The locality is situated in a large outcrop about 700 meters northwest of Kalfa [Calfa, Калфа] village (Anenii Noi district, Moldova) on the southern escarpment of the Byk river valley (Fig. 1). The remains of small vertebrates come from the alluvial deposits consisting of various limestones and clays with shells of mollusks (e.g., marine Mactra podolica, M. fabreana, Plicatiforma fittoni, Solen subfragilis, Cerithium comperei and/or freshwater Lymnea sp., Planorbis sp.). The small extent and thickness of bone-bearing layers, apparently deposited in a short time, suggest an activity of periodical streams in the vicinity of the Sarmatian river (Lungu and Rzebik-Kowalska, 2011). The most detailed studies of the geology, taphonomy, and age of the locality were provided by Lungu, 1978 and Lungu, 1981 and Lungu and Rzebik-Kowalska (2011).

Kalfa has yielded about thirty mammalian taxa (see Lungu, 1978, Lungu, 1981, Lungu, 1984, Lungu, 1990 and Lungu and Rzebik-Kowalska, 2011 for details), but, apart from ochotonids, small mammals are very scanty and give little information about the mammalian age. The available taxa are comparable to those reported from Buzhor 1 (Khynchesht district) that has yielded a much more diverse small mammal community including a few taxa of high biostratigraphic value. Nevertheless, the exact age of Kalfa and Buzhor 1 still remains unclear. The locality under study was originally referred by Lungu (1981) to the Middle Sarmatian. In the eastern Paratethys, the Sarmatian stage (sensu Suess, 1866) is considered to include the Volhynian and (Lower) Bessarabian (Simionescu, 1906) substages; further east to the Caspian Sea, the Sarmatian also includes the Upper Bessarabian and Khersonian (Simionescu, 1906), and is commonly labeled as Sarmatian (s.l.) as proposed by Barbot de Marny (1869). The Middle Sarmatian corresponds approximately to the Early Vallesian (ca MN9). This age for Kalfa was followed by, e.g., Rzebik-Kowalska and Lungu (2009), Lungu and Rzebik-Kowalska (2011), or Delinschi (2014). On the other hand, Topachevski et al. (1997) and Nesin (2004) refer Kalfa, based on the mammals listed by Lungu (1981), tentatively to the Late Vallesian (MN10).

Vangengeim et al. (2006) provided a magnetostratigraphic analysis of Middle Sarmatian deposits in Hipparion localities of Moldova and correlated them to an upper part of chron C5An (upper boundary 11.9 Ma) or, less likely, with subchron C5r2n (base 11.5 Ma); it corresponds in their view to the base of the Vallesian. However Vasiliev et al. (2011) question the age by Vangengeim et al. (2006) and argue that it was only based on the correlation of the Khersonian–Meotian boundary to an age of 9.8 Ma and the predominantly normal polarities of the Khersonian to C5n.2n. The new radio-isotopic data by Vasiliev et al. (2011) for the Khersonian–Meotian boundary indicates a correlation of the normal polarity interval of the Khersonian to the younger chron C4An. The reversed intervals with the Bessarabian–Khersonian boundary may consequently correspond to C4Ar. Thus the Hipparion sites in Moldova may be significantly younger than suggested by Vangengeim et al. (2006).

The results of Vasiliev et al. (2011) are in a good agreement with the view of M. Sinitsa (Sinitsa, pers. comm. 2014; Sinitsa and Delinschi, in press) who, based on the presence of at least two murid genera and Spermophilinus turolensis, and the absence of Democricetodon, Megacricetodon, and Microtocricetus, correlate the small mammal faunas of Kalfa and Buzhor 1 with the Late Vallesian (early MN10), thus being younger than those of Götzendorf (AT), Rudabánya (HU), and Gritsev (UA). This correlation seems to be most probable and is followed in this paper.

The revision of the species kalfense is based on the original material by Lungu (1981) from the Kalfa type locality (Moldova, early MN10) and the material in France published by Sen (2003). The studied material is curated in the collections of the TSU (prefix FTGU Lag/Kfa-) and MNHN (prefix MNHN Lag/Kfa-). The interspecific comparisons, using original material (indicated by “*”) or a bibliographical basis (unless otherwise stated, data were taken from original descriptions of species), were made with: •

Bellatona forsythmajori Dawson, 1961 from Tairum Nor of the Tunggur formation [type locality] (Nei Mongol, CN)—Middle Miocene (Early Tunggurian, Early Astaracian, MN6; Qiu Z.-D. et al., 2013);

All drawings and measurements were taken with the aid of a binocular microscope with camera lucida and ocular micrometer. Dental measurements were taken as maximum anteroposterior or bucco-lingual dimensions of the respective two-dimensional dental structures orthogonal to the prismatic shaft. All measured data are given in millimeters. Dental nomenclature follows Sen (1998), measurements Erbajeva (1988) and Čermák (2004). Jaw measurements follow Erbajeva et al. (2001) with several modifications for the lower jaw (heights of mandibular body are measured on the lingual side). The biostratigraphic terminology used in this paper follows Steininger (1999). Outline maps used in Fig. 1 were taken from http://d-maps.com/. Country abbreviations follow ISO 3166–1 alpha–2 codes.

Class Mammalia Linnaeus, 1758

Order Lagomorpha Brandt, 1855

Family Ochotonidae Thomas, 1897

Subfamily Ochotoninae Thomas, 1897

Genus Ochotona Link, 1795

Ochotona kalfense (Lungu, 1981)

Name-bearing type, type locality, and age: Holotype (sensu Article 73.1 of ICZN, 1999) — left mandible with p3-m3 (Fig. 2A–F), [coll. No. 2 (118)] sensu Lungu (1981), deposited in the TSU; type locality (sensu Article 76.1 of ICZN, 1999) — Kalfa (Late Miocene, Late Vallesian, early MN10). In the original description, Lungu (1981: 46) wrongly referred to the holotype mandible as a right one (cf. Lungu, 1981: pl. 5, fig. 4).

Stratigraphic and geographic distribution: Late Vallesian (early MN10) of Moldova, from localities Kalfa and Buzhor 1.

Material analysed (original material described by Lungu (1981), prefix TSU Lag/Kfa-): Seven mandibular bodies sin with p3-m3: 1, 2; m1-m2: 3; p3-m2: 4; p4-m1: 5; p4-m3: 6; m1-m3: 7. Seven mandibular bodies dex with p3-m3: 8; p3-p4: 9; i-p3-m3: 10; p3-m2: 11; p4-m2: 12; p4-m3: 13; m2-m3: 14. One mandibular ramus dex: 15. Two p3 sin: 16, 17. Five p3 dex: 18-22. Two maxillae sin with P3-M2: 24; P3-M1: 25. One maxilla dex with P2-M2: 23. One P2 sin: 26. Two P3 sin: 28, 29. One P3 dex: 27. One M2 sin: 30. (prefix MNHN Lag/Kfa-) Four p3 sin: 1–4. One p3 dex: 5.

Measurements: For measurements, see Table 1 and Table 2.

Description: Mandible (Fig. 2): In adult individuals — mandibular body relatively robust and high; alveolar length of p3-m2 distinctly longer than diastema; posterior mental foramen oval and located ventral to the area between the trigonid of m2 and m3, in most cases below the talonid of m2 (ca 56%, N = 9); area below p3 richly fenestrated; root end of lower incisor extending in 50% of cases (N = 10) below the trigonid of m1; ventral margin of the mandible variable, predominantly convex; mylohyoid line long, moderately developed, from the ventral view lying in the buccal third of the mandibular body; articular head large; coronoid process weakly developed, dorso-ventrally shortened; mandibular ramus low, antero-posteriorly elongated.

p3 (Fig. 3): Medium to large size (OR = 1.54–1.95); in adult individuals–large anteroconid (M of L × W = 0.79 × 0.90), typically asymmetric (buccally distinctly widened into semicircular or rectangular buccal outline), often anteriorly with depressions, but never with re-entrant folds (anteroflexids); proto- and paraflexid simple without crenulation, the former typically longer than the latter; antero-posteroconid junction moderately wide (M = 0.20); hypoflexid short, shorter than the half of W p3.

Maxilla: premolar foramen small-sized, oval, next to the anteroloph of P4.

P2 (Fig. 4A): single-lobed tooth of rectangular occlusal outline with anteriorly convex lagiconus and sharpened hyperconus; no enamel hiatus; moderately developed paraflexus.

P3 (Fig. 4A–B, D–E): occlusal morphology and widths quite variable; tooth of trapezoidal occlusal outline with PPW ranging from 52% to 66% of W; hypoflexus well-developed, distinctly narrowed buccally.

Upper and lower molariforms (Figs. 2C and 4A–C): P4-M2 and p4-m3 with no taxonomically important features discriminating them from the corresponding teeth of other Ochotona species; M2 posteroloph with well-developed posterolingual process.

Notes on morphological changes of p3 through ontogeny: In all prismatic teeth (Fig. 3A, A′, D–G, I, I′) of the studied sample, the general occlusal outline is comparable with that of the root, and no important morphological changes are visible. In young individuals, with conical teeth (Fig. 3K–N′), the root morphologies — representing a stage that appears with wear in more adult ontogenic phases — differ from the occlusal ones and thus several morphological changes through ontogeny can be traced. The anteroconid displays most changes among all p3 structures. In very young individuals (Fig. 3O), i.e. in very early stage of p3 abrasion, the anteroconid is small and almost symmetrical. Through ontogeny, it grows in relative size. In early permanent teeth, both widened and elongated anteroconids can be recognized in the sample; later in the ontogenetic development the anteroconid extends buccally and becomes predominantly more asymmetric. In adult specimens, a degree of anteroconid asymmetry (buccal extension) is very variable in contrast to relatively stable anteroconid length (Fig. 2). At any rate, a tendency to anteroconid asymmetry is possible to recognize in all ontogenic stages of permanent teeth. Concave anterior walls of anteroconid generally appear in largely asymmetric anteroconids. Proto- and paraflexid are very variable with no recognizable trends in their morphologies, deflections, and relative lengths (depending on the asymmetry of anteroconid) through ontogeny. The lingual side is predominantly concave (e.g., Fig. 3B, C), occasionally with incipient mesoflexid (Fig. 3G); the phenomenon seems not to be related to the age of the individual. The relatively small sample from the type locality does not allow a credible quantification of p3 morphotype composition in O. kalfense.

Comparisons: O. kalfense differs from all available species of Bellatona and from B. eroli of Sinap Tepe in its larger size (Fig. 5), larger and well-developed p3 anteroconid, and longer and narrower P3 hypoflexus; from the former also in its well-developed p3 proto- and paraflexid and constant presence of posterolingual process in M2. The majority of p3 specimens from Damiao (DM01) referred to by Zhang et al. (2012) as Bellatona cf. B. forsythmajori, B. eroli, and Ochotona cf. O. lagreli are generally smaller than those in O. kalfense (Fig. 5). Moreover, in contrast to O. kalfense, p3s attributed to (1) Bellatona cf. B. forsythmajori has an incipient or no paraflexid, a smaller anteroconid, and probably an unstable presence of posterolingual process in M2; (2) B. eroli has a less developed paraflexid, a wider anteroconid-posteroconid connection, and a shorter and more simple anteroconid; (3) Ochotona cf. O. lagreli have a more symmetrical anteroconid (developed lingually) and the anteroconid–posteroconid connection is generally narrower. Two p3s from Bulstyn Kudak (MN) are smaller and show similar morphological differences as Ochotona cf. O. lagreli from DM01. Among the Late Miocene Ochotona of the peri-Paratethyan area: (1) O. ozansoyi has a smaller p3 with the symmetrical anteroconid and narrower anteroconid–posteroconid connection, a more linguo-buccally oriented p3 paraflexid, a shorter and wider P3 hypoflexus, and a slightly longer and thinner posterolingual process in M2; (2) O. eximia has a larger and relatively elongated p3 with the symmetrical anteroconid. The Pliocene species of the peri-Paratethyan area, Ochotona antiqua, O. mediterranensis, and O. ursui, are smaller than O. kalfense (Fig. 6). Moreover, in contrast to O. kalfense, O. antiqua has a more symmetrical p3 anteroconid (if asymmetrical, its lingual part is generally enlarged), a narrower p3 anteroconid–posteroconid connection, a less robust mandible with posterior mental foramen situated more posteriorly, a relatively lower mandibular body (MHL is 69 for O. antiqua vs. 84 for O. kalfense), and a more developed condylar process of mandibular ramus. More detailed comparison with O. mediterranensis and O. ursui is impossible due to lack of sufficient published data. This is also true for O. pseudopusilla (see Čermák and Rekovets, 2010 for details). In the peri-Paratethyan area, there are several scanty findings with unsure taxonomic attribution. Their p3 size relationship to those of O. kalfense is shown in Fig. 6. From a morphological point of view, in contrast to O. kalfense: (1) Ochotona cf. ozansoyi from Inönü (TR) has a narrower p3 anteroconid–posteroconid connection, a smaller symmetrical p3 anteroconid, a shorter and wider P3 hypoflexus, and a more developed mylohyoid line in the mandibular body — from the ventral view situated more medially (positions of the posterior mental foramen and root end of lower incisor in mandibular body are comparable with those of O. kalfense); (2) Ochotona sp. from Inönü (Loc. 45; TR) has a p3 with symmetrical anteroconid and developed mesoflexid; (3) Ochotona sp. from Razeni (MD) has a p3 with the larger and distinctly more triangular anteroconid, a narrower anteroconid–posteroconid connection, and proto- and paraflexid more parallel to linguo-buccal axis; (4) Ochotona sp. from Maritsa I (GR) has a p3 with the symmetrical anteroconid of rhombic outline and narrower anteroconid-posteroconid connection; (5) Ochotona sp. from Frunzovka 1 (UA) has a triangular outline of p3 with the prolonged and markedly narrow anteroconid, a shorter and wider posteroconid, and a narrower anteroconid-posteroconid connection with notably more posteriorly oriented proto- and paraflexid; (6) three p3s (all prismatic) from Gradiste (MD) show a noticeable morphological variability making comparison difficult; the p3 varies (i) in the anteroconid from large, symmetrical, and triangular to small, asymmetrical, and rounded, (ii) in the anteroconid-posteroconid connection from narrow to very wide.

Emended diagnosis: Medium to large-sized p3 with large buccally widened asymmetric anteroconid of semicircular or rectangular buccal outline, often anteriorly depressed, anteroconid-posteroconid junction moderately wide; well-developed proto- and paraflexid without crenulation. Mandibular body robust and high; p3-m2 alveolar row distinctly longer than diastema; posterior mental foramen predominantly below the talonid of m2; mylohyoid line long, moderately developed, from the ventral view in the buccal third of the mandibular body. M2 posteroloph with well-developed posterolingual process.

Remarks: The material under study is attributed here to the subfamily Ochotoninae (first unequivocally separated by Sen (2003) as the “Ochotona group”), a monophyletic lineage derived from Bellatona Dawson, 1961 and comprising here Bellatonoides Sen, 2003; Ochotona; Ochotonoma Sen, 1998; Pliolagomys Agadjanian and Erbajeva, 1983; and Ochotonoides Teilhard de Chardin and Young, 1931 (Čermák, 2010). This subfamily is characterized by the combination of autapomorphies including: (i) the posterolingual process of M2, (ii) well-defined anteroconid with round or more or less triangular outline on p3, and (iii) the presence of the opposite paraflexid and protoflexid on p3 (Sen, 2003). This subfamily demarcation differs significantly from a taxonomical meaning of Ochotoninae defined by Erbajeva, 1986, Erbajeva, 1988, Erbajeva, 1994, Erbajeva et al., 2015 and Erbajeva et al., 2011 [in earlier works she used the name Lagomyinae Lilljeborg, 1866 — see Čermák [2010: 10] for explanation]). In her view Ochotoninae comprises also Marcuinomys Lavocat, 1951; Lagopsis Schlosser, 1884; Albertona López-Martínez, 1986; Alloptox Dawson, 1961; and Paludotona Dawson, 1959 [grouped tentatively here in the separate subfamily Lagopsinae], despite assuming that the subfamily was derived from the Bellatona stock (Erbajeva, 1994 and Erbajeva et al., 2011). Moreover, she places Bellatonoides — a presumed direct descendant of Bellatona and ancestor of Ochotona (Sen, 2003 and Zhang et al., 2012) — in different subfamily Sinolagomyinae. Even though the phylogenetic interrelationships among genera of Lagopsinae is still unclear, the phenotypes in the earliest representatives of Marcuinomys, Lagopsis, or Alloptox, all available already in the middle Early Miocene (i.e. MN2–3; Daxner-Höck et al., 2013, Hordijk, 2010, Lavocat, 1951 and Tobien, 1963), are distinctly different from those available in, more or less coeval, the earliest records of Bellatona. This strongly indicates that these taxa cannot be derived from the Bellatona stock. The present state of knowledge rather supports a hypothesis that Marciunomys most likely branched from a European rooted form like Titanomys von Meyer, 1843 and subsequently gave rise to Lagopsis (Hordijk, 2010 and references therein). Paludotona has been suggested as a possible descendant of presumed relict population of Lagopsis (Dawson, 1967, Hordijk, 2010, López-Martínez, 1986 and López-Martínez, 1989). Phylogenetic affinities of Albertona and particularly Asian Alloptox are less clear and are need of revision. The subfamily Ochotoninae sensu Erbajeva (1994; Erbajeva et al., 2015 and Erbajeva et al., 2011) is not a monopfyletic.

I revised the original ochotonid material from the early Late Miocene Kalfa locality (Moldova) described by Lungu (1981) as “Proochotona” kalfense. The taxonomic analysis of this taxon was supported by a direct examination of type materials of P. eximia, B. eroli, and O. ozansoyi, the relevant Late Miocene ochotonines of the peri-Paratethyan area.

Based on the p3 features in the holotype of P. eximia (the type species of the genus Proochotona), the species eximia undoubtedly belongs to Ochotona. Therefore, the genus Proochotona Khomenko, 1914 is regarded here as a junior subjective synonym of the genus Ochotona Link, 1795. The morphometric analyses of the material from Kalfa confirm the taxonomic validity of the species kalfense. Its p3 features are also typical and discriminative for Ochotona [cf. the emended generic diagnosis given by Erbajeva (1988: 82; cf. Gureev, 1964: 226, Gromov and Erbajeva, 1995: 34–35)], and thus the species kalfense is referred here to that genus. Considering all the available data, I propose here, following Čermák (2010), to retain for now the original concept of Bellatonoides as a distinct genus.

O. kalfense from the early Late Miocene (early MN10) locality of Kalfa represents the oldest record of Ochotona in Europe; it seems to be coeval with O. ozansoyi described from Sinap Tepe (loc. 84). These species demonstrate the occurrence of suprisingly advanced p3 phenotype already in the pre-Turolian of the peri-Paratethyan area. Among the Chinese localities, the age of Kalfa seems to be comparable with Shengou and lower part (Unit/Biozone BH1) of Bahe Fm., thus younger than Amuwusu and older than the Late Bahean Ochotona bearing localities Huitenghe and Shala. The material under study is attributed here to the subfamily Ochotoninae; the monophyletic lineage derived from Bellatona. A more conclusive evaluation of phylogenetic relationships of Ochotona from Kalfa to other relevant species of the clade discussed in this paper is still impossible. Nevertheless, the lateral displacement of p3 anteroconid documented in taxa from Damiao (DM01), Sinap Tepe (Locs. 8A, 120), and Kalfa may indicate a phylogenetic closeness. This feature has not yet been documented in later European taxa of Ochotona recorded from the Late Miocene (MN11–13) of the peri-Paratethyan area.