Carnivora

ABSTRACT The locality of Küçükçekmece has yielded a unique carnivoran fauna representing one of the rare fossil assemblages comprising both marine and terrestrial species. The studied sample comprises thirty-four specimens belonging to three different collections at the Technical University of Istanbul (ITU), Geology Museum of the Istanbul University (IU) and at the National Museum of Natural History of Paris. This low number of fossil remains contrasts with the relatively great taxonomic richness of the fauna, since six families and at least nine species are identified: the ursid Indarctos arctoides (Depéret, 1895), the mustelid Sivaonyx hessicus (Lydekker, 1884), the phocids Cryptophoca sp. and Phocidae gen. et sp. indet., the percrocutid Dinocrocuta senyureki (Ozansoy, 1957), the hyaenid cf. Thalassictis sp., the felids Machairodus aphanistus (Kaup, 1832), Pristifelis sp. cf. P. attica (Wagner, 1857) and Felidae gen. et sp. indet. medium size. A tenth species, a small mustelid identified as Mustela pentelici Gaudry, 1861 by Malik & Nafiz (1933), was probably present. However, the figured (and only?) specimen was destroyed during the 1942 fire at Istanbul University. Most of the species are documented by one or a few specimens, while the phocid material is the most abundant and represents more than half of the described fossils, with many postcranial remains. The list of Carnivora of Küçükçekmece suggests a late Miocene age for the fauna of this site.


INTRODUCTION
Although it was discovered in the early 1930's, the site of Küçükçekmece remains poorly known in the scientific community, with only a handful of recent work discussing or reporting on this locality (e.g., Rückert-Ülkümen & Kaya 1993;Koufos 2006). Of the twenty or so mammalian taxa that are present in the first faunal list published by Malik & Nafiz (1933), about one-third are of the order Carnivora: the mustelids Mustela pentelici Gaudry, 1861 and Lutra sp., the ursid Ursavus sp., the hyaenid Ictitherium sp., the machairodontine felid Paramachairodus orientalis (Kittl, 1887) (named Machaerodus orientalis in their contribution) and the phocid Phoca sp. Sayar (1953) added Machairodus aphanistus (Kaup, 1832) to this list, but Yalçınlar (1954) and Ozansoy (1957) did not list this species and proposed the same faunal list as in Malik & Nafiz (1933). By comparison, Nicolas (1978) provided a somewhat different faunal list based on his own discoveries: the mustelids Lutra pontica Nordmann, 1858 and Sivaonyx hessicus (Lydekker, 1884), the percrocutid Percrocuta senyureki (Ozansoy, 1957), the hyaenid Ictitherium sp. cf. I. orbignyi (Gaudry & Lartet, 1856), the ursid Indarctos arctoides (Depéret, 1895), the felids Machairodus aphanistus (Kaup, 1832) and Felis sp. aff. F. prisca Kaup, 1833, indeterminate mustelids and viverrids, and the phocids Phoca maeotica Nordmann, 1860 and Phoca vindobonensis Toula, 1898. The carnivorans from Küçükçekmece have never been described in detail and are infrequently mentioned in the literature, with one notable exception, however: the dental remains of the ursid were studied and assigned to Indarctos arctoides by Petter & Thomas (1986). The objectives of this paper are to describe in details the material from Küçükçekmece assigned to the Carnivora and to discuss its significance in the context of the late Miocene faunas of Europe and western Asia.

MATERIAL AND METHODS
The material from Küçükçekmece studied in this paper comes from three distinct collections. The specimens stored at the Muséum national d'Histoire naturelle, Paris (MNHN) were collected by the late Father Jean Nicolas between 1957and 1967(Nicolas 1978. Nicolas (1978) collected this material along the southwestern cliffs of the Küçükçekmece Lagoon and the adjacent shores of the Marmara Sea. In the present study this locality is called Küçükçekmece West and the specimens are labelled TRQ. The material housed at the Technical University of Istanbul (ITU) and at the Geology Museum of the Istanbul University (IU) issued from the excavations of Malik & Nafiz (1933) and later discoveries at the original Küçükçekmece locality, which is situated on the southeastern bank of the Küçükçekmece Lagoon. This locality is here called Küçükçekmece East, and the specimens are labelled KÇ and IU, respectively. The 1942 fire at the Geology Institute of Istanbul University, where the collection of Malik & Nafiz (1933) was preserved, destroyed most of the specimens studied and/or figured by these authors (see Sen 2016). Both localities (Küçükçekmece East and West) are in the sandy-conglomeratic horizon of the Çukurçesme Formation, and thus considered of the same age (see Lom et al. 2016;Sen 2016).
Lower case is used for lower teeth, upper case for upper teeth. All measurements are in mm and, unless otherwise noted, were taken with vernier callipers to the nearest 0.1 mm. Estimated measurements are labelled 'est' in tables and text. Anatomical nomenclature follows FIPAT (2011).

AbbreviAtions
TLMd total length of mandible; Lp1-4 length of p1-4; DMp2-3 lingual depth of the dentary between p2 and p3 (or between any other teeth); TMp3/m1 thickness of the dentary across p3/m1; L maximum length; W maximum width; TWm1 maximum width of talonid; H height (from the tip to the dentine-enamel junction on the labial face of teeth); HPrm1 height of the protoconid of m1; TLim1 trigonid lingual length of m1; TLam1 trigonid labial length of m1; LMeP4 length of the metastyle of P4; LPaP4 length of the paracone of P4.
referred MAteriAl froM KüçüKçeKMece. -MNHN.F.TRQ947, subcomplete left hemimandible with the root of the canine, the root of p1, the alveolus of p2, the fused roots of p3, p4-m2, and the root of m3.
description Mandible (Fig. 1A fig. 2). However, it does not display any obvious pathology. The depth of the mandibular body is rather constant below p1-p4 (DMp2-3 = 39.3; DMp3-4 = 40; DMp4-m1 = 41), then increases slightly from m1 and posteriorly (DMm1-2 = 43.8; DMm2-3 = 46). Thickness of the dentary across p3 and m1 is 17.6 mm. The coronoid process is not preserved; the masseteric fossa has its dorsal part missing, is approximately 75-mm long and shallow. The anterior margin of the fossa is situated below the distal border of m3. The ventral margin of the fossa is marked by a crest, which continues posteriorly to the tip of the angular process. The latter is short and extends medially by a small, dorsoventrally flat process. The process displays an additional, ventral crest marking the lateral extension of the internal pterygoid muscle insertion. This insertion is particularly visible in the lingual face of the mandible, where it occupies a dorsally curved, oval-shaped area of 35-mm length. The anterior limit of this area is marked by a conspicuous tubercle located ventrolingually on the mandibular body.
Dentition (Fig. 1A, B) The p1, p2, p3 and m3 are not preserved. The p1, p2 and m3 are single-rooted, with alveolar length of 9, 8.6 and 13 mm, respectively; the p3 has two roots that are mostly fused, for a total alveolar length of 12 mm. The tooth row is 125.5mm long, from the mesial margin of the alveolus of p1 to the distal margin of the alveolus of m3. There is no distinct diastema between cheek teeth but a small one between p2 and p3 (L = 3 mm). The premolar row (L = 52 mm, measured at alveolar level) is much shorter than the molar row (L = 74 mm, measured at alveolar level). The teeth are lightly worn. Peigné S. The main cuspid of the p4 (L = 17.7; W = 9.4) is broken off and a small, mesiolingual fragment of the crown is missing. Distal and mesial accessory cuspids are vestigial; the talonid is long, narrow, and marked by a central, longitudinal crest; it is bordered distally by a distinct cingulid. The m1 (L = 33; W = 15.7; TLim1 = 22.4) is elongated and low, especially the trigonid cuspids. The protoconid is the tallest trigonid cuspid, the paraconid and metaconid being approximately of equal height. The paraconid and protoconid have an orientation parallel to that of the long axis of the body of the mandible. The metaconid is located distolingually to the protoconid and is visible in labial view. The labial face of the tooth is marked by a pronounced concavity at the trigonid/talonid junction. The talonid is much wider than the trigonid and the labial cingulid is particularly developed; its lingual (and not labial, contra Petter & Thomas 1986: 575) crest bears two low cuspids, an entoconid distally, and a slightly larger and mesially located entoconulid. The crista obliqua is low and diverges distad to the hypoconid, which is the only labial cuspid. The talonid basin is shallow. The m2 (L = 25.3; W = 17.6; TWm1 = 16.6) is rectangular in outline, with a trigonid slightly longer and, especially labially, wider than the talonid. The trigonid is composed of a horseshoe-shaped basin outlined mesially by a circular crest and distally by a transverse crest connecting protoconid and metaconid. The lingual margin of the talonid bears two vestigial cuspids, the entoconulid being the most developed one. The hypoconid is partly broken and/or worn. The m3 is not preserved but its root is long and narrow.

coMpArisons And discussion
The genus Indarctos is a typical member of the late Miocene Holarctic faunas. In Eurasia, its species content is still disputed, with 2 to 5 species (see, for example, Hunt 1998; Baryshnikov 2002;Geraads et al. 2005 description A small portion of the dentary is preserved (Fig. 1C-E); at the level of the carnassial it is thick and deeper than the length of the m1. The carnassial is typical of lutrines. It is robust, elongated though low and wide. The trigonid cuspids are low and of approximately equal height; the metaconid is strong and slightly distal to the protoconid; the paraconid is large, triangular in occlusal shape, and occupies nearly the entire mesial half of the trigonid; its distal face is transversely oriented. A prominent crest (= protoconulid, in Morales & Pickford 2005;= postprotoconid cusplet, in Pickford 2007) is present at the distal base of the protoconid; faint notches separate this crest from the protoconid and the hypoconid crest of the talonid. The trigonid basin is deep and opened lingually so that the paraconid and metaconid are deeply separated. A strong cingulid is present on the labial side of the crown and on the lingual side of the paraconid. The talonid is wide and rounded, with a shallow basin; the hypoconid crest is wide (the lingual slope occupies one-half of the talonid width) and tall relative to the lingual ridge. The hypoconid is poorly individualized, but is slightly prominent relative to the distal rim of the talonid. The tooth has its maximum width approximately at the trigonid/ talonid boundary.

coMpArisons And discussion
The generic assignment of the material is based on a comparison with most of the late Miocene and Pliocene species of lutrines. Besides the robustness of the dentary, which is hard to estimate here, species of Sivaonyx are characterized by "an m1 with a low trigonid with cuspids of approximately equal height, deeply separated paraconid and metaconid, a talonid wider than the trigonid, a large hypoconid crest, a shallow talonid basin, a low but distinct entoconid crest, a cingulid marked on the labial side that may extend on to the lingual side of the paraconid" (Peigné et al. 2008: 795). These features are present in KÇ 83. An additional diagnostic feature of the species of Sivaonyx mentioned by Pickford (2007) is mandibular depth greater than m1 length, which is also observed in KÇ 83, but this may also be the case in non-bunodont otters such as Lutra spp. (personal observations). The specimen from Küçükçekmece differs from Vishnuonyx Pilgrim, 1932 (and extant genera such as Lutra Brisson, 1762) in having an m1 that is more bunodont and with a lower trigonid relative to the talonid, and from more bunodont taxa (e.g., Enhydriodon Falconer, 1868, Djourabus Peigné et al. 2008) in having an m1 that is less bunodont and a more slender dentary. Many species of Sivaonyx have been described, especially from Africa and southern Asia (Morales & Pickford 2005;Pickford 2007;Peigné et al. 2008;Werdelin & Peigné 2010). Most of these species have an m1 with a wider talonid and, generally, a larger size than the specimen from Turkey described here (Table 1). In addition, according to Pickford (2007), S. gandakasensis Pickford, 2007 has also a shorter trigonid (57-64%) than in S. hessicus.
The specific assignment is based on a comparison with the holotype and single specimen of the European species originally described as Lutra hessica, BMNH 27486. This specimen is a fragment of right hemimandible with a fragment of p4 and m1 from the German locality of Eppelsheim (MN9, late Miocene). The geographic and stratigraphic proximity of the specimens of Eppelsheim and Küçükçekmece, in addition to their morphological similarity, support their assignment to the same species. Additional material is required to confirm this hypothesis. Because the specimen from Eppelsheim displays the diagnostic features of the genus Sivaonyx mentioned above, I propose to name this species Sivaonyx hessicus, an hypothesis already suggested by Morales & Pickford (2005) and, especially, by Pickford (2007). As noted by Pickford (2007), however, there is not much difference between Sivaonyx hessicus and the type species of the genus, S. bathygnathus. The species from Germany and Turkey may prove to be synonymous with this Asiatic species when more material is known. In his review of the Pliocene and Quaternary European otters, Willemsen (1992) concluded that the morphology of the m1 of Lutra hessica suggested a closer relationship to Lutrini (i.e. Lutra, Lutrogale Gray, 1865, Pteronura Gray, 1837 and their fossil relatives; see Willemsen 1992: 114), but his comparison was biased since he did not compare Lutra hessica with the species of Sivaonyx, of which many species have been described since then.    Table 2) The dentary is broken off approximately one cm posterior to the distal alveolus of m1 and also lacks a small part in the anterior extremity (the canine alveolus is not completely preserved). The mandibular body is overall thick, but its ventral rim is pinched. Dentary depth reaches its maximum below p4-m1 (DMp4-m1 = 18.6), where the ventral rim of the body shows a marked convexity. This convexity has various names in the literature: digastric apophysis (Grigorescu 1976), mental protuberance (Koretsky & Ray 1994), chin prominence (Koretsky & Grigorescu 2002). In the specimen from Küçükçekmece East, the mental protuberance is present though not marked. From this protuberance the dentary tapers off anteriorly and its ventral rim becomes oblique relative to the alveolar border. The symphysis is short and does not extend beyond the level of the mesial alveolus of p2; its surface is rugose only in its anterior part. On the labial face there are many mental foramina below p2 and partly coalescent ones below p3 and the diastema between p3 and p4. According to the preserved alveoli the p1 was large (nearly as large as the canine) and single-rooted, and the p2, p3, p4, m1 were biradiculate. Alveolar lengths indicate an increasing length from p2 to m1. The greatest diameter of the alveolus of p1 nearly reaches the p2 alveolar length. Diastemata are short, a few mm long, but they are present between all teeth; the longest diastema is the one between p4 and m1 (L = 4 mm).
The tooth row (p1-m1, alveoli) is 48-mm long. The tooth row is not set strictly parallel to the long axis of the dentary, but it is slightly twisted, the p1 being set labially and the m1 being set more lingually.

Postcranium
Anatomical descriptions are based on the description of extant phocids (Howell 1929;Piérard 1971) and middle-late Miocene taxa of Europe (Koretsky 2001;Koretsky & Grigorescu 2002;Berta et al. 2015). Table 3). The humeri are robust and, though none is complete, they are complementary: MNHN.F.TRQ930 and TRQ933 preserve the proximal part of this bone and TRQ935 preserves the distal two-thirds. The greater and lesser tubercles are greatly enlarged and projected far from the shaft. The greater and lesser tubercles extend proximal to the head. The humeral head is rounded, with mediolateral and anteroposterior (dorsoventral) diameters of similar size. The lesser tubercle is projected essentially to the medial and posterior sides. The intertubercular sulcus (or bicipital groove) is deep and has a maximum width of 7 mm. The deltopectoral crest is strongly developed and thick, especially at maximum crest height, where the deltoid tuberosity is prominent and slightly swollen laterally in TRQ933 and TRQ935; the crest is long and extends distally to approximately 10 mm from the distal articular surface. The epicondylar crest is prominent and anteroposteriorly thick, though not as developed and long as the deltopectoral crest. Nevertheless, it  extends proximal to the distalmost point of the deltopectoral crest. The epicondylar crest projects mainly posteriorly and poorly laterad; the lateral rim of its anterior face is marked by attachments for the extensor muscles of the manus. It is not possible to compare directly the distal and the proximal epiphysis, but TRQ933 and TRQ935 appear to be adequately complementary, hence providing a good idea of the complete bone. The distal epiphysis is narrow, probably narrower than the proximal epiphysis, or of approximately equal width. The medial epicondyle is broad and anteroposteriorly flattened. An entepicondylar foramen is present. The olecranon fossa is nearly flat and the coronoid fossa is shallow. In anterior view, the capitatum is much wider than the trochlea. Table 3). The olecranon process is extremely expanded posteriorly and narrow in MNHN.F.TRQ925, much more so than in TRQ926, which probably belongs to a juvenile specimen. The olecranon crest is convex in lateral view. In anterior view, the proximal surface of the process is medially flared, with an anconeal process projected medially. The medial surface of the olecranon is depressed. The coronoid process of the trochlear notch is prominent, while the one of the radial notch is not. The trochlear notch is as tall as it is wide, while the radial notch is taller than wide. The shaft is broken off 3 cm distal to the radial notch; at this level it is reduced in diameter, oval in cross section.
Radius (Fig. 2H, I; Table 3). Only the distal articulation and a part of the shaft are preserved in MNHN.F.TRQ927. The shaft is strongly flattened mediolaterally and expanded anteroposteriorly. The anterior rim of the expanded region is convex, while the posterior rim is slightly concave. At least four grooves are present on the lateral surface for the tendons of extensor muscles of the manus. I interpret these grooves following the anatomical descriptions of Howell (1929) and Piérard (1971). The groove that obliquely crosses the distal region of the lateral side of the radius in a posteroproximal to anterodistal direction is particularly deep in the anterodistal corner. The position and the course of this groove indicate that this is the one for the tendon of the M. abductor pollicis longus (see Piérard 1971: fig. 20; Fig. 2H). A wide, smooth groove located just posterior to it is interpreted as the groove for the M. extensor digitorum communis and/or that for the tendon of the M. extensor carpi radialis; these two tendons are very close one to another but are not always associated with a groove in extant phocids (Piérard 1971). In the proximal part of TRQ927 this groove is visible along the anterior rim of the radius; distally it crosses the groove for the tendon of the M. abductor pollicis longus. A wide and low ridge separates this groove from two distinct grooves that are certainly for the tendons of the M. extensor digitorum lateralis and, more posteriorly, the M. extensor pollicis longus, as in extant phocids (Howell 1929: fig. 10). A small facet for the scapholunar occupies the distal epiphysis; it is concavoconvex and, in distal view, figure-8 shaped. Distally, the posterior side is flattened and occupied by the facet for the ulna, which is elliptical and slightly depressed. Table 4). The two specimens from Küçükçekmece West do not preserve the epiphysis and certainly belong to juvenile or young adult individuals. The absence of a deep trochanteric fossa in MNHN.F.TRQ944  is most probably a juvenile feature, hence has no pertinent taxonomic value. The lesser trochanter is absent. The greater trochanter is large, anteroposteriorly flattened and extended proximal to the femoral head. Its proximomedial part extends posteriorly above the trochanteric fossa (nearly absent/not yet developed? in TRQ944). The neck is short. The shaft is strongly anteroposteriorly flattened and has a lateral and, especially, a medial rim that is particularly thin at mid-length. The intercondyloid fossa is shallow. The main difference between TRQ944 and TRQ945 is that the latter shows a greater medial expansion of the medial epicondyle of the distal epiphysis. As TRQ945 belongs to a larger individual this difference may result from an ontogenetic difference. Indeed Koretsky (2001: 25) mentioned that juveniles tend to have limb bones where the proximal and distal epiphyses are not much different, while in adults the distal epiphysis is "distended considerably" compared to the proximal epiphysis. Table 4). The four distal extremities assigned to Cryptophoca sp. have nearly the same size and proportions, MNHN.F.TRQ917 being slightly larger. They all have similar morphology, except in the length of the preserved portion of the shaft. In its proximalmost part, the shaft is triangular in cross-section, then becomes more quadrangular distally. Distally on the medial face, there is a flattened, elongated triangular area corresponding to the surface of contact with the fibula. The medial side of the posterior face is marked distally by two grooves, interpreted here as those for the tendon of the M. flexor digitorum longus and, more medially, for the tendon of the M. tibialis posterior (Piérard 1971: fig. 32). The grooves are well distinct in TRQ916 and TRQ913, but less so in TRQ917 and TRQ918, where only one groove is clearly identified. Lateral to these grooves, the tibia is convex and slightly bulged. The anterior face is marked by a wide groove, assigned to the one for the tendons of Mm. tibialis anterior and extensor hallucis longus. The anterior distal rim of the tibia extends distally to the level of the medial malleolus. The distal epiphysis is occupied almost entirely by the facet for the astragalus, which is strongly concave. Laterally the articular surface is inclined proximally, contrasting with the medial malleolus, which extends distally.
Astragalus (Fig. 4A, B; Table 4). The bone is poorly preserved. The calcaneal process is broken off and the articular facets are not well visible. The superior facet is proximodistally convex and mediolaterally flat. The neck is long and poorly deviated medially. The head, with an articular facet for the navicular, is flat distally and convex ventrally: the facet extends along the ventrolateral face of the neck. The facets for the calcaneus are not well distinct; the sustentaculum facet is narrow, convex and connected to the distal facet (for the navicular); the ectal facet is concave, wider, and separated from the sustentaculum facet by a shallow groove where a minute foramen is visible. The facet for the fibula occupies the entire lateral face of the body.
coMpArisons And discussion Malik & Nafiz (1933: pl. 13, figs 1-3) figured and briefly described a craniodental material that was much more complete than that available today. Since their study, a fire destroyed all the specimens figured. The material found by Nicolas (1978)  There is a great variability of body size among these late Miocene species. Cryptophoca sp. from Turkey is a rather small species though a part of the material probably belongs to juveniles. Nevertheless I do not consider size as a pertinent character for taxonomic assignment. Pinnipeds generally show a marked sexual size (and morphological) dimorphism, but this is partly due to relative development of soft tissue (muscles, fat). Sexual dimorphism of bones does exist, and deals not only with size but also with the anatomy, which is worth knowing in the present context. In fossil seals, sexual dimorphism is generally not documented due to a lack of pertinent material. However, Koretsky (2001) provides differential measurements and traits for (supposed) males and females of Monachopsis pontica, which is one of the best-known late Miocene species. Praepusa vindobonensis also shows a great size variability (Koretsky 2001). Below I therefore discuss chiefly anatomical differences rather than size differences.

Dentary
Morphologically the hemimandible from Küçükçekmece East closely resembles that of Cryptophoca maeotica illustrated by Koretsky & Ray (1994: fig. 1) and Koretsky (2001: fig. 33): same development and position of the mental protuberance, largesized and single-rooted p1, presence of diastemata between cheek teeth, development and length of the symphysis (straightness of the ventral rim, expansion not posterior to p2 mesial alveolus). The features that may represent autapomorphies for KÇ 101 and C. maeotica are the large size of the single-rooted p1 and the morphology of the anterior part of the dentary (symphysis length, ventral border straight). In these features these taxa differ from the species assigned to Praepusa Kretzoi, 1941, Histriophoca Gill, 1873, Monachopsis Kretzoi, 1941and Pontophoca Kretzoi, 1941 Postcranium Comparison with C. maeotica. The humerus (MNHN.F.TRQ930, TRQ933, TRQ935) also fits the description of the same element in C. maeotica by Koretsky & Ray (1994) and Koretsky (2001): deltopectoral crest strong and long (though not reaching the coronoid fossa), lesser and greater tubercles at the same level and higher than the head, epicondylar crest reaching the distal part of the deltopectoral crest proximally. The fragmentary femora (TRQ944, TRQ945) belong to juveniles. They have a greater trochanter that is much higher than the head and a distal epiphysis that is only slightly wider than the proximal epiphysis, like in though not from eastern Europe) in having a single-rooted p1, a humerus with a longer deltopectoral crest, a more reduced epicondylar crest, a distal epiphysis (probably) narrower than the proximal epiphysis, and a femur with a higher greater trochanter and a distal epiphysis that is proportionally narrower, though exceeding the width of the proximal epiphysis.
The ulnae, radii, tibiae and astragali from Küçükçekmece West cannot be compared with all the late Miocene phocids of Europe due to lack of comparative data; only some of these bones may be compared to one or several taxa (see above the comparisons with Praepusa spp. and M. pontica). They are provisionally assigned to the same species as the dentary, humeri and femora based on their size. Given the difficulty of generic assignment of fragmentary specimens, the material assigned here to Cryptophoca sp. may prove to represent more than one species. description Humerus (Fig. 4C, D; Table 5). MNHN.F.TRQ932 belongs to a much larger specimen than those represented by Cryptophoca sp. The tubercles are developed and slightly higher than the humeral head. The latter is less rounded than in the specimens of Cryptophoca sp. described above, showing dorsoventral compression.
The tibia and fibula are fused proximally (MNHN.F.TRQ946). Both articular facets for the femoral condyles are well de- fined and slightly concave. The lateral condyle is larger and more elliptical than the medial one. Just distally, and rather lateral, to the proximal epiphysis, there is a deep and short tibial fossa on the anterior and posterior faces. The anterior fossa is much shorter than the posterior fossa. The shaft is triangular in cross section; minimal diameters are measured slightly distal to mid-length (minimum mediolateral width is 14.6 mm, anteroposterior thickness is 13 mm at the same level). The distal part of the shaft and the distal epiphysis in TRQ922 and TRQ946 are not different from those of the specimens assigned to Cryptophoca sp. described above. The distal tip of the medial malleolus of TRQ922 is broken off. The grooves observed in the latter are visible in TRQ922, but in TRQ946 a single wide groove is distinct on the posterior face. Approximately 30 mm of the distal fibula are preserved in TRQ946. The shaft is nearly T-shaped, with a posterior face divided by a sharp crest into two subequal parts. The anterior face of the shaft is slightly depressed by a fossa that may be the insertion area of the M. extensor digitalis lateralis, as in extant phocids (Piérard 1971). Relative to the proximal face of the tibia, the proximal face of the fibula is inclined laterally distad.
Mt I (Table 5) MNHN.F.TRQ923 may represent the proximal half of an Mt I. This identification is based on the morphology of the proximal epiphysis, which shows a strong concavity that may correspond to the articulation with the medial cuneiform; a small apophysis extending proximally is also present in the ventrolateral corner. Laterally, the facet for the Mt II is hardly visible.
coMpArisons And discussion The fragmentary humerus MNHN.F.TRQ932 belongs to a much larger specimen than those represented by Cryptophoca sp. Besides size, its morphology differs from that of TRQ930 and 933 by having a deltoid tubercle located more proximally, a lesser tubercle oriented more medially, hence a much wider intertubercular sulcus, and a dorsoventrally compressed head. The latter character is also a characteristic of Monachopsis pontica and of some species of Praepusa (Koretsky 2001).
MNHN.F.TRQ946 and TRQ922 differ from those assigned to Cryptophoca sp. (TRQ913, TRQ916, TRQ917 and TRQ918) mainly by larger size (see Tables 4, 5). Fragmentary tibiae of Cryptophoca sp. are close in size to the tibiae of Praepusa vindobonensis, while TRQ922 and TRQ946 are of the size of Leptophoca lenis True, 1906(Koretsky 2001. In addition, compared to the specimens assigned to Cryptophoca sp., TRQ922 and TRQ946 have a proportionally wider distal epiphysis.
I consider that the observed difference in size and anatomy between the humeri and tibiae of these two species cannot be attributed to sexual dimorphism and/or ontogeny.
Both the anatomical and taxonomic assignments of MNHN.F.TRQ923 are judged to be provisional.  description Mandible (Fig. 5A, B; Table 6) Only an anterior fragment of the dentary is preserved, but it is thick (TMp3 = 23 mm) and robust. The post-canine diastema is long and there is no alveolus for a p1.

Dentition
Lower dentition (Fig. 5A-D; Table 6). The two premolars preserved in MNHN.F.TRQ685 display a moderate, hori-zontal wear pattern, the p2 being less heavily worn than the p3. The p2 is set slightly obliquely in the dentary and relative to the p3 (see Fig. 5A). A major part of the mesial root is visible well above the dorsal rim of the dentary so that in lateral view the tooth seems to be oriented backwards. It bears a strong, but not particularly bulbous, main cuspid; the crown width increases distad; the mesial accessory cuspid is small, slightly lingual, and displays a small wear facet; the distal accessory cuspid is slightly more prominent than the mesial one, centrally located, and is followed by a short cingulid. The p3 is both worn and damaged so that it is not possible to assess the relative development of the accessory cuspids. However, the tooth does not show the bulbous morphology of the p3 of durophagous species such as Adcrocuta spp. On the contrary, p3 appears to be an enlarged version of the p2. The isolated p4 KÇ 58 is moderately worn. It differs from the p2 and p3 by a more slender main cuspid and more developed accessory cuspids, where the mesial cuspid is larger than the distal cuspid. The distal accessory cuspid is followed by a short cingulid. The distal rim of the tooth is curved distally but rectilinear more mesially along the contact with the carnassial (see Fig. 5D).
Upper dentition (Fig. 5E, F; Table 7). The fragmentary right maxilla KÇ 60 preserves the P2. Only the alveoli of the I3, canine, and P1 are preserved. That of the I3 is not complete but indicates a tooth larger than the P1. The canine alveolus is oval, approximately 32 mm in length and 25 mm in width. The single, rounded alveolus of P1 indicates that it was relatively large (alveolus length is 13.9 mm, alveolus width is 12 mm). Very short diastemata separate the P1 from the adjacent teeth. The main cusp and distal accessory cusp of P2 display a moderate, horizontal wear pattern. The tooth is wide relative to its length; it has roughly the same width mesial to the main cusp as distal to that cusp. There is no mesial accessory cusp but a strong cingulum that is particularly prominent mesiolingually. Contrary to the p2, the distal accessory cusp of P2 is lingually located; the distal cingulum is very reduced.

Postcranium (Fig. 6)
A fragmentary ulna is the only postcranial bone from Küçükçekmece that might be assigned to D. senyureki. Unfortunately, the olecranon is not preserved. Overall the shaft is slender and the articular part is narrow compared to that of Indarctos and Machairodus, two genera also present at the site. On the anconeal process, the articular surface for the humerus extends proximally less than in the bears, felids, and hyaenids used in comparison (Ursus sp., Indarctos arctoides, Panthera tigris (Linnaeus, 1758) and P. leo (Linnaeus, 1758), Machairodus aphanistus, Amphimachairodus giganteus (Wagner, 1848), Crocuta crocuta (Erxleben, 1777), Hyaena hyaena (Linnaeus, 1758)), especially on the lateral side. The morphology of the radial notch is typical of that in hyaenids: the lateral extremity of the coronoid process is very short and poorly projected, the articular facet for the articular circumference of the radius forms a very thin surface, and a deep, marked area (presumably for ligament attachment at the proximal radioulnar joint) is present just distal to the radial notch. Overall MNHN.F.TRQ948 looks very like the ulna of roughly contemporaneous hyaenids such as Adcrocuta eximia (e.g., MNHN.F.PIK3247, proximal fragment of right ulna from Pikermi, late Miocene of Greece) and that of extant hyaenas (Crocuta crocuta, Hyaena hyaena) though the shaft of the fossil specimen is clearly more robust than in the extant taxa.
coMpArisons And discussion We follow previous authors by recognizing the family Percrocutidae as distinct from the Hyaenidae Gray, 1821 (see Werdelin 1996, for a historical background). Here I restrict comparisons to the late Miocene middle and large-sized percrocutids, i.e. the species of the genus Dinocrocuta. Sayar (1953: fig. 1) figured KÇ 60 as a maxilla fragment of the sabre-toothed felid Machairodus aphanistus, but the premolar preserved and the size of the alveoli indicate a clear distinction from those of a sabre-toothed cat. This specimen (KÇ 60, Fig. 5E, F) is assigned to Dinocrocuta senyureki based on the overall size of the dentition, the morphology of P2, with its distinctive mesial cingulum, and the presence at the site of additional specimens of the same species. Dinocrocuta senyureki was erected by Ozansoy (1957) 1961,1965). In his publications Ozansoy did not clearly select a type specimen, however. Howell & Petter (1985) regarded TRQ1010 (the only specimen figured by Ozansoy in 1957) as the type specimen of Dinocrocuta senyureki. This specimen is a lectotype, by subsequent designation (Howell & Petter 1985). The material from Küçükçekmece is fragmentary, but the assignment of this material to Dinocrocuta senyureki is well supported by the comparison with specimens from the type locality. I did not find any difference between the lectotype TRQ1010 (Yassiören) and TRQ695 (Küçükçekmece). Both specimens share the absence of p1, the oblique orientation on the dentary and the morphology of p2, the absence of marked bulbous morphology on p3.
In the upper dentition, the distinct mesiolingual cingulum around the crown basis of P2 has the same development in TRQ1007 from Yassiören and KÇ 60 from Küçükçekmece. This is also true for the p4 KÇ 58 from the latter site, which shows cuspids and cingulum as developed as in the p4 of the specimens from Yassiören listed above. Additional specimens from the Sinap were collected more recently and described by Viranta & Werdelin (2003). This sample comes from Loc. 108 (10 Ma in age; Kappelman et al. 2003) and Loc. 12 (9.6 Ma in age; Kappelman et al. 2003), both in the MN9 (early Vallesian) biostratigraphic level and from Loc. 37, MN10 (late Vallesian). These localities yielded a few postcranial bones and many dental remains (Tables 6,  7). Among this new sample from Yassiören the elements that also are known from Küçükçekmece are morphologically very similar. In addition to the type locality, D. senyureki has been described from the Turkish sites of Eşme Akçaköy (MN9, early Vallesian), Kayadibi (MN11, early Turolian) and Inönü (MN10-11, late Vallesian-early Turolian), which yielded dental elements that are not comparable with those from Küçükçekmece (Schmidt-Kittler 1976). Elsewhere, the species is also known at Sahabi (late Miocene, Libya) from right and left hemimandibles of a single individual, an isolated P2 (Howell 1987: fig. 4), and several postcranial specimens. There are differences in proportions or size of teeth or between teeth (Table 6), but the intraspecific variability of the sample from Yassiören alone indicates great variability in size of D. senyureki. The P2 from Küçükçekmece is larger than in the other specimens of the species, but this is the sole difference and the known sample is limited. Therefore the material from Küçükçekmece is assigned to the same species. Other species of Dinocrocuta are known from late Miocene sites in the Old World, especially D. minor (Ozansoy, 1965), D. salonicae (Andrews, 1918, although generic assignment of this species is still debated (Koufos 1995;Zhang 2005), D. gigantea (Schlosser, 1903) and D. algeriensis (Arambourg, 1959). The morphological distinction between the species of Dinocrocuta, which has been discussed elsewhere (e.g., Howell & Petter 1985;Spassov & Koufos 2002;Zhang 2005), is not easy. The main reason to assign the material from Küçükçekmece to D. senyureki rather than to D. minor is the size, with D. minor being much smaller based on the holotype (by monotypy) and probably single known specimen of this species (but see Viranta & Werdelin 2003: 182), MNHN.F.TRQ1011 (fragment of left hemimandible with i1-m1 = Yas 58 in Howell & Petter 1985;Ozansoy 1965: pl. 4, fig. 1). The p2 of the holotype of D. minor is narrower distally than in D. senyureki, but it is hard to tell whether this difference is significant considering the limited samples. Except size, one of the diagnostic features of D. senyureki is the presence on P3 of a strong mesial accessory cusp, but this tooth is not preserved in the material from Küçükçekmece. According to Zhang (2005), compared to D. senyureki, D. gigantea is much larger and has proportionally larger p2/P2. The specimens from Küçükçekmece fit this picture and show a more reduced p2 compared to p3 than in most specimens of D. gigantea. Dinocrocuta algeriensis is based on a sample from Oued el Hamman (= Bou Hanifia; late Miocene, Algeria; Arambourg 1959: fig. 8A, 9D, 10C, 11, 12D, pl. 2, pl. 3, fig. 1-4) representing at least three individuals, two (young) adults (the holotype MNHN.F-1951.9-174, the paratype MNHN. F-1951.9-172, MNHN.F-1951.9-29, MNHN.F-1951.9-75, MNHN.F-1951 and one juvenile (MNHN.F-1951.9-19 and a couple of isolated, unnumbered teeth). An isolated P3 (MNHN.F.AMA9) from Menacer (= Marceau in Arambourg 1959) is also assigned to this species. Like D. gigantea, Dinocrocuta algeriensis differs from our material by a enlarged p2, especially relative to p3, and a thicker P2.

description
The tooth (Fig. 7A-C) is complete except for the distal root. The paraconid and protoconid blades and the hypoconid tip show slight wear. The crown is low relative to its length. The paraconid blade is horizontal and similar in length to the protoconid one; the paraconid is lower than the protoconid. The metaconid is strongly reduced and lower than the paraconid; in occlusal view, it is slightly distal to the protoconid. The talonid is short (approximately 20% of m1 length; Table 8) and narrow, and displays only two cuspids of approximately equal height, although the hypoconid (a little worn here) probably was taller than the entoconid. The hypoconid crest is individualized and well distinct from the basal cingulid; mesially it starts between the metaconid and protoconid and becomes taller and cuspid-like distally. Mesially, the entoconid crest is very low and emerges in the middle of the metaconid distal side. It then runs obliquely and distally forms a small cuspid that is less individualized than the hypoconid. The distal margin of the talonid is nearly transversely oriented and no cuspid is present: the hypoconulid is clearly absent/vestigial in this tooth. A developed cingulid is present on the labial side of the crown and extends slightly on the lingual side of the paraconid.
cf. Thalassictis sp.    1991: 51). The latter is not informative as the metaconid is present in all the species considered here and in the tooth from Küçükçekmece. The number of talonid cuspids is more diagnostic as the latter specimen has only two cuspids in the talonid, the hypoconid and the entoconid; the hypoconulid is absent/vestigial. All the type species listed above retain the primitive feature, i.e. three cuspids in the talonid of m1 (see also Semenov 2008: fig. 6; analysis of Tseng & Wang 2007: Appendix 2, character 6). However two genera were not included in the cladistic analysis of Werdelin & Solounias (1991): Miohyaenotherium and Thalassictis. These authors placed the species of these genera in a clade comprising Hyaenotherium, Hyaenictitherium and more derived genera (e.g., hunting hyaenas and bone-cracking taxa; Werdelin & Solounias 1991: fig. 38), considering, however, that the type and only species of Miohyaenotherium, M. bessarabicum Semenov, 1989, is a possible synonym of Hyaenictitherium parvum (Khomenko, 1914). In a recent analysis, Tseng & Wang (2007) considered the two latter species as synonyms of H. wongii. The close relationship of M. bessarabicum with Hyaenictitherium spp. and/ or Hyaenotherium wongii suggest that it is distinct from the species from Küçükçekmece. The genus Thalassictis includes the type species Thalassictis robusta Gervais, 1850 ex Nordmann MS and according to Werdelin & Solounias (1991) perhaps also: T. certa (Major, 1903), T. sarmatica (Pavlow, 1908), T. proava (Pilgrim, 1910) (= T. chinjiensis (Pilgrim, 1932); see Werdelin & Solounias 1991: 35), T. montadai (Villalta & Crusafont, 1943) and T. spelaea (Semenov, 1988). Thalassictis sarmatica is based on a skull and not comparable to the material from Küçükçekmece. The m1 of the species of Thalassictis is described as having a protoconid taller than the paraconid, a reduced metaconid, and a short and low talonid with two or three cuspids. Generally, late Orleanian-Astaracian  The m1 from Küçükçekmece resembles Thalassictis species in having a protoconid taller than the paraconid, a reduced metaconid, and a short and low talonid with two cuspids. In fact in the number and relative development of the talonid cuspids, it differs from T. proava and T. certa and resembles late Miocene specimens assigned to T. montadai and T. spelaea. The m1 from Turkey differs from species of Thalassictis in having a longer paraconid relative to the protoconid (measured parallel to the alveolar margin) and a shorter talonid. Pending discovery of more material, I maintain the taxonomic assignment open. Malik & Nafiz (1933: pl. 12, fig. 5) assigned to Ictitherium sp. an incisor and a p4 (L: 19 mm; w: 11 mm) that could belong the same species. This is reported in the published faunal lists of Küçükçekmece (e.g., Yalçınlar 1954;Ozansoy 1957). These specimens were destroyed by fire during the last century. Nicolas (1978) also listed Ictitherium sp. cf. I. orbignyi, which is now Plioviverrops orbignyi, but this species is clearly distinct from cf. Thalassictis sp. described here. In Nicolas' collection stored in the MNHN, there is no evidence supporting the presence of P. orbignyi or a closely related species. description Dentition Lower dentition (Fig. 7D-G). The p4 (Fig. 7D-E) is very well preserved and nearly unworn. In occlusal view the crown forms roughly an elongated rectangular. The main cuspid is sharp, with a convex distal crest and a nearly rectilinear mesial crest. In labial view it is oriented somewhat distally relative to the crown-root junction. The accessory cuspids are large and sharp, the mesial cuspid being slightly larger than the distal cuspid. Both are separated from the main cuspid by deep notches. The distal cingulid is marked but short and extended lingually rather than labially. The crown of m1 (MNHN.F.TRQ1213, Fig. 7F-G; Çukurçesme unnumbered) and the mesial root (Çukurçesme unnumbered) are preserved. In both specimens it is heavily worn labially along the paraconid and protoconid crests at the level of the carnassial notch and on the distolabial face of the protoconid. The protoconid is taller and longer than the paraconid. In TRQ1213 there is no metaconid distinct from the talonid. The latter appears very short and sharp, and separated from the distal rim of the protoconid by a deep notch.

T. robusta
Upper dentition (Fig. 7H-J). The P4 presents a similar size and wear pattern as the p4 MNHN.F.TRQ1214 and could belong the same individual. The paracone tip and the protocone are broken off. The labial cusps are narrow and trenchant. The metastyle/metacone blade and the paracone are approximately the same length. The notches that separate the metacone, paracone and parastyle are deep. The parastyle is very large and tall. There is a small ectoparastyle (or ectostyle) mesial to the parastyle. Postcranium 8) Three postcranial specimens are assigned to M. aphanistus. IU 01 is a fragment of left humerus broken across the middiaphysis (Fig. 8). The shaft and the olecranon fossa are also damaged. The maximum width of the distal epiphysis is 88.3 mm and the maximum width of the articulation is 58.6 mm.
The anteroposterior diameter of the medial part is 44.4 mm, while that of the lateral part is 49.5 mm. The overall morphology of this specimen is similar to that of specimens assigned to Amphimachairodus giganteus (Wagner, 1848) fig. 10). Our specimen is a little bit more robust, especially compared to the specimen from Pikermi figured by Roussiakis (2002: fig. 12-1). The medial epicondyle is medially and proximodistally more extended than in the specimen from Montredon (M. cf. aphanistus), but not especially more than in the compared specimens of A. giganteus that display great morphological variability regarding this feature.
MNHN.F.TRQ700 (Fig. 7M, N) is a proximal fragment of a proximal phalanx with a proximal mediolateral width of 21.5 mm and a proximal dorsoventral height of 17 mm. The phalanx is robust, with a dorsoventrally compressed shaft and a slightly asymmetrical proximal articulation. It is very similar to the proximal phalanges of the associated forelimb of A. giganteus MNHN.F.PIK3241 from Pikermi (see Table 11). The fragmentary Mt III (Fig. 7K, L)  description Mandible (Fig. 9A, B; Table 12) The dentary is broken across the middle of the masseteric fossa. The postcanine diastema is long (7 mm) and dorsally concave. There are two mental foramina, one large at the mid-length of the diastema and a smaller one beneath p3. The anterior rim of the masseteric fossa is at the level of the distal border of the m1.
Dentition (Fig. 9A, B; Table 12) There is no premolar between the canine and p3. The p3 is short and low compared to p4. It has no mesial cuspid and a small distal accessory cuspid. The distal cingulid is long and extended lingually and labially. The tooth is much wider distally than mesially. There is a tiny diastema between p3 and p4. The p4 is taller than the m1 paraconid and of the same height as the m1 protoconid. The main cuspid has a straight mesial rim and a markedly convex distal rim. Accessory cuspids are large and similar in size. The distal accessory cuspid is located labially in the crown. It is separated from the main cuspid by a deep notch and its distal rim is vertical. The distal cingulid is more expanded than in p3. The carnassial is elongated and low compared to p4. Its paraconid is lower than the protoconid and approximately of the same length. The blade of these two cuspids forms an obtuse angle in labial view. The distal face of the protoconid is vertical. The talonid is short and the metaconid is not individualized. The derived characters that support the assignment of the hemimandible from Küçükçekmece to Pristifelis sp. cf. P. attica are the absence of premolars anterior to p3 and a distally wide p3 lacking a mesial cuspid. The first character distinguishes the specimen from Küçükçekmece from Styriofelis vallesiensis (and Peigné S. from the other species assigned to Styriofelis). The development of the mesial cuspid on the p3 is highly variable, especially in P. attica (Roussiakis 2002;Salesa et al. 2012b), but in KÇ 61 this tooth lacks a mesial cuspid and is wider distally than in the contemporaneous S. vallesiensis. The Turkish specimen differs from the Turolian P. attica (i.e. specimens included in the comparison: see Table 12) in having a less reduced talonid and in being smaller in size; it is, however, from a site that is several million years older.

description
The phalanx (Fig. 9C, D) is relatively long (Table 13). It is relatively slender with a dorsoventrally compressed proximal articulation and a distal epiphysis that is narrower than the proximal one (Table 13).
coMpArisons And discussion The length of MNHN.F.TRQ701 is much greater than the proximal phalanges of the known small felines from the late Miocene of Europe. On the other hand this phalanx fits with the size range of the proximal phalanges of species such as Promeganteron ogygia (Salesa 2002) Malik & Nafiz (1933) and Yalçınlar (1954) and Machairodus (Paramachairodus) orientalis by Ozansoy (1957). One of the specimens assigned to this species was figured by Malik & Nafiz (1933: pl. 12, fig . 7); the only descriptive information is the length of P3 (12 mm) and of the diastema between the canine and P3 (5 mm). Unfortunately this specimen was not recovered and was certainly destroyed by fire (see Sen 2016). MNHN.F.TRQ701 described here and the specimen figured by Malik & Nafiz (see above) both support the presence of a third, medium-sized felid at Küçükçekmece. Both the size and the morphology of this felid exclude an assignment to either Mach-airodus aphanistus or Pristifelis sp. cf. P. attica already identified from the site. In contrast, the size of the phalanx TRQ701 and/ or that of the P3 figured by Malik & Nafiz (1933) fits well with that of the same elements in late Miocene felids such as Paramachairodus orientalis, Promegantereon ogygia, Metailurus parvulus and 'Felis pamiri' Ozansoy, 1965 from Yassiören. Compared to these species, however, the P3 of the specimen figured by Malik & Nafiz (1933) is taller relative to its length. The phalanx is not diagnostic at the genus level.

discussion
The 1942 fire at the Geological Institute of Istanbul University destroyed most of the fossil specimens described by Malik & Nafiz (1933). Among Carnivora, this occurred for specimens assigned to Mustela pentelici and a (more or less important) part of the material assigned to the Phocidae, Indarctos arctoides, cf. Thalassictis sp. and Felidae gen. et sp. indet. medium size.
Nevertheless my study of the few specimens still stored at the ITU and IU and at the MNHN allows identification of 9 taxa of Carnivora at Küçükçekmece: the ursid Indarctos arctoides, the mustelid Sivaonyx hessicus, the phocids Cryptophoca sp. and Phocidae gen. et sp. indet., the percrocutid Dinocrocuta senyureki, the hyaenid cf. Thalassictis sp. and the felids Machairodus aphanistus, Pristifelis sp. cf. P. attica, and Felidae gen. et sp. indet. medium size. It is highly probable that a tenth species close to Mustela pentelici was present, but it is known today only from illustrations (Malik & Nafiz 1933: pl. 12, fig. 3). As a consequence, my work was mostly based on fossils found by Nicolas (1978) and stored at the MNHN. It is not surprising to find similarities between my faunal list and that which Nicolas proposed, though he did not provide any description nor illustration to support his assignments. Nicolas (1978) mentioned Sivaonyx hessicus, Percrocuta senyureki, Indarctos arctoides and Machairodus aphanistus. Except for the generic assignment of the percrocutid, I confirm the presence of these species (but Sivaonyx hessicus is based, in my study, on a single specimen stored at the ITU, not at the MNHN). However, contrasting with Nicolas' faunal list, I did not find any evidence supporting the presence of Lutra pontica, Ictitherium sp. cf. I. orbignyi, Felis sp. aff. F. prisca and indeterminate viverrids at Küçükçekmece. The faunal list of Nicolas also included two phocids, Cryptophoca maeotica and Praepusa vindobonensis (= Phoca vindobonensis). The material stored at the MNHN is probably not diagnostic at the species level, but I agree that it represents at least two species. One is probably close to Cryp- 27.2-33.7 prox ML W 11.7 11.7-14.55 11.7-19.66 9.9-11.5 9.6-12.9 prox AP T 9.1 9.32-11.29 9.41-11.91 8.1-9.2 8.7-9.9 dist ML W 8.7 9.28-10.95 9-11.3 7.2-9 7.9-9.9