Large mammals of Fouvent-Saint-Andoche (Haute-Saône, France): a glimpse into a Late Pleistocene hyena den

ABSTRACT The paleontological site of Fouvent-Saint-Andoche (Fouvent-le-Bas, Haute-Saône, France) is a large bone accumulation in a karstic context known since the early nineteenth century. This article focuses on the analysis of 14977 remains from both the historical collection (excavations in 1842) and the material of the recent excavations (1989–1992). A faunal list of 18/19 large mammal species has been established: 12 carnivores (Crocuta crocuta spelaea (Goldfüss, 1832), Panthera (Leo) spelaea (Goldfüss, 1810), Ursus spelaeus Rosenmüller & Heinroth, 1794, Canis lupus Linnaeus, 1758, Vulpes vulpes Linnaeus, 1758, cf. Alopex, Gulo gulo Linnaeus, 1758, Meles meles Linnaeus, 1758, Martes sp., Mustela eversmanii Lesson, 1827, Mustela sp.) and seven/eight ungulates (Mammuthus primigenius Blumenbach, 1799, Coelodonta antiquitatis (Blumenbach, 1799), Megaloceros giganteus Blumenbach, 1799, Cervus elaphus Linnaeus, 1758, Rangifer tarandus (Linnaeus, 1758), Equus germanicus Nehring, 1884, Bos primigenius Bojanus, 1827 and/or Bison priscus Bojanus, 1827). The palaeontological analysis allows us to characterize the evolutionary stage related to each species and to comment their biochronological significance. The detailed study of particular species, such as C. c. spelaea, E. germanicus, or C. antiquitatis improves our knowledge on species associations and their implication in Late Pleistocene hyena dens.

Within these deposits with both lithic industries and faunal remains, the cave hyena drew particular attention. Its role in bone accumulations (e.g., den sensu stricto) as well as its impact on bone stock (e.g., aven, anthropic layers) had been discussed at length (see Fosse 1995 for a general review). First studies already showed that hyenas were bone collectors as well as consumers (e.g., Cuvier 1812; Buckland 1822; Bertrand de Doue 1828; Tournal 1833; Regnault 1885; Harlé 1892, 1899). The description of the bone accumulation at Kirkdale Cave (England) by Buckland (1822) is certainly the first characterization of hyena den. The diversity of carnivores, the overrepresentation of juvenile and senile ungulates (attritional mortality), the abundance of their antlers and acropods, the teeth marks, the number of individuals (many juveniles), and the presence of numerous coprolites permitted all together to give a precise definition of a den. Fouvent was also of paleontological interest (Cuvier 1825; Gervais 1870) and just after the work of Buckland at Kirkdale, this site allowed for precise determination of carnivore sites (Thirria 1828(Thirria , 1833Nodot 1858Nodot -1859. More recently, the renewed attention in the predator habits and hyena dens was stimulated by the development of neotaphonomy (Binford 1981;Brain 1981) to understand hu-man and carnivore interactions with their prey (Fosse 1994). In this context, new excavations were launched in Fouvent during the late 1980s and early 1990s. Thanks to this new material, we present here an up-to-date paleontological study of Fouvent, notably in a palaeoenvironmental perspective.

Location and geoLogicaL context
The palaeontological site of Fouvent, also named Abri Cuvier, is located in the karstic region of Fouvent-Saint-Andoche, in the Northwest of Haute-Saône department, precisely in the village of Fouvent-le-Bas, near Le Vannon River, at an estimated height of 200 meters above sea level (Fig. 1). The site is located in a karst area in Bathonian limestone diaclases (Detrey 1992). The cave has now disappeared and the limestone blocks were used to build a house. However, Fouvent has been described in previous publications by the time of its discovery. These descriptions indicated a cavity of moderate size: the cave stretched 10 m long, 4 m wide, and a height of about 2 m and it had two entrances. The main entrance was about 1 m 2 wide and the second was slightly smaller (Nodot 1858(Nodot -1859. New excavations in the late 1980s revealed a passageway probably belonging to the original cave (Detrey 1992;Fosse 1997).
HistoricaL context: previous researcHes The bone accumulations of Fouvent have a famous and long history starting from the beginning of the nineteenth century until recent excavations in 1992 (Fourvel 2012;Fourvel et al. 2014). The site highlighted changes in the way to study and to interpret bone layers through time (see synthesis in Fosse 1995). Since early works, Fouvent played a key role in Quaternary palaeontology: osteological distinction between Large mammals of Fouvent-Saint-Andoche current and fossil hyenas (Cuvier 1825), first reference of fossil wolverine in France (Gervais 1870) and first definition of a Pleistocene carnivore den (Thirria 1828(Thirria , 1833Nodot 1858Nodot -1859. Nodot (1858)(1859) advanced several arguments in favour of bone accumulation by cave hyenas. He highlighted the correlation between tooth marks, morphology of bone fractures and consumption of bones by carnivores. The first synthesis carried out on Fouvent was published by Bouillerot (1881). In comparison with observations and interpretations given at Kirkdale Cave (Buckland 1822), Bouillerot concluded that Fouvent was undoubtedly a hyena den cave.
The most recent excavations took place at the end of the 1990s. This area has been systematically excavated under the direction of J. Detrey between 1989and 1992(Detrey 1992Fosse 1997). A very abundant faunal material and a small lithic assemblage composed of nearly 300 elements (cores, tools, and splinters) were reported to Châtelperronian or Mousterian facies (Fosse 1997). To date, only a small part of the "ancient" cavity has been excavated and studied.

stratigrapHy
The recent excavations (dir. Detrey) have been focused on a 20 m 2 surface. These fieldworks allowed recovering an abundant and species-rich paleontological material, in-situ within a cryoclastic sedimentary matrix (typical for cold intervals).
Four main stratigraphic levels have been recognized, based on the observation and description of the sediment (Fig. 2).
The upper levels (E) are characterized by silty clay layers. The sedimentological analysis concludes to a deposit by runoff water and/or colluvial phenomenon (poor vegetal cover). Levels A and B are largely composed of rocks and cryoclastic gravels. The different grain sizes of the cryoclastic gravel as well as the collapsed blocks in the matrix permit to make a subdivision of the layer B into two levels: B1 and B2. The layer C is a silty clay horizon interbedded with fine cryoclastic gravel with angular elements. The major part of the sequence indicates relatively cold climatic conditions (Morin in Detrey 1992). A betterment of climatic conditions (temperature and humidity increase) throughout the documented sequence is suggested by the identification of silt levels which alternate with cryoclastic material and trace of gelifraction.

MATERIAL AND METHODS
Our study includes both the 1842 sample corresponding to Dubois' excavations during the 1800s, first published by Nodot (1858Nodot ( -1859 and revised by Lovis (1968), and the bone material unearthed during modern excavations, between 1989excavations, between -1992. We have observed a total amount of 14977 bone remains (both identifiable and unidentifiable) distributed as follows: 145 specimens in the 1842 sample and 14832 specimens in the 1989-1992 sample. The identifiable specimens (3347 NISP) have been referred to 27 taxa, at species, genus or family level, including 12 carnivores and 7/8 ungulates (Table 1).
The palaeontological analysis, focused on the identification of the faunal spectrum, aims at characterizing paleoenvironmental parameters, mostly based on osteometrical variables. The measurement protocol follows Von den Driesch (1976). However, the detailed analysis of peculiar species (e.g., hyena or horse) implies using specific measurements. Even if we could find a wide array of published data for these species, it appears that each author uses his   -    -47 -177 -1544 -242 -4883 -240 -2348 -2148 -11629  Large mammals of Fouvent-Saint-Andoche own measurement protocol, especially for hyenas (e.g., Ehrenberg 1938Ehrenberg -1940Clot 1980;Argant 1988;Dufour 1989;Cardoso 1994;Ambros 1998;Turner 2001;Baryshnikov & Tsoukala 2010). Accordingly, we selected several measurements that we considered as relevant for this study.
The results are expressed in millimeters, in tables including the number (n) of measured specimen, the smallest (min) and largest (max) dimensions, the Mean and the standarddeviation (SD). Three main quantification units are used in this study: the number of identified specimens (NISP), the minimum number of elements (MNE) and the minimum number of individuals (MNI). For horses, confidence interval for the mean (95%) is also provided. Test and specific routines are calculated using both softwares R© (versions 2.13.2 and 2.14.0) and XLStat© (Version 11.4.07). We also used Kruskal-Wallis test of the one-way analysis of variance by rank (Kruskal & Wallis 1952). This method can be considered as the non-parametric equivalent of the ANOVA. It is used to compare more than two samples of different sizes and makes no assumptions about the shape of the distributions. The observation of 1110 carnivore remains allows us to recognize 11 distinct taxa (either at species or genus level) related to Felidae, Hyaenidae, Canidae, Ursidae, and Mustelidae. The cave hyena is the main species, as it is represented by 850 bone and tooth remains (Fig. 3). Other carnivore species are quite rare: less than 50 elements have been observed for each species, excluding the small canids (Fig. 4) description Among the small canids remains (n=76), some of them have been associated to the red fox Vulpes vulpes (n=13) corresponding to five adults. The low number of elements determined at species level is related both to the state of conservation of the material and to the difficulties to distinguish Vulpes and Alopex Kaup, 1829. The red fox is a small common carnivore in Pleistocene bone assemblages (e.g., Gönnersdorf in Poplin 1976;Gerde in Clot 1980). This predator is ubiquitous and well-represented during different Pleistocene glacial and interglacial climatic phases. In France, the red fox is found at Nauterie II (layer 11: Mindel-Riss), at Montmaurin-la-Niche, Nestier, and Gerde (Riss); it then cooccurs with the arctic fox Alopex lagopus (Linnaeus, 1758) throughout Würmian times, before it becomes a common carnivore during Holocene times (Clot & Duranthon 1990). Morphometrical similarities and potential co-occurrence of red and arctic foxes during the Oxygen Isotopic Stage 3 (OIS3; Cohen & Gibbard 2011) necessitate performing a careful analysis of the bone material. At Fouvent, cranial and dental elements are represented by a left mandibular branch with p2-m2, a fragment of left mandible with m1 and m2, and a broken right mandible with m1 and m2 with two canines. As regards the red fox, according to Poplin (1976), the lower canines are longer and thinner in comparison with isatis. At Gerde, Clot (1980) also pointed out the great difficulty to make the distinction between Vulpes and Alopex but determined the canines of fox based on their general proportions. We attributed the material of Fouvent to the red fox on the base of the concerned morphometrical data. Our three m1s (G8.B.509b, D11.E.3 and G8.B.507) leave no doubt for their specific assignment. Their large dimensions (respectively B = 6.4-6-6 with L = 16-16-16) match perfectly those of red foxes as provided at Gerde (Clot 1980: table 49) and at Gönnersdorf (Poplin 1976: 48 fig. 28). Postcranial elements of red fox include two humeri, one ulna, one metacarpal II, two tibiae, one calcaneus and one metatarsal V. The measurements are summarized in Table 2. According to Altuna (2004), the breadth of the distal humerus of the red fox ranges between 17.2 and 23.8 mm while Clot (1980) gives an interval comprised between 19.2 and 22 mm. According to these studies, the distal breadth of isatis is not greater than 19.5 mm, consequently the humerus of Fouvent (1842.118) is quite similar in size to the red fox. Published metric data available for the ulna are rare but the dimensions of the specimen of Fouvent suggest without no doubt an attribution to Vulpes.
The total length (47 mm) of the left metacarpal II (G8.A/B.165) of Fouvent is larger in comparison to the measurements of red fox (Altuna 2004). However, the left metacarpal falls into the variation range as provided by Clot (1980) for Pleistocene populations of Vulpes. The same is true for the measurements of both a proximal and a distal tibiae (respectively n°1842.119 and 1842.92) that we also assigned to the red fox. In the Table 3, the maximum length (56 mm) of the left metatarsal V (F9.B.135) is roughly similar to the mean value observed in living foxes (Altuna 2004) corresponding to their smaller pleistocene representatives (Clot 1980). We observed the same thing with the maximum length of the calcaneus G8.E.26 (only 29 mm) that could be associated to a small red fox, while populations of isatis only reach a maximum of 28.4 mm.
cf. Alopex sp. description A small right metatarsal V (Fvt 90-92 G8.B.288) could be related to the polar fox. Osteometrical comparison of this element with other modern or Würmian populations of both red and polar fox allows us to refer the concerned remain to as Alopex (Table 3). The metatarsal length (GL=51 mm) matches the larger specimens recorded in Pleistocene samples (Clot 1980;Baryshnikov 2006) and modern populations (Altuna 2004). However, the proximal breadth (Bp) is large (8 mm) and could coincide with measurements of the red fox. But, in contrast, metatarsals V of Vulpes with a length lower than 55 mm seems to be uncommon. This kind of small-sized red fox has been only recorded in extant populations (Altuna 2004). However, the recorded dimensions overlap those of both Vulpes and Alopex.
As such, the metatarsal V from Fouvent could either belong to a very small red fox or to a large polar fox. The geochronological context of Fouvent and the proportions/dimensions of the concerned bone are compatible with the presence of A. lagopus. Accordingly, we prefer to let this specimen in open nomenclature, referring it to as cf. Alopex sp. Seven remains brought to light some additional morphometrical features of the knowledge of this species (Table 4). For example the P3 (F11.B.223) shows a fairly massive crown which is made of a single antero-posterior cusp crossed by a protruding edge. Moreover, a cingulum is strongly developed  at the base of the lingual side. The dimensions of this tooth compared to large recent and fossil mustelids (Wolverine and Badger) permit us to identify it as G. gulo (Fig. 5). Furthermore, morphometrical data allow us to exclude the Badger Meles meles Linnaeus, 1758. Finally the specimen of Fouvent has relatively small dimensions compared to the Pleistocene gluttons but fits well among the greatest living representatives of G. gulo.

description
The bone sample from the recent excavations includes several remains attributed to small Mustelidae, among which a pine marten or a beech marten (NISP=2) and a smaller one related to genus Mustela (NISP=8). Two mandibles of greater size (related to a single individual) has been observed  Large mammals of Fouvent-Saint-Andoche in detail. Based on their general characters and size, the specific determination of these pieces could only be related to a polecat (European polecat Mustela putorius or steppe polecat M. eversmanii). These species at Fouvent do not imply the same environmental conditions for the deposit. Morphologically, the mandible of steppe polecat is distinguished from the common polecat by a bulge of the horizontal branch under the m1 just below the paraconid (Hugueney 1975;Delpech 1989). The mandible F11.B.214 (Fig.4G, H) clearly shows this morphological feature. In addition, the dimensions of the two hemi-mandibles as well as the carnassials (L/B of m1 respectively 10/4 and 9.3/3.7 mm) suggest large-sized animals, similar to those of the steppe polecat (Table 5). Accordingly, everything concurs unambiguously to assign those remains to Mustela eversmanii. description Bear remains are far from being well represented in the Fouvent assemblage. Seventeen elements, referred to nine individuals (adult and old bears) have been attributed to Ursidae. The original work of Lovis (1968), based on the 1842 sample, reports the presence of two different species: the brown bear Ursus arctos and the cave bear U. spelaeus. Re-examination of both this material and bone remains recovered from the 1989-1992 excavations allows identifying a single species, U. spelaeus. Morphometric data (Table 6) and morphological features (e.g., general shape and size of raw teeth, high development of numerous cuspids) of unworn teeth (1842.66 left maxilla including M1 and M2; G9.B.558 right M2 (Fig.4J, K); F10.C.263 left m3) relate unambiguously this material to the cave bear. The wide size range as recorded for a given dental locus (e.g., length ranging between 25.3 and 31 mm on M1 and between 40.5 and 48 mm on M2) might be due to either intraspecific variability or sexual dimorphism (marked in recent ursids). Such variability is often recorded for tooth length: Spahni (1954)  description A fragment of a right upper canine is reported to a Pantherine (F9.C.226). The general size of the fragment suggest a large size species like leopard Panthera pardus or cave lion Panthera (Leo) spelaea but the intense surface modification (dissolution) and the difficulty to take any measurements do not allow us to make precise attribution. Consequently this speciment is related to Panthera sp.   description Thirteen cranio-dental remains (upper and lower teeth, mandible) have been attributed to the cave lion. This material and its stratigraphical distribution within the locality suggest a minimum number of seven individuals. Teeth measurements reveal significant size variability ( The taxonomic status of the small morph is a matter of debate, given that such size discrepancy may either reflect ecomorphotypy or sexual dimorphism. At Fouvent, the ratio B/L of the P4 and the m1 compared with fossil and living populations leads to some comments on the size of the different clines (Fig. 6). Thus, on the base of the m1 of Jaurens, a clear distinction appears between large-sized lions (Jaurens in Ballesio 1980) and a smaller form (Jaurens in Ballesio 1980;Espèche in Clot et al. 1984). In addition, many osteometrical datasets for Late Pleistocene cave lions confirm significant variability for the m1 (specimens smaller than P. spelaea var. cloueti and also larger than the biggest form of Jaurens). Moreover, current data confirm the presence of a strongly marked sexual dimorphism, increasing the probability of significant overlap between osteometrical dimensions. In our opinion, the different sizes observed at Jaurens seem to be more related to intraspecific sexual dimorphism than to any evolutionary stage or stratigraphical age-based discrepancy. Consequently, if we consider the small form of Jaurens as characteristic of females, the m1 of Fouvent which are very close in size could belong to females. The same thing is true concerning the two P4s of Fouvent, both presenting extreme values. The larger one could be associated to a large-sized male and the smaller to a female. However, it would be necessary to undertake a thorough revision of cave lion intraspecific variability in order to validate the concerned hypothesis.

description
We could not get access to the complete mammoth collection from Fouvent. However, the palaeontological analysis of the proboscideans was already realized by the late P. Paupe during the 1989-1992 excavations. The present study is largely inspired from his unpublished report. Our own analysis was focused on highly fragmented deciduous teeth (ridges of enamel) and long bone fragments which were not identified anatomically in spite of their large proportions. We have observed directly 222 remains attributed to M. primigenius but were not able to describe thirty-three isolated teeth and two tusk fragments. Based on P. Paupe's observations and our own analysis, eight up to ten individuals are represented in Fouvent. Juveniles are dominant: four individuals died between two and seven years old. Four teeth are referred to prime adults (between 10-20 and 30 years old). Two molars point to the presence of two old individuals, aged of 45 and 50 years, respectively. In our analysis, dental remains are mostly represented by fragments of lamella mostly belonging to very young animals (n=81). This did not allow us to deduce neither their rank, nor their laterality and even less the MNI. Although the preliminary study of P. Paupe is based on a small dental sample (n=11), our observations confirm his taxonomic assignment with certainty. Thus, both the morphological description of teeth as well as the morphometrical indices (e.g., lamellar frequency between 10 to 16, length/height ratio) demonstrate a high evolved degree of the mammoths that are associated without no doubt to M. primigenius (Paupe in Detrey 1992).  1946;Martin 1968;Bonifay 1971;Ballésio 1979;Clot 1980;Guadelli 1987;Dufour 1989;Argant 1991;Cardoso 1993;Blasco & Ramirez 1997;Fosse 1997;Ambros 1998;Baryshnikov 199;Altuna & Mariezkurrena 2000;Testu 2006;Barycka 2008;Fourvel 2008Middle Pleistocene Kurtén 1962, 1972Bonifay 1971;Kurtén & Poulianos, 1977Baryshnikov 1999 ;Testu 2006 ;Baryshnikov & Tsoukala 2010OIS5: Clot 1980Baryshnikov 1999;Fourvel 2012 Fouvent-Saint-Andoche  Lower teeth are under study and will be included later in a more complete analysis with post-cranial elements. For now, 283 upper cheek teeth have been studied (Table 9). In this table, teeth are compared to a reference dataset of Pleistocene equids. For each measurement the number of teeth, minimum and maximum values, mean and confidence interval at 95% and standard deviation are given. The morphometrical distinction between ranks of premolars and molars (e.g., P3-P4 and M1-M2) is usually difficult or impossible with isolated teeth. Accordingly, we have chosen not to distinguish P3 from P4 and M1 from M2 in order to compare the sample of Fouvent to a larger number of cheek teeth from literature.
As a first step, we wanted to know if the cohorts of horses from the different levels of Fouvent could be considered as coming from the same species. Thus, we grouped the stratigraphic units (e.g., Ab, Ba, C2...) in three levels namely A, B and C. Dental remains without stratigraphic location were excluded, as well as few teeth from the levels E and E9. We compared the Protocone Index using the Kruskal-Wallis test of one-way analysis of variance by ranks detailed previously. The analysis clearly indicates that there was no statistically significant difference between levels A, B, and C (α=0.05). Thus, dental material from those levels can be considered as originating from the same demographic/evolutionary unit, which is confirmed by the lack of differences in pairs of Dunn (1964) ( Table 10). The method of Dunn (1964) compares the mean of the ranks, the latter being those used in the calculation of k according a normal asymptotic distribution for the standardized difference of the average of the ranks.
The dimensions of most upper teeth of Fouvent match the variation range of both E. germanicus and E. gallicus (Table 9). This is also the case in the upper part of the sequence of La Quina (Charente, France) which hosted the two species and made impossible their distinction based on their dimensions (Armand 1998). Here, we propose a new quantitative approach using biometric  For each measurement following a normal distribution, the confidence interval on the mean was given using the basic t-test of Student with the software R (2.14.0; t.test function). For the measurements which did not satisfy the conditions of normality, the confidence interval we derived was estimated from the theoretical median (wilcox.test function in R). Results indicate that intervals associated to Würmian species are well individualized from those related to ante-Würmian  Table 12).
To identify the horse of Fouvent at species level through dental measurements, we used the routine package knn.cv from R software (version 2.13.2). The program corresponds to one of the more efficient non-parametric methods for data classification in data mining: the so-called k-Nearest Neighbors (or k-NN for short; see details of the method in Mathieu-Dupas 2010). When there is little or no prior knowledge about the distribution of the original data, the rationale consists of finding among the predefined training samples (e.g., measurements of well known species) the closest distances of the new points that may be assigned to the original data. In this study, we applied the Euclidean distance, which is most commonly used. As an example, for a data point x of Fouvent, we computed the distance between x and all the data points from the training samples, in order to attribute the species determined by the nearest points of x according to k. This number k, usually an odd number, ranks the nearest neighbors from the training data. It determines the species to be assigned on the base of the majority vote using cross validation.
When k is small (e.g., k=1), it improves the power of association even if noise may somewhat affect the results. However, when k increases it is less sensitive to noise and makes the borders of the classes less distinct but necessarily requires large training samples (Mathieu-Dupas 2010). Table 13 shows unambiguously and whatever k is, that the nearest species for the ratio of protocone length to occlusal length of M1M2 [M-pl/M-ol] is always E. germanicus. The same is true for the ratio of the length of protocone to the occlusal average length of P3P4M1M2 [PM-pl/PM-ol], except for an isolated case (k=7) which is associated to Equus antunesi Cardoso & Eisenmann, 1989. Nevertheless, E. steinheimensis appears for the ratio of the length of the protocone to the occlusal length of P3P4 [P-pl/P-ol], as the closest species except for k=9, k=13 and k=14 which are attributed once again with E. germanicus. However, the dental morphology of E. steinheimensis allows undoubtedly excluding such an assignment because caballoid and stenoid characters are not observed at Fouvent, contrary to what occurs at Châtillon-Saint-Jean (Drôme, France; Mourer-Chauviré 1972). At Fouvent, dentition shows styles with splits on the premolars, molars with simple parastyle and mesostyle, concave interstylar surfaces, and bilobed protocones. Even so, the proximity between E. steinheimensis and E. germanicus is not trivial and refers to the hypothesis of a possible phylogenetic relationship, as mentioned by Prat (1968: 520). Finally, our analysis does not either reveal a possible association between the horse of Fouvent and the more evolved E. gallicus.
At the end of this study, it appears that dental morphometry, coupled with high resolution analytical tools can account for evolutionary stages of Pleistocene horses. We have shown that the horse of Fouvent was associated to the typical species E. germanicus but did not yet reach the evolutionary stage as observed in E. gallicus. In conclusion, in an anagenetic perspective, it is quite reasonable to consider that the deposition E. germanicus of Fouvent is probably associated to the time interval from the very end of OIS 4 to the end of OIS 3. description More than 90 isolated teeth (mainly deciduous teeth), and several postcranial elements have been attributed to the woolly rhinoceros C. antiquitatis (Blumenbach, 1799) (Fig.8). The  morphology of this Eurasian Pleistocene species is well-known (for review, see Guérin 1980). The Fouvent sample presents all the morphological features of this species (large size, high crowned teeth, quadrangular M3s with distinct ectoloph and metaloph).
We have used the tooth wear stage-based protocol developed by one of us (POA) for reconstructing mortality curves in both extant and extinct rhinoceroses, based on isolated teeth (Bacon et al. 2008). The age classes as characterized on the white rhinoceros by Hillman-Smith et al. (1986) are used in this study because of the phylogenetic, chronological and ecological close relationships between this extant species and the woolly species (Antoine 2002). We have restricted this analysis to upper teeth, more abundant in Fouvent (69 specimens), as they display a much more homogenous eruption and wear pattern than the lower teeth, both in the white rhino and in the woolly rhino. The mortality curve obtained for C. antiquitatis is trimodal, with a majority of individuals ranging from classes I to IX (1.5 months up to 9 years) (Fig. 10); the first mode is comprised between 1.5 and 4 years (classes V-VI), which coincides with a period encompassing weaning and abandonment of juvenile individuals by the mother (Groves 1972). The second mode includes specimens the individual age of which ranges from 4 and 9 years, e.g., more or less the subadult-adult transition (classes VIII-IX; Hillman-Smith et al. 1986). The third and last mode, with less specimens, includes adult individuals (14-28 years; classes XII-XIII). Based on upper teeth, juveniles (61%) are far overrepresented with respect to subadults (23%) and adults (16%). Such a structure is significantly distinct from that of natural populations of recent Indian rhinos (Rhinoceros unicornis Linnaeus, 1758), consisting of 27% juveniles, of 21% subadults, and of 52% adults (Laurie 1982;Laurie et al. 1983). On the other hand, the Fouvent mortality profile is much similar to what is observed for the late Pleistocene rhino tooth sample of Duoi U'Oi, in Vietnam and other Southeast Asian cave localities, for which the accumulating factor is also of biological origin (porcupines; Bacon et al. 2008;Antoine 2012). Nevertheless, 14 woolly rhino teeth from the 1989-1992 excavation sample in Fouvent are eroded, which attests to a by-pass and a hydrodynamical erosion, both post-mortem and pre-accumulation (n°32, 42, 78, 87, 107, 124, 158, 192, 225, 269, 293, 330, 337, and 480). A lower tooth was split into two pieces then eroded, before both halves were recovered from two distinct excavation units .

k=1
[1] P-pl/P-ol     and/or m3 also show individualized interlobar column (Fig.9A, B). Measurements are summarized in Table 14 and their comparison with published data confirms our attribution to the Megaceros. The basal length and the width of six left P2 are quite similar of the ones recorded at Tournal (23-23 mm for the length and 22-23 for the width in Magniez 2010) and at Conives (22 mm for the length and 21.4 mm for the width in Fourvel 2008

description
The red deer C. elaphus is represented by 30 cranial and postcranial remains which are related to 10 individuals. The dental material is attributed undoubtfully to C. elaphus. Premolar and molar size, proportion, and morphology are consistent with the descriptions of Bouchud (in Lavocat 1966) and easily distinguishable from the reindeer. In addition, their size clearly does not correspond to the roedeer and there is no interlobar column on molars as for the giant-deer. The morphometrical features, in particular on M3, do not match the descriptions of the small form Cervus simplicidens (Guadelli 1996). Unfortunately, with the exception of a left talus (1842.97; Fig. 9D, E), bones are too much fragmented for any biometric analysis. However, diagnostic criteria for the red deer as described by Bouchud (in Lavocat 1966) were observed at Fouvent (e.g., the welldeveloped median gutter of the metapods). The occurrence of a red deer at Fouvent is not surprising, as this ubiquist species has been recognized in Europe from the Middle Pleistocene (e.g., Mosbach, Mauer) until recent times.  description One hundred and thirteen bone remains (both cranial and postcranial elements) have been referred to as the reindeer, R. tarandus. Their stratigraphical location at Fouvent suggests a minimal number of 16 individuals. All the typical reindeer morphological features (see Magniez 2010 for a review) could be observed on the material (including teeth, antler and postcranial elements). Unfortunately, this material is too much altered and the sample is not sufficiently abundant to meet the conditions for any population structure analysis (age-structure, sex-ratio...). Be as it may, a left basilar fragment of shed antler (FVT.1842.106) can be associated to a male adult (Fig. 9M). Another small fragment (FVT.1992.F10.B.329) could match a female or a young male (Averbouh pers. comm.; Fig. 9N). Finally, distinct tooth wear stages as observed on isolated teeth suggest various age classes (young adults, adults, old adults, and seniles). description Roe deer is represented by 11 remains (including one isolated tooth), which correspond to three adult individuals. Appendicular skeleton is not sufficiently abundant and well preserved to allow any taxonomic identification. However, the size and the general morphology of metatarsals are typical of Capreolus (i.e. presence of a narrow longitudinal dorsal gutter; proximal epiphysis distinct from other comparable small ungulate species). The roe deer C. c. suessenbornensis has been attested in Europe (e.g., Süssenborn) since the Middle Pleistocene, and it became abundant with C. c. capreolus during the Cromerian stage. This latter subspecies has been widespread in Europe from the late Pleistocene until today (Kurtén 1968;Delpech & Guérin in Guérin & Patou-Mathis 1996).

DISCUSSION
The bone collection coming from previous excavations of Fouvent led to famous works in the field of palaeontology. As an example, taking into account the material of Gaylenreuth and Fouvent, Georges Cuvier was the first to make the distinction between fossil hyenas and their current representatives, starting from morphological cranial comparisons (Cuvier 1812). In this study, Cuvier made a synthesis of the previous works (six studies involving three collections from the different campaigns of excavations) and described the faunal association. The first species association, based on the material collected by Mr. Le Febvre de Morey and transmitted to Cuvier, reported at least three species: the horse, the elephant and a hyena. The successive works suggested an open steppic environment under a cold climate, with an assemblage encompassing horses, mammoths, woolly rhinoceroses, reindeers, and a wolverine. The faunal association of Fouvent reflects a steppe-tundralike open landscape under a severe cold climate as reported in the preliminary study of Fosse (1997). The association of Dicrostonyx Gloger, 1841 and Lemmus Link, 1795 (lemmings) as well as the presence of marmots is also characteristic of cold climates (Roger unpublished). From a strict biochronological point of view, the species of Fouvent are quite similar to OIS3 faunal assemblages integrating gregarious species (B. priscus, E. germanicus, R. tarandus) in a steppic environment (C. antiquitatis and M. primigenius). The evolutionary stage of ungulates of Fouvent points to the Würmian period and more probably OIS3. Most carnivores of Fouvent are ubiquist species (P. (Leo) spelaea, U. spelaeus, C. lupus, C. c. spelaea, M. meles and V. vulpes). The hyena of Fouvent is robust and quite similar to the large hyenas of OIS3 the body size of which increased during recent phases of the Pleistocene (Kurtén 1963;Klein & Scott 1989). The recognition of the wolverine (G. gulo) and the steppic polecat (M. eversmanii), as well as the probable presence of Isatis, are clear indications of a cool phase.

CONCLUSION
The palaeontological site of Fouvent could be considered as an original example of a Late Pleistocene hyena den (Fourvel 2012;Fourvel et al. 2014). The association of various criteria such as predator abundance, in particular the cave hyena (850NISP, 121MNI), the diversity of ungulate prey-species (ranging from small-sized species such as the roe deer to mega-herbivores, like woolly rhinos and mammoths) and the general context of the bone accumulation (horizontal karstic cavity) suggests a long-termed occupation of the site by the predator, which in turns defines a cave hyena den. The bone accumulation is particularly species-rich in terms of large mammals, encompassing seven or eight ungulate species and eleven carnivore species. These faunal elements point to the latest phase of the Late Pleistocene (Würmian stage). The evolutionary stage of equids (referred to E. germanicus), the abundance of woolly rhinoceroses and reindeers, and the large dimensions of the cave hyena allow us to refer this faunal assemblage to the Isotopic Stage 3. Palaeoenvironmental conditions as inferred by the co-occurrence of horses, mammoths, woolly rhinoceroses, reindeers, and wolverines, coincide with a steppe/tundra-like open landscape under a cold climate, which are perfectly compatible with the Oxygen Isotopic Stage 3 (OIS3; 57-29 ka BP; Fourvel 2012).