuka, 2005; LACM 150150, Barnes, 2013) is well preserved. Thus, the morphology of the paleoparadoxiine forelimb, especially the proximal humerus, has been insufficiently described thus far. Therefore, comparing the new specimen provides valuable anatomical and taxonomical information on the paleoparadoxiine shoulder girdle.
Paleoparadoxia has been reported from the lower to middle Miocene in the Northwest Pacific (Figure 4; Ogasawara, 2000; Inuzuka, 2005) and other areas around the North Pacific Rim, including Baja California, Mexico (Barnes and Aranda-Manteca, 1997) and California, USA (VanderHoof, 1937; Reinhart, 1959; Mitchell, 1963; Mitchell and Repenning, 1963). The previous record of the oldest Paleoparadoxia came from the lower Miocene (19 Ma) in the Chichibu Basin, Saitama, Japan (Ogasawara, 2000). The age of the present specimen is considered to be the earliest Miocene (23.8–20.6 Ma, see below), significantly older than the oldest record of Paleoparadoxia ever reported. Some previous studies emphasized that Paleoparadoxia originated in the early Miocene in North America and then spread throughout the North Pacific Rim, including Japan (Inuzuka, 2005, 2013), based on the earliest and basal paleoparadoxiine Archaeoparadoxia weltoni found from the lowest Miocene in North America (Clark, 1991). This specimen extends the earliest record of Paleoparadoxia in the Western Pacific to a time comparable to that of Paleoparadoxiinae as a whole (represented by Archaeoparadoxia weltoni) in the Eastern Pacific. The present specimen provides new information for interpreting the early evolution and distribution of Paleoparadoxiinae.
In addition to its morphological and evolutionary implications, this specimen contributes to our knowledge of the spatial distribution of the genus Paleoparadoxia. The locality of the specimen is the northernmost among those reported for the genus. Previous reports show that the occurrences of Paleoparadoxia range from 35°5′N (Shimane, southwestern Japan) to 43°08′N (Akan, southern Hokkaido, Japan; Ohkubo et al., 1980; Taguchi, 1984; Kimura et al., 1998; Inuzuka, 2000; Ogasawara, 2000). This range overlaps, but is limited to the southern portion of that of Desmostylus (35°5′N, Shimane, Japan to 55′N, Sakhalin, Russia). Chinzei (1984) and Ogasawara (2000) suggested that the habitats of Paleoparadoxia were subtropical to warm waters in the temperate zone based on such occurrence patterns of desmostylians and the information provided by molluscan and pollen fossils found with Paleoparadoxia. The Chikubetsu area is located farther north than the previously reported geographical range of Paleoparadoxia; thus, the new specimen contributes new information on the paleogeography of Desmostylia.
Materials and methods
A partial but well-preserved shoulder girdle of a desmostylian and other associated bones probably belonging to a single individual (Figure 5; UMUT CV31059, proximal right humerus; UMUT CV31060, distal end of right scapula; UMUT CV31061, proximal right rib) were found from the lower Miocene marine deposit in the Chikubetsu area, Hokkaido, Japan. To infer their taxonomic identity, the following reference specimens were studied:
(i) NMNS PV-5601, the neotype of Paleoparadoxia tabatai (Tokunaga, 1939; Shikama, 1966), a nearly complete skeleton including an incomplete left humerus from the lower Miocene of the Mizunami Group, Gifu, Japan.
(ii) SMNH VeF-61, a partial skeleton of Paleoparadoxia sp., including a nearly complete left humerus from the lower Miocene in the Chichibu Basin, Saitama, Japan (Saegusa, 2002).
(iii) UHR 18466, a nearly complete skeleton of Desmostylus hesperus (Marsh, 1888), including a nearly complete left humerus from the middle Miocene Uchiboro coal-bearing Formation in Sakhalin, Russia. This is the type specimen for D. mirabillis (Nagao, 1935) that was later synonymized with D. hesperus (Inuzuka et al., 1994) and redescribed by Inuzuka (1982).
(iv) GSJ-F7743, a nearly complete skeleton of Desmostylus hesperus, including left and right humeri from the middle Miocene Tachikaraushinai Formation in Japan, described by Inuzuka (2009).
(v) AMP 21, a nearly complete skeleton of Ashoroa laticosta (Inuzuka, 2000b), including nearly complete right and left humeri from the Oligocene Morawan Formation in Hokkaido, Japan, described by Inuzuka (2000b, 2011). This specimen is the species holotype.
(vi) AMP 22, a nearly complete skeleton of Behemotops katsuiei (Inuzuka, 2000b; Inuzuka, 2006), including ribs from the upper Oligocene Morawan Formation in Hokkaido, Japan. This specimen is designated as the species holotype.
(vii) AMP AK0011, a nearly complete left humerus of Paleoparadoxia sp. from the middle Miocene Tonokita Formation in Hokkaido, Japan.
(viii) LACM 150150, a nearly complete skeleton of Neoparadoxia cecilialina (Barnes, 2013), including nearly complete right and left humeri from the lower upper Miocene Monterey Formation in California, USA. This specimen was designated as the species holotype.
(ix) HMG-343, a nearly complete skeleton of Desmostylus “japonicus” (Kimura and Akamatsu, 1984; Kimura, 1985), including ribs from the lower to middle Miocene Takinoue Formation in Hokkaido, Japan.
Figure 2.
Map showing the locality of UMUT CV31059-31061 and geology of the Chikubetsu area in northwestern Hokkaido, Japan.

Figure 3.
Stratigraphic column of the Chikubetsu area showing the inferred approximate stratigraphic position of UMUT CV31059-31061 (marked with a star; after Kurita et al., 1992; Noda, 1992; Kafanov et al., 2005).

Figure 4.
Localities of Paleoparadoxia specimens from Japan showing UMUT CV31059-31061 from the lower Miocene (star); and other Paleoparadoxia from the lower Miocene (circles), middle Miocene (16.3–14.2 Ma, squares), and middle Miocene (<14.2 Ma triangles). Modified after Kurita et al. (1992) and Kafanov et al. (2005).

Linear measurements on specimens were made following Inuzuka (2005, 2009) with a vernier caliper with a precision of 0.01 mm (Table 1). The anatomical terminology follows Kato and Yamauchi (1995). To infer the environmental conditions of habitats of Paleoparadoxia, we compiled data on molluscan fossils associated with specimens of this genus in the Northwest Pacific, including those associated with UMUT CV310596 in the float.
Institutional Abbreviations—AMP, Ashoro Museum of Paleontology, Hokkaido, Japan; AOPM, Aomori Prefectural Museum, Aomori, Japan; APM, Akita Prefectural Museum, Akita, Japan; FM, Fukushima Prefectural Museum, Fukushima, Japan; FMI, The Folk Museum of Itsukaichi, Tokyo, Japan; GMNH, Gunma Museum of Natural History, Gunma, Japan; GSJ, Geological Survey of Japan, Ibaraki, Japan; HMG, Hobetsu Museum, Hokkaido, Japan; IPMM, Iwate Prefectural Museum, Iwate, Japan; KS, Kimachi Stone Museum, Shimane, Japan; LACM, Los Angeles County Museum, California, USA; MFM, Mizunami Fossil Museum, Gifu, Japan; NID and NOD, NANAO Science Museum of Children, Ishikawa, Japan; NMNS and NMST, National Museum of Nature and Science, Tokyo, Japan; SMNH, Saitama Museum of Natural History, Saitama, Japan; UCMP, University of California Museum of Paleontology, California, USA; UMUT, The University Museum, The University of Tokyo, Tokyo, Japan; UHR, The Hokkaido University Museum, Hokkaido, Japan.
Table 1.
Measurements (in mm) on the humerus, scapula, and rib of three Paleoparadoxia specimens.

Geological setting and age
The specimen was found in a float on the bank of the Chikubetsu-gawa River in the Chikubetsu area (Figure 2; 44°23′N and 141°52′E). In the upstream region of Chikubetsu-gawa River, a large anticlinal structure called the Chikubetsu Anticline is present (e.g., Kurita et al., 1992; Moriya and Hirano, 2004; Okamoto et al., 2003). The Mesozoic Yezo Group and Paleogene and Neogene deposits crop out in this region (Yamaguchi and Matsuno, 1963; Noda, 1992; Kurita et al., 1992; Kurita and Obuse, 1994). In the western wing of the Chikubetsu Anticline, the Paleogene deposit is divided into the Haboro and Sakasagawa formations, and the Neogene deposit consists of the Sankebetsu and Chikubetsu formations (Figure 3; Kurita et al., 1992; Noda, 1992; Kafanov et al., 2005).
The float was found in an area where the Sankebetsu Formation crops out (Figure 2). The Sankebetsu Formation consists of massive, tuffaceous, fine- to medium-grained gray sandstone. Alternating beds of sandstone and siltstone are also common and marine molluscan fossils occur abundantly in the sandstone (Noda, 1992). The lithology of the float is grayish, massive fine sandstone containing diatoms, molluscan shells, and wood fragments. Mulluscan taxa found from the float are consistent with those listed in Noda (1992) for the Sankebetsu Formation, and the lithology of the float closely matches that of the lower Miocene Sankebetsu Formation too. It is highly probable that the float was derived from the Sankebetsu Formation outcrop where the float was found (Figures 2 and 3). Depositional environments of the Sankebetsu Formation were shoreface to outershelf (Noda, 1992; Kurita et al., 1992). The molluscan fossils from the Sankebetsu Formation suggest that the climate was cool and temperate (Kurita et al., 1992).
The age of the Sankebetsu Formation was estimated based on fission-track (FT) dating, marine molluscs, and diatom fossils (Kurita et al., 1992; Noda, 1992; Kafanov et al., 2005). FT ages for the upper part and base of the formation were reported as 20.6±1.0 Ma and 23.8±1.5 Ma, respectively (Kurita et al., 1992). The molluscan fauna (e.g. Mytilus tichanovitchi, Macoma osakaensis, Megayoldia (Hataiyoldia) tokunagai, and Neilonella (Borissia) sakhalinensis) indicates that the formation was deposited in the early Miocene (Kafanov et al., 2005). The diatoms indicate ages from 24.0 Ma to 20.3 Ma (Thalassiosira praefraga zone; Hata et al., 1988; Kurita et al., 1992), consistent with the FT ages. The locality of the present specimen is more upstream than the sampling point for the FT analysis. The float's age is likely older than the FT ages, as the ages of strata become older upstream in this area. In conclusion, the possible age of the present specimen can be constrained between 25.3 Ma (oldest estimate based on FT, including the possible error of 20.6±1.0) and 20.3 Ma (youngest estimate based on diatoms).
Systematic paleontology
Order Desmostylia Reinhart, 1953
Family Paleoparadoxiidae Reinhart, 1959
Subfamily Paleoparadoxiinae (Reinhart, 1959) Barnes, 2013
Genus
Paleoparadoxia
Reinhart, 1959
Type species.—Paleoparadoxia tabatai (Tokunaga, 1939).
Emended diagnosis for the genus (humeral characters only).—The greater tubercle extending toward the proximal side above the head, a large greater tubercle, a distinct, medially projected lesser tubercle located on the medial side, the intertubercular groove located on the medial side, a shallow and narrow intertubercular groove, an oval head that is slightly convex at the distal end, and absence of the deltoid tuberosity.
Paleoparadoxia
sp.
Figure 5 A–C
Materials.—UMUT CV31059 (right humerus, Figure 5A), UMUT CV31060 (scapula, Figure 5B), and UMUT CV31061 (rib, Figure 5C). UMUT CV31059 and CV31060 were found articulated in situ, confirming that these bones definitely belong to the same individual.
Locality.—The riverbank where the Chikubetsu-gawa and Panke-zawa rivers meet in the Chikubetsu area, Hokkaido, Japan (Figure 2), where the lower Miocene Sankebetsu Formation outcrops (Figure 3; 44°23′N, 141°52′E).
Description.—The present specimen, although given three separate catalog numbers, is considered as a single individual because all bones were found from a single float and the humerus and scapula were articulated in situ.
Right humerus (UMUT CV31059).—The proximal part is well preserved, but most of the distal part is missing. The cranial side of UMUT CV31059 is partly compressed, whereas the medial side is slightly damaged. The epiphyseal line is obliterated. The head of the humerus is large as in Desmostylia in general, and is oval and slightly convex distally. The shaft of the humerus is thick and nearly straight. The greater tubercle is large as in Desmostylia in general, and is higher than the head of the humerus, located on the cranial side. The lesser tubercle is slightly higher than the middle of the head and situated anteromedial to the head, creating a short ridge extending in an anterolateral to posteromedial direction. Additionally, the head of the lesser tubercle is round. As observed in Desmostylia in general, the intertubercle groove separating the greater and lesser tubercles is shallow and narrow. The tricipital line is sharply defined. The angle of the neck shaft is 122°.
Right scapula (UMUT CV31060).—Although only the partial distal end is preserved, this preserved part is in good condition. The coracoid process is fully ossified and located on the medial side of the scapula. The glenoid cavity is oval and almost circular, and is depressed at the center. The glenoid cavity is relatively deep for Desmostylia because its margin is more ossified in this specimen than in others.
Rib (UMUT CV31061).—The head and tubercle of a rib are partially preserved. The detailed structure is unclear, and the distal part is largely absent. The rib head is larger than the tubercle and the shaft of the rib is twisted. The cross-section shows that the shaft is anteroposteriorly flat.
Identification
The specimen is referred to Desmostylia based on diagnostic characteristics in the humerus such as a straight humeral body and an unexpanded deltoid ridge (Shikama, 1966). Moreover, it can be referred to the genus Paleoparadoxia based on the following putatively diagnostic characteristics proposed in previous studies: (1) large greater tubercle of the humerus; and (2) slightly inwardly curved humeral crest. Saegusa (2002) recognized a large greater tubercle (69 mm in SMNH VeF-61) as a characteristic of Paleoparadoxia observed on the humerus. Inuzuka (2005) showed that the inwardly curved humeral crest is characteristic of “Paleoparadoxia media” (SMNH VeF-61 and NMNS PV-5601).
UMUT CV31059 differs from the humerus of Desmostylus hesperus (UHR 18466 and GSJ-F7743) in lacking the following features: intertubercle groove placed just behind the humeral head, nearly even heights of the greater tubercle and humeral head, and a long ridge running in an anterolateral to posteromedial direction. UMUT CV31059 can be also distinguished from Ashoroa laticosta because the former lacks the following characters observed in A. laticosta (AMP 21): slightly swelled lesser tubercle on the mid-medial side, sharp deltoid crest, well developed deltoid tuberosity, and undeveloped line of the triceps muscle. While Neoparadoxia ceciliaena (LACM 150150) has a relatively large deltopectoral crest and a sharp deltoid ridge, these features of UMUT CV31059 are smaller and weaker than those of N. ceciliaena. Thus, the new specimens also are distinguishable from N. ceciliaena based on these features.
Comparison and diagnostic features of Paleoparadoxia
Right humerus (UMUT CV31059).—The height of the greater tubercle of UMUT CV31059 is comparable to that of Paleoparadoxia tabatai (NMNS PV-5601) and Ashoroa laticosta (AMP 21), and relatively larger than that of Desmostylus hesperus (juvenile, GSJ-F7743, and adult, UHR 18466). The greater tubercle in UMUT CV31059, P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61), and A. laticosta (AMP 21) extends dorsally. Because the proximal part is poorly preserved, the greater tubercle of Paleoparadoxia sp. (AMP AK0011) is unrecognizable.
The morphological features of the lesser tubercle cannot be compared with those of other desmostylians because of poor preservation (NMNS PV-5601, SMNH VeF-61, and AMP AK0011) or incomplete ossification (juvenile, GSJ-7743). The position of the lesser tubercle of UMUT CV31059 is on the medial side and almost the same as its position in P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61), and A. laticosta (AMP 21). The position of D. hesperus (UHR 18466) is more cranial than in other specimens.
The depth of the intertubercular groove is typically shallow in desmostylians. The intertubercular groove of UMUT CV31059 is situated on the medial side and that of P. tabatai (NMNS PV-5601) and Paleoparadoxia sp. (SMNH VeF-61 and AMP AK0011) are in a similar position. Unlike UMUT CV31059, the intertubercular groove of D. hesperus (adult: UHR 18466) and A. laticosta (AMP 21) develops cranially. In D. hesperus (juvenile, GSJ-7743), it cannot be observed because the lesser tubercle was unfused and lacking in this specimen.
The head of the humerus is oval and slightly convex on the distal side in UMUT CV31059, P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61), D. hesperus (UHR 18466), and A. laticosta (AMP 21). That of D. hesperus (juvenile: GSJ-7743) is not oval but almost circular. In Paleoparadoxia sp. (AMP AK0011), the head of the humerus is incompletely preserved.
The humeral shaft becomes thinner in the middle in UMUT CV31059, P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61 and AMP AK1002), D. hesperus (UHR 18466), and A. laticosta (AMP 21). The shaft width of D. hesperus (juvenile, GSJ-7743, +28 mm) (Table 1) is thinner than that of UMUT CV31059 and almost constant throughout the entire length.
The humeral necks of UMUT CV31059, P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61), D. hesperus (UHR 18466), and A. laticosta (AMP 21) are deeply constricted. On the other hand, the degree of constriction is slightly smaller in one D. hesperus (juvenile, GSJ-7743). The humeral neck of Paleoparadoxia sp. (AMP AK0011) is poorly preserved and cannot be identified.
The trace of the triceps muscle (tricipital line) on UMUT CV31059 is more sharply demarcated than that on D. hesperus (juvenile, GSJ-7743, and adult, UHR 18466) and A. laticosta (AMP 21). Because the medial side of the humerus of P tabatai (NMNS PV-5601) is lacking and humeri of Paleoparadoxia sp. (SMNH VeF-61 and AMP AK0011) are incompletely preserved, we could not identify the tricipital line in these specimens.
A deltoid tuberosity is not developed among UMUT CV31059, P. tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61 and AMP AK0011), D. hesperus (juvenile, GSJ-7743, and adult UHR 18466), or A. laticosta (AMP 21).
Based on comparisons of a wide range of desmostylian humeral specimens described above, we newly propose the following diagnostic characteristics of Paleoparadoxia (Table 2): large greater tubercle, distinct lesser tubercle, medially projected lesser tubercle located on the medial side, shallow and narrow intertubercular groove located on the medial side, slightly convex oval head at the distal end, and absence of a deltoid tuberosity.
Right scapula (UMUT CV31060).—The overall shape of UMUT CV31060 is similar to the shape of scapulae in Paleoparadoxia tabatai (NMNS PV-5601), Paleoparadoxia sp. (SMNH VeF-61), and Ashoroa laticosta (AMP 21), In contrast, the neck of the scapula in Desmostylus hesperus (UHR 18466) is thicker than that of UMUT CV31060. The size of UMUT CV31060 is slightly smaller than that of Paleoparadoxia sp. (SMNH VeF-61) and much smaller than that of P. tabatai (NMNS PV-5601) (Table 1).
Rib (UMUT CV31061).—The cranial and dorsal outlines of UMUT CV31061 are similar to those of Desmostylus hesperus (UHR 18466) and Paleoparadoxia tabatai (NMNS PV-5601). Although the cross-sectional surface of D. hesperus (UHR 18466) and Desmostylus “Japonicus” (HMG-343) show a spongy structure inside, UMUT CV31061 and P. tabatai (NMNS PV-5601) show osteosclerosis. The rib shafts of Ashoroa laticosta (AMP 21) and Behemotops katsuiei (AMP 22) differ from that of UMUT CV31061 in that the former specimens are almost straight but the latter is slightly arched.
Discussion
Spatial distribution and habitat of Paleoparadoxia
The locality of UMUT CV31059-31061 is at the latitude of 44°23′N, thus representing the northernmost locality of the genus Paleoparadoxia. Previously reported latitudinal ranges of Paleoparadoxia occurfences were between 35°5′N and 43°08′N and 32°34′N and 37°34′N in the northwestern and northeastern Pacific regions, respectively (VanderHoof, 1937; Mitchell, 1963; Mitchell and Repenning, 1963; Barnes and Aranda-Manteca, 1997; Inuzuka, 2000a, 2005). The locality of UMUT CV31059-31061, therefore, extends the latitudinal range of Paleoparadoxia in the Pacific area farther north than previously recognized (Figures 3 and 6).
Table 2.
Comparison of character states found in the desmostylian humerus.

Figure 6.
Main localities of Paleoparadoxia specimens in the Pacific region (gray circles: see Table 4). UMUT CV31059-31061 is also indicated by a star.

Several studies suggested that the habitat of Paleoparadoxia was in subtropical to warm-temperate zones, including the frontal fringe of the tropical zone (Chinzei, 1984; Itoigawa, 1984; Ogasawara, 2000). The fossil pollen and mollusc assemblages of the Sankebetsu Formation indicate that it was deposited under cool to temperate conditions (Kurita et al., 1992; Noda, 1992; Suzuki, 2000), suggesting that Paleoparadoxia from the Chikubetsu area inhabited an environment cooler than those previously inferred for the genus. Through reviewing molluscan assemblages associated with other specimens of Paleoparadoxia (Table 3), we found that the Kintaichi specimen (IPMM32669, a nearly complete skeleton of Paleoparadoxia from Ninohe City, Iwate, northern Japan; Oishi et al., 1990) was also found associated with a cold-water molluscan assemblage (Masuda and Oishi, 1990), suggesting it inhabited a cool environment. Therefore, UMUT CV31059-31061 and the Kintaichi specimen demonstrate that habitats of Paleoparadoxia include a wide range of environments, from the warm-temperate to cool-to-warm—temperate zones, and are not restricted to the subtropical to warm-temperate zones as inferred previously.
Table 3.
Paleoclimates of the paleoparadoxian localities.

Table 4.
Geological ages of Paleoparadoxia specimens from Japan.

Temporal distribution of Paleoparadoxia and the early evolution of Pareoparadoxiinae
UMUT CV31059-31061 occurred in the Sankebetsu Formation, whose age estimate is bracketed between 25.3 and 20.3 Ma based on the FT method and diatoms (Kurita et al, 1992), as discussed above. Previously, Paleoparadoxia from the Chichibu Basin (SMNH Ve-61) dated to 19 Ma was the oldest specimen known from Japan (Ogasawara, 2000). The estimated age of UMUT CV31059-31061 represents the oldest record of Paleoparadoxia in the Northwest Pacific (Table 4). In the Northeast Pacific, Paleoparadoxia was found from the uppermost lower Miocene (VanderHoof, 1937; Mitchell, 1963; Mitchell and Repenning, 1963; Barnes and Aranda-Manteca, 1997; Inuzuka, 2005). The estimated age of UMUT CV31059-31061 also nearly matches the oldest record of the basal paleoparadoxiine Archaeoparadoxia weltoni in the Northwest Pacific, ca. 23.0–20.8 Ma (Clark, 1991). This indicates that the currently known oldest occurrences of basal and derived paleoparadoxiines (i.e., Archaeoparadoxia from the Northeast Pacific and Paleoparadoxia from the Northwest Pacific, respectively) overlap in age. Accordingly, it is likely that the geographic range of Paleoparadoxiinae had already expanded from the northeastern to northwestern Pacific regions in the early stage of their evolution by at least 20 Ma.
Inuzuka (2005, 2013) and Inuzuka et al. (2006) hypothesized that Paleoparadoxia (i.e., “Paleoparadoxia media” (SMNH VeF-61 and NMNS PV-5601)) evolved from Archaeoparadoxia weltoni (= “Paleoparadoxia weltoni”) in the early Miocene based on the temporal differences between these two species. However, the age of UMUT CV31059-31061 shows that both Paleoparadoxia and Archaeoparadoxia coexisted during the early Miocene, and thus this hypothesis is not supportable. Instead, Paleoparadoxia, a derived paleoparadoxiine, likely had already diverged from Archaeoparadoxia, a basal Paleoparadoxiinae, at the latest by 20.6 Ma (earliest early Miocene).
Acknowledgments
We are grateful to N. Kohno (NMNS), H. Naruse (Kyoto University), and T. Tsuihiji (University of Tokyo) for giving us helpful advice. Thanks are also due to T. Yamada and Y. Tajima (NMNS), N. Kaneko (GSJ), Y. Kobayashi and T. Tanaka (UHR), H. Sawamura and T. Shinmura (AMP), T. Nishimura and K. Sakurai (HMG), and O. Sakamoto and H. Kitagawa (SMNH) for allowing us to study specimens under their care. T. Okamoto (Ehime University) and K. Ogasawara (University of Tsukuba and NMNS) provided us with geological information about the Chikubetsu area. We also thank Y. Nakajima (Bonn University), Y. Takeda (UMUT), T. Karasawa (Kyoto University), T. Hiroki (Kyoto University), W. Anzai (UMUT), R. Fukiage (UMUT), and S. Kobayashi (UMUT) for their help and supports for this project. This manuscript was greatly improved by comments from Y. Iryu (Tohoku University), B. L. Beatty (New York College of Osteopathic Medicine), and T. Ando (AMP).
References
Appendices
Appendix 1. Diagnostic characters of referred specimens.
(i) NMNS PV-5601
NMNS PV-5601 is identified as P. tabatai based on the following characters: downwardly bent upper margin of the mandibular body in front of the second premolar, and slightly inwardly curved humeral crest. NMNS PV-5601 was considered an adult because the epiphyses in the humerus were completely fused (Hayashi et al., 2013; Barnes, 2013).
(ii) SMNH VeF-61
Saegusa (2002) identified SMNH VeF-61 as Paleoparadoxia based on the following characters: bunodont premolar, molars brachyodont and stylodont with extremely long roots, cingulum shape varies from subtle swelling to distinct shelf-like structure, greater tubercle of humerus is higher than the head, thin and wide humeral trochlea, and kesser trochanter of the femur is small and projecting. SMNH VeF-61 had fully fused epiphyses in the humerus and was considered an adult.
(iii) UHR 18466
UHR 18466 is referred to as D. hesperus based on such characters as high-crowned stylodont teeth with more cusps than other desmostylians and no cingulum, a linear lateral epicondylar crest of the humerus, and the large diameter of the humeral trochlea. UHR 18466's humerus was fully fused and was considered an adult (Hayashi et al., 2013).
(iv) GSJ-F7743
GSJ-F7743 has neurocentral sutures of the vertebrae and long bone epiphyses were partially fused, thus indicating a juvenile (Hayashi et al., 2013). GSJ-F7743 was identified as D. hesperus because of its diagnostic characteristics such as high-crowned stylodont teeth with more cusps than other desmostylians and no cingulum, and a narrow narial opening high on the maxillary rostrum.
(v) AMP 21
AMP 21 possesses diagnostic characters such as bunodont teeth and very stout ribs relative to their length. AMP 21 had fully fused humeri and was considered an adult (Hayashi et al., 2013; Barnes, 2013).
(vi) AMP 22
AMP 22 possesses the following diagnostic characters: cheek teeth are disproportionately smaller for a larger mandible, canine is minimally compressed in cross-section, and the buccal surface of the mandible is slightly expanded. AMP 22 has third molars (M3) and was therefore considered a subadult (Hayashi et al., 2013).
(vii) AMP AK0011
This specimen is identified as Paleoparadoxia because the lateral epicondylar crest of the humerus is anteriorly bent and the diameter of the humeral trochlea is small. AMP AK0011 had a fully fused humerus and was considered an adult (Hayashi et al., 2013).
(viii) LACM 150150
LACM 150150 was designated as the species holotype based on the following diagnostic characters: larger dorsal naris, transversely wider nasal bones, larger and more dorsally placed orbits, laterally farther extended margin of orbit because of the flared ventral edge of the jugal at the anterior end of the zygomatic arch, and dorsoventrally deeper jugal ventral to the orbit (Barnes, 2013). Epiphyses of LACM 150150 humeri are not fused and the specimen was thus considered a juvenile (Barnes, 2013).
(ix) HMG-343
HMG-343 is identified as Desmostylus based on high-crowned stylodont teeth with more cusps than other desmostylians and no cingulum. Because HMG-343 has third molars (M3), it was considered a subadult.