Fish – Actinopterygii indet.

The centrum of an amphicoelous vertebra (Fig. 3/1) demonstrates the presence of an actinopterygian fish, and suggests the nearby presence of a permanent water body.

Spiny toad – Bufo cf. spinosus

The Mediterranean common toad is represented by one humerus, two radioulnae and 2 phalanges. The humerus (Fig. 4/1) shows a straight and strong diaphysis with a humeral condyle lacking here but that may have been displaced radially. The other elements display the general morphology of the genus Bufo like the radioulnae with a wide collum antibrachii (Fig. 4/2) and the short and robust phalanges. Cautious attribution among bufonids to B. spinosus can be done based on the size and robustness of the remains: B. spinosus reaches generally a larger size than B. bufo and especially B. calamita. The spiny toad is currently restricted to Iberia and southern France, and inhabits all environments, but with a preference for stable, cooler and woody biotopes (Lizana, 2002 and Masó and Pijoan, 2011).

Indeterminate lizard – Lacertidae indet.

A small and somewhat elongated procoelous trunk vertebra (centrum length: CL = 2.1 mm) is attributed to a lacertid lizard. The interzygapophyseal constriction is very poorly marked. The neural spine is long, thin and low. Posteriorly, it ends as an interzygapophyseal point that does not project back beyond the postzygapophyses. In ventral view, the centrum is cylindrical with a haemal keel more or less wide and with indistinct lateral edges. The cotyle and condyle are slightly flattened dorsoventrally. Vertebrae are not diagnostic to make a separation between the different lacertids (Barahona Quintana, 1996), besides the small size of our fossil preventing an attribution to large lizards as Timon lepidus, Lacerta schreiberi, and Lacerta bilineata. The size of the Valdavara 3 material fits the species Iberolacerta monticola, Podarcis muralis, Podarcis hispanica s.l., and Podarcis bocagei, which live today in the North of the province of Lugo (Masó and Pijoan, 2011).

Slow worm – Anguis fragilis

The two fossil vertebrae are small (CL < 3.6 mm) and procoelous, with a posterior condyle and anterior cotyle that are dorsoventrally flattened. In ventral view, the centrum, which is much longer than wide, has a flat ventral surface and has lateral margins that are parallel. One of the most complete vertebrae (Fig. 4/3) shows the bases of the hemapophyses fused to the centrum on its posterior half, and transverse processes that are well-developed and dorsoventrally flattened. The other fragments correspond to a posterior and an anterior part of a caudal vertebra (very probably the two parts of the same vertebra) divided at the level of the autotomy line. These vertebrae can be distinguished from those of other anguines (Dopasia and Pseudopus) mainly by their smaller size (CL < 5 mm), but also by their relatively elongated shape and the lateral margins of the centrum, which are more or less parallel. The fossil vertebrae do not display any morphological difference from those of the present-day species A. fragilis.

The slow worm is a Eurasiatic species with a wide European distribution, with the exception of Ireland, a large part of Scandinavia, the south of the Iberian and Balkan Peninsulas. In Iberia, its southern limit corresponds to the Ebro valley, Central System and Portugal (Galán, 2002a). It is a species with crepuscular and nocturnal habits, whose presence seems to be linked with a high degree of environmental moisture. It lives in areas with a dense vegetal cover, on the edges of the forest and clearings, but even in humid meadows, with loose soils.

Smooth snake – Coronella cf. austriaca

The only fossil vertebra (Fig. 4/4) is small (CL = 2.3 mm), bears an incomplete hypapophysis (character of anterior trunk vertebrae), and has a wide neural arch that is markedly depressed dorsoventrally but with a slightly convex dorsal margin and does not show a well-developed interzygapophyseal constriction, as in the genus Coronella. The precondylar constriction is somewhat marked, a characteristic of the genus Coronella that differs from the juveniles/adults of the other Iberian colubrines (Blain, 2005). According to several authors, C. austriaca has prezygapophyseal processes that are wide at the base and shorter than C. girondica; the parapophysis is also longer in relation to the diapophysis in the former species (Bailon, 1991, Blain, 2005 and Szyndlar, 1984). However, these characters may show a certain degree of variability, and the attribution is made with caution.

The smooth snake is a medium to southern European species. In Iberia, it mainly lives in the northern half, in particular in northern Portugal, Galicia, Asturias, Cantabria, Basque Country and Pyrenean region. In the Euro-Siberian bioclimatic area, it occurs in the collinean and mountain bioclimatic levels (Galán, 2002b). Its distribution in the North of Iberia seems to be correlated with mean annual precipitation higher than 600–800 mm (Galán, 2002b). It lives in humid-cool environments and in particular in the beech grove and Atlantic oak grove, taking refuge in stony or bushy areas.

Ladder snake – Rhinechis scalaris

The only trunk vertebra (Fig. 4/5) is the largest of the vertebrae attributed to snakes in Valdavara 3 (CL = 4.1 mm), robust and rather short (CL/centrum width NAW = 0.98). The neural arch is not so dorsoventrally depressed as in genus Coronella. The precondylar constriction (CNW) is well marked with a ratio CNW/CL = 0.46: larger than in adults of genus Coronella and concordant with a sub-adult of larger Iberian colubrid snakes. The haemal keel, poorly discernable because of the probable action of gastric juices, seems to have been rather wide and not well delimited laterally. Among Iberian colubrids, this vertebra is consistent with Rhinechis scalaris; all other colubrids generally show longer trunk vertebrae.

Rhinechis scalaris is currently present in the province of Lugo, one of its current northernmost occurrences in Iberia (Pleguezuelos and Honrubia, 2002). It lives in Mediterranean environments of rocky bushlands and woodland edges. In northern Iberia, it frequents sunny, dry and opened areas with an orientation facing to the south (Masó and Pijoan, 2011).

Viper – Vipera sp.

One incomplete trunk vertebra (CL = 3.2 mm) documents a viper (Fig. 4/6). It bears, although broken, a straight hypapophysis. The neural arch is posteriorly depressed, and the centrum is ventrally convex in cross section with indistinct lateral margins. The condyle and cotyle are rather large and the zygapophyseal articular surfaces inclined. Unlike the vertebra attributed to Coronella cf. austriaca, this vertebra is much more gracile with a well-pronounced interzygapophyseal constriction, the zygosphene is much slenderer, the dorsal margins of the neural arch in posterior view are rather straight, and show at least one preserved long and slender parapophyseal process. No more precise attribution is possible among Iberian vipers.

There are currently three species of vipers in Iberia: V. latastei, V. aspis and V. seoanei, but only the latter is represented today in the province of Lugo. Vipera seoanei lives in areas with an Atlantic or mountain climate, with not very harsh winters and short and rainy summer and frequents open woodlands, bushlands, and humid meadow edges with a good low vegetal cover, but where insolation is possible (Braña, 2002 and Masó and Pijoan, 2011).

White-tailed eagle – Haliaeetus albicilla

Due to the fracture, the recovered portion does not allow us to take exact measurements (Fig. 3/3). Nevertheless, we have measured the width of the lateral condyle (18 mm) and it is within the parameters established for this species taking into account its sexual dimorfism. Its morphology coincides with the skeletons of this species in the reference collection from the Universidad Autónoma de Madrid. The white-tailed eagle tends to be an over-wintering species (Sánchez-Marco, 2008 and Van der Made et al., 2003) that lives on coastal cliffs or close to inland deep water bodies.

Alpine chough – Pyrrhocorax graculus

Two fragments, a right humerus (Fig. 3/4) and a left ulna represent the alpine chough. The measurement of the diagonal distal width of the ulna (7.04 mm) is within the parameters defined for P. graculus by Tomek and Bochenski (2000). According to the same authors, the humerus distal width (12.2 mm) is within the parameters of this species. The alpine chough lives in the mountains, and usually does not go down to the lowlands nor present in the coasts.

Crow – Corvidae indet.

A right immature tibiatarsus is ascribed to an indeterminate corvid. It has been compared to other immature specimens from other sites and current collections. The measurements of both the distal width (6 mm) and the depth of the distal lateral condyle (5 mm) are within the parameters established for the small corvidae. These birds live in very diverse habitats, especially in extensive forests and rocky areas. Their diet is very varied and they look for areas where they can find everything they need.

Indeterminate birds – Aves indet.

The bird remains are mostly distal epiphyses with a part of the diaphyses. The remains are too poor for a precise identification.

Hedgehog – Erinaceus cf. europaeus

The P⁴ is large with a tall protocone, the metacone oriented to back, a short heel and three roots. The M¹ and M² are broken labially, but both have a quadrangular contour in occlusal view. These characteristics allow us to ascribe our remains to the genus Erinaceus. There is only one Erinaceus species living in Iberia (Erinaceus europaeus) and the remains from Valdavara 3 are similar in size to this species, therefore it is likely that these teeth belong to this species. However, because of the scarcity of the remains we prefer assign them to Erinaceus cf. europaeus. Erinaceus europaeus is distributed over almost all of Iberia and has a very diverse range of habitats, both open and wooded. However, it prefers wet areas.

Shrew – Sorex sp.

The fragmented mandibles recovered are characterized by being medium-sized and by the pigmentation of the molars, typical of the genus Sorex (Fig. 5/1). Given the scantiness of the remains, we have assigned this specimen to Sorex sp. Three species of the genus Sorex live in Iberia: S. araneus, S. coronatus, and S. granarius. They are distributed across the northern half of the Peninsula and are characteristic of humid environments with good herbaceous or shrub vegetation cover (Cuenca-Bescós et al., 2008, Pemán, 1990a, Pokines, 1998 and Zubeldia Garmendia, 2006).

Mole – Talpa sp.

There are currently two species of this genus in Iberia: Talpa europaea and Talpa occidentalis. According to Cleef-Roders and van den Hoek Ostende (2001), the differences in size of the humerus are useful to separate both species. The four Talpa humeri identified have a small size more typical of Talpa occidentalis than Talpa europaea, but due to the scarcity of remains we preferred to assign our specimens to Talpa sp. The joint distributions of Talpa europaea and T. occidentalis cover nearly all of the Iberian Peninsula. Their habitat requirements are linked to soft soils, mainly dealing grasslands, from sea level to 2000 m.

Greater horseshoe bat – Rhinolophus ferrumequinum

The fourth lower premolars are characterized by a triangular shape with a convex lingual edge, a concave distal edge and a thin cingulum (Fig. 5/2); the M₁ and M₂ have a thin cingulum and show a nyctalodont pattern (the hypoconulid connected with the entoconid), typical of the genus Rhinolophus. This combination of characters and the big size of the teeth coincide with R. ferrumequinum, a species with a broad distribution over the Peninsula. This species is found in a wide range of environments. However, it has a preference for wooded areas with open spaces.

Mouse-eared bat – Myotis sp.

The only well-preserved tooth is the M₃, characterized by a myotodont pattern, a thick and regular cingulum, a closed trigonid and an open and slightly reduced talonid. This combination of characters suggests that the mandible belongs to genus Myotis. The scarcity and poor preservation of the remains do not allow us to identify the species, but the small size of the mandible suggests a small-sized Myotis, like Myotis nattereri or M. daubentoni. The genus Myotis is widely distributed in Iberia and occurs normally in forest environments.

Rabbit – Oryctolagus cuniculus

The anatomical elements mainly correspond to basipodia and acropodia. Bones and teeth with morphology typical of the Lagomorpha have a small size. This small size indicates the rabbit Oryctolagus cuniculus, rather than a hare (Lepus), which tend to be larger. Rabbits may live in forests, although they prefer extensive fields covered by shrubs where they can hide. They build burrows in the ground for their shelter. Their main source of food is herbaceous plants or herbs. They prefer warm, dry climates, where sandy areas abound, although this species has successfully adapted to other ecological contexts.

Beaver – Castor fiber

The humerus (Fig. 3/2) has a very wide distal end with a wide articulation, wide olecaranon fossa, wide medial and lateral epicondyles and wide epicondylar crests. There is no epicondylar foramen. The specimen is much smaller than a humerus of Meles and Lynx with which we compared it and differs also in not having an epicondylar foramen. The humerus of Lutra also has an epicondylar foramen. The humerus of Vulpes is much smaller and narrower. The humerus of Hystrix is slenderer. The medial and lateral epicondyles are much wider than in Macaca. The greatest resemblance is with Castor fiber. Castor fiber lives in north and eastern Europe, while its distribution in western Europe is discontinuous. Historically beavers were more widespread and there are geographic names referring to them in areas where today they do not live (e.g., Beverwijk in the Netherlands). In historic times, they went extinct in Spain, but have been reintroduced (Ceña et al., 2004). Their fossils are known from a number of localities, such as Atapuerca TE8-14 and TD6 (Rodríguez et al., 2011).

Southwestern water vole – Arvicola sapidus

The acute angulation of the lingual re-entrant angle 4 (LRA4) in the two lower first molars, together with the presence in molars of an enamel thicker in the posterior than the anterior parts of the triangles allows us to ascribe the material to Arvicola sapidus (Fig. 5/3). Arvicola sapidus is a semi-aquatic rodent represented throughout Iberia and associated with flowing or stationary bodies of water.

Field vole – Microtus agrestis and Common vole – Microtus arvalis

The M₁ assigned to M. agrestis are characterized by their reduced size and by the presence of three closed triangles in the posterior complex (TTC) and four triangles that are closed or minimally confluent in the anterior complex (ACC) (Fig. 5/4–5). There is also a marked asymmetry and alternation in the T4/T5 and T6/T7 triangles, as well as marked alternation in the re-entrant angles; the salient angles are pointed in shape and curved towards the anterior part of the tooth, and the T9 triangle is strongly developed. These features distinguish it from the M. arvalis specimens, which are characterized by the almost symmetrical and parallel arrangement of T4/T5 and especially of T6/T7, as well as the corresponding re-entrant angles, giving a rounded shape to the ACC by contrast with the angular form in M. agrestis.

The field vole lives in the northern third of Iberia, from the Pyrenees to Galicia, but it is absent in the Mediterranean area. It inhabits open spaces close to deciduous woodland, in humid areas with abundant grass, such as ponds and riverbanks up to altitudes of 1900 m. The common vole displays a similar distribution, although it is a more generalist species and more widely distributed. In Iberia, it is more extensively spread over the southern region and may be found in many environments, from pasture to deciduous and coniferous woodland. Its most frequent habitat is subalpine and alpine meadows; it prefers open, not very humid spaces without high vegetation, and it does not live in closed woodland and swampy areas. It is found at altitudes between 900 m and 2000 m, with an annual rainfall greater than 800 mm (Palomo et al., 2007).

Pine vole – Microtus (Terricola) sp.

The characteristics that distinguish the subgenus Terricola from other Microtus are basically the presence in the first lower molar of an open T4 triangle, as well as T4 and T5 forming what is known as the pitymyan rhombus (Fig. 5/6). At present, there are three species belonging to this subgenus in Iberia: M. (T.) duodecimcostatus, M. (T.) pyrenaicus gerbei, and M. (T.) lusitanicus (Brunet-Lecomte and Chaline, 1993). Morphological differentiation of the species is complicated (Brunet-Lecomte, 1988, Brunet-Lecomte et al., 1987, Brunet-Lecomte and Chaline, 1990 and Brunet-Lecomte and Chaline, 1993) and due to the scarcity of remains from Valdavara 3 we prefer to ascribe our specimens to the subgenus Terricola. The living Microtus (Terricola) species have diverging geographical distributions in Iberia, but all are associated with open spaces and stable, moist soils that are easy to dig and that have abundant herbaceous cover.

Bank vole – Clethrionomys glareolus

The M₁ has roots, cement on the labial and lingual re-entrants of the triangles, enamel that is continuous and with same thickness over the entire occlusal perimeter, except for a small enamel-free area on the labial part of the posterior lobe (PL), and confluence of the triangles. These characteristics allow us to distinguish our specimen from the extinct genus Pliomys and assign it to the living species Cletrhionomys glareolus (Fig. 5/7). The bank vole is one of the few extant arvicolines that live in more or less open deciduous woodland or in areas of shrubs or high grasses. It is a good indicator of the presence of vegetation cover in a temperate climate with a humid atmosphere (Cuenca-Bescós et al., 2008, Pemán, 1985, Pemán, 1990a and Pemán, 1990b). Within Iberia, its current distribution is northern (Pokines, 1998 and Ventura et al., 1993), although it may be found in mountainous massifs of Euro-Siberian climate in eastern Iberia (Arrizabalaga et al., 1986).

Extinct vole – Pliomys coronensis

The presence of roots, the absence of cement on the labial and lingual triangles, the enamel-free areas in the mesial anterior cusp (AC) and the labial and lingual part of the posterior lobe (PL) and its relative big size in relation with other arvicolids (with the exception of Arvicola species) allow us to ascribe our material to the extinct vole Pliomys coronensis (Fig. 5/8). The fossil genus Pliomys is probably related to the extant Balkan endemic species Dinaromys bogdanovi (Chaline et al., 1999). D. bogdanovi is a high mountain species ecologically comparable with Chionomys nivalis.

Field mouse – Apodemus sp.

The presence in the lower molars of a low occlusal surface with six main cusps, the anterolingual and anterolabial cusps confluent in an X-shape and separated by a deep, narrow groove, in conjunction with a posterior cusp (cp) that is low, rounded and well-developed and the presence of secondary cusps (c) on the labial face of the m1 and of a mesial tubercle (tma) allow us to assign our specimens to the genus Apodemus (López-García et al., 2011), but due to the scarcity of the remains we can not assign our material to a species (Fig. 5/9). There are two extant species of this genus in Iberia (A. sylvaticus and A. flavicollis), both living in areas with good shrub or tree cover. In regions of homogeneous woodland, however, they prefer the marginal areas.

Red fox – Vulpes cf. vulpes

The P⁴ (Fig. 6/1–3) has a small protocone, situated anteriorly. There is no parastyle. This is a typical morphology in the Canidae. The size is smaller than in the species of the genera Canis and Cuon, larger than in most species of foxes and similar in size to Vulpes praeglacialis and the living red fox V. vulpes (Fig. 6). The former seems unlikely because of the age of the remains and therefore these are assigned to Vulpes cf. vulpes. Vulpes appeared probably during the early middle Pleistocene (Vulpes “angustidens” in Przletice and Hundsheim and Vulpes jansoni in L’Escale – Bonifay, 1971, Fejfar, 1995, Thenius, 1954 and Von Koenigswald and Heinrich, 1999).

Wolf – Canis lupus

These remains have a homogenous size and the morphology of a carnivore. The M₁ (Fig. 7/5) is the most indicative fossil. It has a high protoconid, unlike Ursidae and large Mustelidae, and a talonid, which is larger than in Felidae and Hyaenidae. Its morphology is as in the Canidae. It has an entoconid, as in Canis and Vulpes and unlike in Cuon and Xenocyon lycaonoides, which lost this cusp. The M₂ (Fig. 7/6) has the hypoconid on the buccal side of the talonid, while in Cuon and X. lycaonoides it is larger and situated more in the middle. In side view, the cusps of the premolars tend to be blunt as in Canis lupus, and unlike most other species, including Canis mosbachensis, Canis aureus or the foxes. This is reflected in these teeth being more elongate, or longer relative to the molars (for the P₄ see Fig. 7). The teeth are larger than in raccoon dogs and jackals, tend to be in the ranges of Canis mosbachensis, the earlier Canis lupus and the living wolves from Spain, but are small compared to the large latest middle and late Pleistocene wolves and Xenocyon (Fig. 7 and Fig. 8). Other remains have the morphology and size of a small wolf. The European Pleistocene Canidae include the small foxes (Vulpes and Alopex), the raccoon dog (Nyctereutes), wolves and jackals (Canis), a large wolflike dog (Xenocyon) and the dhole (Cuon) (Crégut-Bonnoure, 1996a). Morphology and size indicates that the dog from Valdavara 3 belongs to the wolf lineage (C. mosbachensis-lupus).

Small species of the genus Canis appeared about 2 Ma ago. One of these gave rise to Canis mosbachensis, which was common in the late early and middle Pleistocene. Around 400–300 ka ago, size started to increase, the premolars became blunter and Canis lupus appeared (García García, 2003, Van der Made, 2010b and Van der Made et al., 2017). Bonifay (1971) named the not yet very large wolves Canis lupus lunellensis. The recent wolves of Spain are again relatively small. The dog from Valdavara 3 is close in size to C. mosbachensis and the living C. lupus in Spain, but it is small compared to the late Pleistocene C. lupus. Clot and Duranthon (1990) mentioned similar small late Pleistocene Canis from the Magdalenian of Isturitz, from the Azilian from Tourasse and various other localities of this age or younger. Some of the small remains have been attributed to a domestic dog. Here we have abundant remains of an adult and some of a juvenile individual, most of them collected in situ from a layer dated to the early late Pleistocene. While the final size of the juvenile bones was not yet reached at the age of death, the teeth show that the adult individual was smaller than other wolves of the same geological age, just like the remains discussed by Clot and Duranthon (1990). In this case, the explanation that the remains belonged to a domestic dog, does not apply, since domestication occurred much later (Thalmann et al., 2013). The premolar morphology is different from that in C. mosbachensis and similar to that in C. lupus: we interpret this individual as a small wolf.

Brown bear – Ursus arctos

The upper fourth premolar (P⁴) has a simple morphology. The tooth is narrow, with an oval outline, and with the protoconid as the only relief (Torres, 1988). Small enamel spikes (Argant, 1991) flank the protoconid. The lower canine is small. The three specimens identified are among the smallest brown bears (Fig. 9/2). The lower fourth premolar (P₄) also has a simple morphology, with three well-separated cusps (Torres, 1988). The deuterocone is clearly separated from the metacone and is positioned at the level of the metacone. The specimen fit within the range of variation of the European brown bears (Fig. 9/3). The lower third molar (M₃) of U. arctos has a crown with symmetrical edges. The hypoconid displays a single relief and the ornamentation of the talonid is relatively weak (Torres, 1988). Although rare, the brown bear known in Europe since the early middle Pleistocene (Argant, 2009) becomes more abundant during the late Pleistocene (especially MIS 5) and it is a common species in the Pleistocene NW Iberian record (García-Vázquez et al., 2011).

Cave bear – Ursus spelaeus

All the specimens assigned to this species correspond to very large individuals. Only the canines could be measured and compared to other populations. The two lower canines are significantly larger than the largest specimens from the European late Pleistocene (Fig. 9). The cave bear appeared at the end of the middle Pleistocene (MIS 6) (Argant, 2009). During the late Upper Pleistocene (MIS 3) some specimens reached impressive sizes. The extinction of the species in Europe occurred at the end of the last glacial. In Galicia, it is one of the most abundant species (Grandal-d’Anglade and Vidal-Romaní, 1997).

Otter – Lutra lutra

A lower canine (Fig. 10/2) had a low crown, unlike in the fox (Vulpes) and wild cat (Felis silvestris). Female monkeys have canines with the lower border of the crown curving down much on the buccal side and there is a depression on the buccal wall where the upper canine fits. The Valdavara 3 specimen has a straighter lower border and lacks that depression. It resembles the canines of mustelids and is a little smaller than the canines of Meles and similar to those of Lutra (Fig. 10). A broken upper canine (Fig. 10/1) matches the lower canine in size and in crown height. The fossil record of the Mustelidae is not well-known. From the early or early middle Pleistocene, some relatively large mustelids are known, but the later large mustelids seem to have been Meles, Gulo and some otter species (Crégut-Bonnoure, 1996b). The fossil Meles seem to have been similar to the living species (Wolsan, 2001). The known otters of the European main land either belong to Lutra and are of similar size as the living species, or to smaller species (Crégut-Bonnoure, 1996b and Willemsen, 1992). The material from Valdavara 3 fits the living species L. lutra.

Spotted hyaena – Crocuta crocuta

The P₄ (Fig. 11/1) does not have a large paraconid, which is unlike in Felidae and it is also less developed than in Pachycrocuta and Hyaena. Several bones fit the spotted hyaena and this is also the case for some incisors, but the latter are also not very different from those of Ursus. The Pleistocene Hyaenidae from the mainland of Europe include Chasmaporthetes (until about 2 Ma), Pachycrocuta brevirostris (about 2 Ma till the early middle Pleistocene), Hyaena (similar) and Crocuta (about 0.9 Ma till the latest Pleistocene). Earliest Pleistocene hyaenas have been placed in Hyaena perrieri or Pliocrucuta perrieri and in Hyaena brunnea. A similar hyaena appeared from the beginning of the Middle until the beginning of the Late Pleistocene and has been assigned to Hyaena prisca, Hyaena striata (= H. hyaena), H. hyaena or P. perrieri (Bonifay, 1971, Crégut-Bonnoure, 1996c, García and Arsuaga, 2001, Kurtén, 1968, Varela et al., 2010 and Von Koenigswald and Heinrich, 1999). The small paraconid on the P₄ from Valdavara 3 is as in Crocuta and smaller than in the other European Pleistocene hyaenas. Within Europe, there was a minor gradual increase in the size of Crocuta, reaching a size much superior to that of the living spotted hyaenas. The specimen from Valdavara 3 is relatively large.

Leopard – Panthera pardus

The reduction of the P_1–2 and M_2–3, and the absence of a talonid in the lower carnassial indicate that the material (Fig. 11/3) belongs to a felid. The carnassials are wider than in Acynonyx and the Machairodontinae. The European Pleistocene Felinae are small (Felis, Lynx), or medium size (Acynonyx, Puma, P. pardus) or large (P. spelaea, P. gombaszoegensis or P. onca gombaszoegensis) (Argant, 1996). The cheek teeth from Valdavara 3 have sizes that fit P. pardus and Puma. There is a large cusp behind the main cusp and behind this one there is a still lower third cusp in P₄. This is like in P. pardus, while in Puma the cusp behind the main cusp is lower and placed further backwards; behind it, there is no third cusp. Panthera pardus appeared shortly after the early middle Pleistocene transition and lasted until close to the end of the Pleistocene (Sanchis et al., 2015).

Cave lion – Panthera spelaea

Two fragments of a P₄, found in 2010 in level B, fit together and fit in an alveolus in a mandible fragment with remains of a M₁ found not in situ in 2009 (Fig. 11/4). The carnassial does not have a talonid and behind it there is no other tooth or alveolus as in the Felidae. The M₁ and a P⁴ (Fig. 11/5) are large and wide as in Panthera leo. A metapodial and phalanges are larger than in recent lions (IPHES). Fossil lions are known from Europe under various species or subspecies names, but more recently the name P. spelaea is applied (Argant, 1996 and Stuart and Lister, 2010). The oldest cave lion may have appeared in isotope stage 17 (Parfitt et al., 2005 and Van der Made et al., 2017). The species survived until about 14–14.5 ka (Stuart and Lister, 2010).

Indeterminate carnivore – Carnivora indet.

A coprolite (Fig. 11/2) consists of two pellets stuck together. It is well mineralized. Hyaenas and other carnivores may produce droppings with a high mineral content. Carnivore excrements are often sausage-shaped and tend to consist of a series of ball shaped elements stuck together. Diameters are variable, but those of Caracal (an African cat similar in size to Lynx) may be up to 20 mm, Acynonyx 25–35 mm, Panthera pardus 20–30 mm, P. leo > 40 mm (Stuart and Stuart, 2000). The diameters of hyaena scat and coprolites (both Hyaena and Crocuta) range between 15.5 and 44 mm (Horwitz and Goldberg, 1989). Crocuta coprolites have diameters from 31 to 44 mm (Fernández Rodríguez et al., 1995). The different pellets in a complete Crocuta excrement differ in shape and size according to their position in the defecation sequence, with type A first leaving the gut and type G last (Diedrich, 2012). The two pellets from Valdavara 3 show great similarity to types B and C, situating them somewhere before the middle of the complete excrement. This is where the diameters tend to be largest, but with 20.8–24.7 mm, the diameters of the Valdavara 3 specimen are relatively small. The excrements of Lynx may consist also of a series of pellets, with morphologies not very unlike those of Crocuta (van Diepenbeek, 1999, fig. p. 251). Sanz et al. (2016) compared shape, length, and width of scat and coprolites of canids, felids and hyaenids. Canis scat tends to be more elongated than that of hyaenas and felids. The C type pellet from Valdavara 3 have a length of about 25 mm and a somewhat shorter diameter, while the B type pellet have a length of 15–20 mm and a diameter of 24.7 mm. These pellets would be in the lower end of the distribution Crocuta (which more likely would correspond to types D–G).

Morphologically, the Valdavara 3 coprolite resembles most those of Crocuta, but they are relatively small. It is to be expected that juveniles produce smaller pellets. Alternatively, they could have been produced by a smaller species of carnivore. Since fresh scat, even with a high mineral content, is fragile, this coprolite suggests that a living individual of carnivore was present at this place.

Horse – Equus cf. torralbae

The remains with the morphology of an equid seem to be homogenous and to belong to a moderately large species. A lower premolar (Fig. 12/2) has a U-shaped linguaflexid, as is common in caballoid horses, while stenonid species have a V-shaped linguaflexid (Forsten, 1992). A complete first phalanx (Fig. 12/1) has a moderately robust aspect. This is confirmed metrically (Fig. 12). Equus hydruntinus has much more gracile phalanges. Phalanges assigned to E. mosbachensis and E. taubachensis are much larger, while specimens assigned to E. germanicus and similar species are of similar length, but are on average more robust. The specimen from Valdavara 3 is in the middle of the phalanges assigned to E. torralbae. The upper cheek teeth tend to have relatively well-developed styles and a relatively large protocone.

Caballoid horses include the domestic horse (Equus caballus), the tarpan (E. ferus) and Przewalski's horse (E. przewalksii or E. ferus przewalskii). The wild forms were previously included in E. caballus, but now the name E. ferus is used for the wild ancestor of the domestic horse. There are different opinions on the classification of the fossil European caballoid horses. Azzaroli (1990) recognized a single species (E. caballus), Forsten (1988) recognized three species (E. mosbachensis, E. germanicus and E. caballus) and Eisenmann (1991) recognized three types with eleven species. These are still the prevalent options. Eisenmann's (1991) types are primarily based on the relative sizes of the protocones. If protoconal indices are calculated based on the averages in her Table 2, it appears that differences between the samples tend to be small, while the samples that have somewhat more different values tend to be small samples. The average of a sample tends to resemble that of the population from which it is taken, but differences may occur and small samples have a greater chance not to be representative than large samples. This effect may have contributed to the more complex classification by Eisenmann (1991).

All specialists in equids recognize the importance of the size and proportions of the metapodials and the first phalanges. The clearest patterns are seen in the metapodials, but to some extent these are also observed in the first phalanges. Fig. 12 (diagram) shows the proximal width (DTp) and length (L) of the first phalanx. Stenonid horses of the E. altidens/petralonensis/hydruntinus group are very gracile. The caballoid horses show much variation in size and robustness (Fig. 12), which is difficult to reconcile with the single species concept of Azzaroli (1990). These horses include a large and robust species to which the names Equus mosbachensis and E. taubachensis are applied, but which probably all correspond to Forsten's (1988) E. mosbachensis. Then there is a group of robust horses of smaller size, to which the names E. germanicus, and E. gallicus are applied. This group corresponds to Forsten's E. germanicus. It is known from France, but no similar horses have been reported from Spain (Maldonado Diaz, 1996). Then there is a group with phalanges of similar length, which are more gracile (Torralba, Atapuerca TD10, TG and TE19, Solana de Zamborino, Pinilla del Valle, Cueva del Buho, etc.), to which the names E. torralbae and E. antunesi have been applied (Maldonado Diaz, 1996 and Martín Penella, 1988), the former being the older name. These could correspond to Forsten's (1988) E. caballus (E. ferus or E. przewalskii). We tentatively assign the Valdavara 3 material to E. torralbae, but recognize the need of a more thorough study.

Steppe rhinoceros – Stephanorhinus hemitoechus

The P_3/4 (Fig. 13/2) has smooth enamel, while in Coelodonta it is much more rugose. The tooth is within the metrical ranges of the P₃ of Stephanorhinus kirchbergensis and of the P₄ of the other species of this genus. The second phalanx of the third toe (Fig. 13/1) is much smaller than in S. kirchgerbensis, but well in the metrical ranges of S. hundsheimensis and close to that of the remaining species (diagram Phal III-2 in Fig. 13). The first phalanges of the side toes are in the ranges for S. hundsheimensis, S. hemitoechus and the earlier S. etruscus, but are clearly larger than the later samples attributed to this species (the two smallest specimens in Fig. 13). The poor remains of a second upper premolar suggest that its crown had a length of about 24.4 mm at the crown base, which is very short.

The rhinoceroses of the Western European Pleistocene belong to four species of the genus Stephanorhinus, to Coelodonta or to the giant Elasmotherium (Guérin, 1980 and Guérin, 1982). The Valdavara 3 material differs from: Coelodonta in having smoother enamel; S. kirchbergensis and the giant Elasmotherium in being smaller; the very small S. etruscus in being larger; and S. hundsheimensis in having a very small second premolar. The best fit is with S. hemitoechus. Stephanorhinus hemitoechus replaced S. hundsheimensis between about 500 and 450 ka. Initially the species may have been more gracile (S. hemitoechus falconeri), while after about 250 ka it was more robust (S. hemitoechus hemitoechus) (Van der Made, 2010a). The material from Valdavara 3 does not allow an attribution to one of these two subspecies. The youngest record of S. hemitoechus is around 45 ka (Stuart and Lister, 2012).

Wild boar – Sus scrofa

The fossils with the morphology of a suid include a maxilla with a M³ with elongate shape and complex talon (Fig. 14/2). The P³ (Fig. 14/1) has a narrow paracone with high parapre- and parapostcristas and relatively well-developed pre- and poststyles. There is a low protocone with its protopre- and protopostcristas, enclosing a fossa, which separates the proto- and paracone. A primocone (nomenclature of Van der Made, 1996) with well-developed pre- and postcristas projects a little lingually. This morphology occurs in Sus scrofa, while Sus strozzii tends to have a paracone that is conical and much more massive, a protocone that is more conical and with less developed crests and a fossa that opens antero-lingually and a less well-developed primocone. The P³'s from Valdavara 3 are narrow as in Sus scrofa, while those of Sus strozzii are wider (Fig. 14). One of the Valdavara 3 specimens is the longest one in the graph, the next longest one being from the early middle Pleistocene of Koneprusy, while the other fossils are from late Pleistocene or Holocene localities.

At present two species of Suidae are recognized from the Pleistocene of the European mainland, Sus strozzii from the early Pleistocene and Sus scrofa from the latest early Pleistocene onwards (Faure and Guérin, 1984, Van der Made, 1992 and Van der Made and Moyà-Solà, 1989). These differ in various features, including the width of their premolars (Van der Made, 1999a). The narrow P³ indicates that the pig from Valdavara 3 is the common wild boar Sus scrofa. How and when Sus scrofa replaced the earlier pigs in Europe is still a matter of discussion, but it was present shortly after 1 Ma (Van der Made et al., 2017). Sus scrofa was subject to significant changes in its body size. The earliest populations had large body sizes (Parfitt et al., 2010), while later it was generally small, though in Taubach and Jaurens it was also large (Faure and Guérin, 1983 and Hünermann, 1977) and at present it is very small. Though sample sizes are small in the comparison (Fig. 14), the wild boar from Valdavara 3 seems to have been very large.

Roe deer – Capreolus sp.

The pedicle of an antler fragment (Fig. 15/A1) is wholly situated on top of the brain case and not partially over the orbita. This is common in the Capreolinae, where the posterior limit of the frontals and antler base, moved backwards. The antler itself has a strongly perlated burr and surface of the main beam. This is typical of Capreolus. The upper molars (Fig. 15/A2) are small. However, the P₂ is not outside the ranges of the earlier samples (Fig. 15/B). A maxilla with two molars, which is part of the first collection and has not been found in situ, seems to fit a molar that was excavated from level B (Fig. 15/2). While some recognize only one species, C. capreolus, up to four species are recognized by other authors, which differ in size and body proportions (Kahlke, 2001 and Pfeiffer, 1998). The size of Capreolus decreased in time and this is documented for the M₃, astragalus, metatarsal and first phalanx (Van der Made, 1998 and Van der Made et al., 2017). The exact moment when the present small size was reached is not clear, but this may have been very late. Here the size changes in the P₂ are documented (Fig. 15B), but it is difficult to see a size trend. Maybe this is so because this tooth is more variable than other teeth. We assign the material to Capreolus sp.

Red deer – Cervus elaphus

A lower premolariform tooth (Fig. 15/3) has the morphology of a D₂, rather than a P₂. The low and elongate tooth is comparable in size to the Cervus elaphus D₂ while the P₂ of this species are wider and the P₂ of Dama shorter (Fig. 15/B). Cervus elaphus appeared in western Europe around 1 Ma (Van der Made et al., 2017). Initially it had antlers with no crown, but after about 500 ka, it developed a crown. It had significant fluctuations in size (Van der Made, 2011 and Van der Made et al., 2014): it was small from the later part of MIS5 until MIS3, large in MIS2 and at present it is small, but there is overlap between the populations with large and small sizes. The Valdavara 3 material is insufficient to estimate the average size of the bones or teeth of the population.

Chamois – Rupicapra cf. pyrenaica

A horn core (Fig. 16/2) is straight with a small diameter as in Rupicapra, while other small bovids of the European Pleistocene tend to have larger and curved horn cores with flattened or triangular sections. The upper molars and premolars (Fig. 16/1) are worn, but must have had high crowns. The molars have no lingual interlobular columns. This morphology occurs in the Caprinae. The molars have well-developed paraexocristas. The size is smaller than in Ovis, Capra and Hemitragus, but fits Rupicapra. The small straight and round horn core and the small dentition of caprine morphology clearly indicate Rupicapra. Two living species are recognized: R. pyrenaica from the Cantabrian Mountains, the Pyrenees and central Apennines and R. rupicapra from the Alps, central Europe, Balkans, Caucasus, and NE Turkey (Aulagnier et al., 2009). It has been suggested that R. pyrenaica dispersed during the middle Pleistocene into Europe and R. rupicapra in the late Pleistocene, displacing the former in most of Europe, but leaving relict populations in Spain and Italy (Crégut-Bonnoure and Guérin, 1996). Analyses of the DNA and microsatellites suggest the existence of one to three species (Pérez et al., 2017, Pérez et al., 2014 and Rodríguez et al., 2009). Pleistocene fossils from Iberia have recently been assigned to R. pyrenaica (Arceredillo-Alonso, 2015). We cannot discuss the taxonomy of the Pleistocene Iberian chamois based on the Valdavara 3 material, and therefore assign those specimens to R. cf. pyrenaica.

Bison – Bison aff. priscus

The molars have high crowns, well-developed interlobular columns and well-developed buccal styles and ectocristas or lingual stylids and ectocristids (Fig. 17/2). This is as in the Bovini, while the Caprinae lack interlobular columns and have less pronounced ectocristas and lingual stylids and ectocristids. Two distal metatarsals (Fig. 17/1) have the lateral and medial side with the shape of a Bordeaux wine bottle as in Bison and not of a Bourgogne bottle as in Bos (Gee, 1993). In side view, the D₂ (Fig. 17/3) has an anterior edge (the parastylid or precristid), which lowers gradually from the tip to the base. This is as in the P₂ and D₂ of Bison, whereas in Bos (and Bubalus) it extends more anteriorly at midheight (Fig. 17.4). The roots of the tooth are not preserved, but based on its small size, the tooth is interpreted as a D₂ (Fig. 17 bivariate diagrams). The tooth is very small, even for a D₂.

The morphology of the remains indicates Bison. Different authors recognize different species or subspecies of Bison in Europe (Crégut-Bonnoure and Guérin, 1996, Sher, 1997 and Van der Made et al., 2017), which differ in size, robusticity of the metapodials, and in skull and horn core shape. Bison schoetensacki survived till the early late Pleistocene, but after its extinction B. priscus became the most widespread European bison (Van der Made et al., 2017). Work in progress suggests that this species may have arisen from a smaller middle Pleistocene form. Recently, B. bonasus and another similar small form have been interpreted to have arisen from interbreeding of B. priscus and Bos primigenius (Soubrier et al., 2016). The bison from Valdavara 3 probably represents a still not yet very large stage of the B. priscus lineage.

Indeterminate mammal – Mammalia indet.

A phalanx is gracile and curved with a very wide proximal end (Fig. 18). The proximal articular facet has some damage at the dorsal side, because of which it is not possible to observe it shape there. The facet is concave in all directions. Near the plantar side of the facet, there is a furrow in the middle, but it does not continue dorsally. The distal articulation is convex in dorsoventral direction and completely flat in transverse direction. Its size is comparable to that of the central phalanges of a cat. However, it differs from those of small carnivora such as Felis and mustelids in that these have a clear furrow in the middle of the distal articulation. In Vulpes and the Lagomorpha, the phalanx is straight and its proximal facet is oriented more vertically with parallel lateral edges. The first phalanges of Macaca have a shallower depression in the middle of the distal articular surface and has the proximal surface of a completely different shape and oriented more vertically (in dorsoventral direction). They are also clearly larger. The first phalanges of Castor are less curved and those of Hystrix tend to be more robust with more vertically oriented proximal facets. Their phalanges have less widened proximal ends. The first phalanges of Erinaceus, the largest insectivore in the Pleistocene of Europe, are much smaller. We did not find a taxon with similar features and therefore cannot assign the phalanx from Valdavara 3 to a taxon.

Although a complete taphonomic study has not been performed, some preliminary observations can be made. Punctures and scores associated with biting and chemical corrosion by digestion have been identified in the epiphysis of the immature tibiotarsus and the humerus of Pyrrhocorax graculus (Fig. 3/4c). These alterations may be caused by nocturnal raptors, like owls (Andrews, 1990). Evidence of human activity has not been identified on the bird bones, which seem to have accumulated due to the action of non-human predators. The damage on corvid and rabbit bones indicates that carnivores living in the immediate surroundings of the cave were the primary accumulator agent. All the rabbit remains exhibit manganese deposits and fissures produced by humid conditions (Fernández-Jalvo and Andrews, 2016). The tarsal bones show a slight thinning of the articular surfaces produced by digestion. One of the phalanges presents carnivore damage. However, the predator cannot be identified due to the paucity of remains. The fish vertebra indicates a size in the range for a prey of the fish eagle. Alternatively, it could have been brought into the cave by the otter. Except the Haliaeetus albicilla bone, all the bird remains show the typical black surfaces associated with manganese deposits. This indicates the presence of water during the formation of the deposit (Fernández-Jalvo and Andrews, 2016).

The taphonomical analysis of the small-vertebrate remains has included 571 elements. The skeletal element frequencies indicate a good representation of small-vertebrate skeleton elements, especially for rodents. The differential anatomical representation, the presence of breakage and, especially, the digestion marks show that the origin of these remains can be related with predation. Although the remains show lower skeletal representation and higher breakage rates than those expected for any predator according to Andrews (1990), it results coherent with fossil assemblages and probably related with post-depositional agents. The percentage of small-vertebrate digested elements is low (21.7%): 22.9% of incisors, 19.8% of molars and 39.1% of postcranial remains. The most significant pattern is the clearly predominant light digestion degree (16.5%), in less proportion the moderate degree (4.4%) and the heavy digestion is rare (0.9%). These few number of digested elements, reinforced by the slight signs of digestion, indicate that the fossils were accumulated by some type of owl, inside the Category 1 or 2 (sensu Andrews, 1990 and Fernández-Jalvo and Andrews, 1992), probably the Tawny Owl (Strix aluco) or the Eurasian Eagle-Owl (Bubo bubo). Both species are opportunistic and sedentary predators (Mikkola, 1983). The fossil association is therefore representative of the ecosystem in the vicinity of the cave at the time when the remains were deposited.

No digestion marks are observed in bat teeth, but such marks are not easily detectable because of the high resistance of the enamel (Fernández-Jalvo, 1995). The species identified frequent caves, so it could be related with in situ mortality, but occasional predation by nocturnal birds of prey cannot be discounted (Kowalski, 1995). The same applies to the amphibian and reptiles. Some of them tend to use caves as refuges (Blain et al., 2008 and Mikkola, 1983), but some remains exhibit digestion marks that related them directly with the predation. As it happens with fish remains, small mammal, amphibian and reptile remains show manganese and fissures related to humid conditions and flooding, which is coherent with a fossiliferous cave environment.

Eighteen species are represented by a single individual (Table 2). The 11 exceptions are: Microtus agrestis, (14), M. arvalis (3), Terricola sp. (13), Pliomys coronensis (9), Sorex sp. (4), Talpa (2/3?), Canis lupus (1 adults, 1 juvenile), Ursus arctos (2), Equus cf. torralbae (2 adult, 1 juvenile), Sus scrofa (1 adult, 1 juvenile) and Bison sp. (1 adult, 1 juvenile). The small-vertebrates are recovered in a different way and the MNI is thus not comparable to that of the large mammals. The species vary greatly in number of specimens between one or a few to 23 (Stephanorhinus hemitoechus), 25 (Bison aff. priscus), 27 (Equus cf. torralbae), and 39 (Canis lupus). Quite often, a fossil association is dominated by two or three species, which are well represented in numbers of specimens and individuals, but this does not seem to be the case. Generally, it is assumed that species living in larger groups tend to be better represented in a fossil association. The representation in Valdavara 3 might reflect this to some extent.

Human activities in Valdavara 3 are testified by a small but diagnostic lithic assemblage (Fig. 19) and some cut-marks on bone remains. The lithic assemblage is composed of 18 artifacts, twelve from the collapsed sediments and six recovered during the excavation. Quartz is the dominant raw material (n = 11), but quartzite (n = 5) and sandstone (n = 2) are also represented. The absence of flint is noteworthy, since this material is well represented in the Upper Paleolithic assemblage of the nearby Valdavara Cave. Rolled cortical surfaces indicate that these materials were collected in alluvial formations. Flakes and flake fragments are the most abundant artifacts (n = 15), and three cores have been also identified. Instead, we have not found any retouched artifact. Large flakes are clearly dominant and most of them show cortical surfaces. Neither Levallois nor blade products have been identified. Two of the cores exhibit very expedient reduction sequences and yielded only a few flakes. However, the third one is most significant from the technical point of view. It is a discoidal core that shows a clearly hierarchical structure (Fig. 19/6). One surface was used for flaking, while the opposite surface, which remains mainly cortical, was used to prepare the striking platforms. This reduction strategy is widespread in Middle Paleolithic technologies.

Modifications of faunal remains by humans included cut-marks and bone breakage. We have identified cut-marks on 28 bone fragments (1.9% of the faunal remains) (Fig. 20), including some remains corresponding to bison, red deer, and rhinoceros. They are slicing and sawing marks that are related to different actions, like defleshing, disarticulation and removal of viscera. Carnivore damage is less common. Only six elements showed tooth-marks associated with the action of large carnivores. One bone, a proximal femoral epiphysis of S. hemitoechus, showed both cut-marks and carnivore damage. Concerning the non-biological processes, the most common modifications are associated with weathering (30.1% of bone remains), manganese coatings (29.7%) and trampling (24.6%).

Pre-Pleistocene fossil localities are common and do not have indications of human presence. However, towards the end of the Pleistocene humans became increasingly important in the formation of sites with fossils. Archeological assemblages characterized by huge amounts of faunal remains and scarce evidence of human presence are common throughout the middle and late Pleistocene. They have been called “paleontological sites with indices of human presence” (Brugal and Jaubert, 1991) and indicate mobility strategies that were frequent in the Lower and Middle Paleolithic. Most assemblages correspond to cave sites, although open-air occurrences are not unknown (Daura et al., 2013).

Two processes are normally used to explain the formation of such bone assemblages. On the one hand, animals may enter the caves by accidental falls in vertical shafts acting as natural traps (see Bar-Oz et al., 2012, Marder et al., 2011, Moncel et al., 2008 and Wolverton, 2006). On the other hand, faunal accumulations may be related to non-human predators using the caves as dens (Baryshnikov and Tsoukala, 2010, Diedrich and Žák, 2006, Fosse, 1996, Fosse et al., 1998, Stiner, 1991 and Villa et al., 2010). Among the latter, the caves used by bears for hibernating and denning tend to exhibit huge amounts of bear remains (see Andrews and Turner, 1992, Argenti and Mazza, 2006, Grandal-d’Anglade, 1993, Quilès et al., 2006, Stiner, 1999, Stiner et al., 1996 and Torres et al., 2007). However, these natural formation processes are not mutually exclusive and faunal assemblages are normally palimpsests formed by different inputs.

In these sites, human presence is attested to small lithic assemblages and some modified bones, resulting from short visits for exploiting the animal carcasses accumulated by natural agents. In general, introduction of animal resources by humans and residential occupations are discarded. However, some scholars have questioned whether the lithic artifacts found in these sites really indicate human presence. Villa and Soressi (2000) suggested that the lithics from carnivore dens might correspond to redeposited artifacts introduced into the caves by natural processes. The association between bones and lithics would be purely accidental. However, this can hardly be considered as a general explanation and the nature of the bone and lithic association must be considered on a case-by-case basis, taking into account taphonomical data and artifact refitting.

These assemblages are common throughout the middle and late Pleistocene in the Iberian Peninsula: Trinchera Galería (Ollé et al., 2005), Torrejones (Díez et al., 1998), Lezetxiki (Alvarez-Alonso and Arrizabalaga, 2012), Tossal de la Font Cave (Olària et al., 2005), Teixoneres (Rosell et al., 2010), 120 Cave (Terradas and Rueda, 1998), Ermitons (Maroto, 1985), Mollet (Maroto et al., 2012). Some sites have been interpreted as natural traps (Trinchera Galería) and others as hyaena (Torrejones, Teixoneres, Mollet) or bear (Lezetxiki) dens. Most of them have been dated to the end of the middle and the early late Pleistocene. Valdavara 3 is another site with scarce indications of human presence.

None of the species is indicative of cold or glacial environments. In Europe North of the Pyrenees, the climatic cycles resulted in an alternation of “glacial” and “interglacial” faunas. Initially, this was not the case in Iberia, which was a refugium for the “interglacial” faunas. However, during the latest glacial periods, “glacial” taxa penetrated the Peninsula. This is the case with Mammuthus primigenius, which is known from many sites in Spain, the southernmost being Padul in the province of Granada (Aguirre et al., 1973, Álvarez-Lao et al., 2009 and Van der Made and Mazo, 2001). Its presence in the gravel pit Arriaga near Madrid (Álvarez-Lao and García, 2011a) shows it was present in Spain during MIS6 (Silva et al., 2013). The wooly rhinoceros Coelodonta spread as far south as Madrid and is known from MIS6 in La Parte in the North of Spain (Álvarez-Lao and García, 2011a and Arsuaga Ferreras and Aguirre Enriquez, 1979). Other taxa, which were widespread during cold periods, such as Rangifer, Gulo, Alopex, and Saiga, reached the Cantabrian Mountains and the Spanish Pyrenees (Altuna, 1993, Altuna and Mariezkurrena, 1996 and Álvarez-Lao and García, 2011b). The reindeer reached these areas by MIS6 (Álvarez-Lao and García, 2011a). Rangifer is known from the Galician sites of Valiña and Furada dos Cas, Mammuthus primigenius from Buxán (López-González, 2003), and Coelodonta is also present at the Cova Eirós faunal record (Fábregas-Valcarce et al., 2012). On the one hand, “glacial” taxa are thus to be expected in Spanish sites dating to the cold periods as far back as MIS6. On the other hand, the Iberian Peninsula was a refugium for the “interglacial” fauna, but at present it is impossible to know how far these taxa retreated to the south. The low number of taxa of the Pleistocene faunas from Galicia (Fig. 21) might be because many taxa retreated from mountainous and possibly cold areas. For these reasons the interpretation of a faunal association as indicative of glacial or interglacial conditions in this part of Europe is not straightforward. However, the great diversity of “interglacial” taxa and the total absence of “glacial” taxa suggest that Valdavara 3 represents interglacial conditions. Among the large mammals, Ursus, Sus, and Cervus are indicative of more closed and Stephanorhinus hemitoechus, Equus, Crocuta and Panthera spelaea of more open environments. The presence of Lutra and Castor suggests the presence of permanent water bodies.

Although there are few small-vertebrate remains, from a palaeoenvironmental point of view the assemblage of Valdavara 3 is dominated by taxa linked with open wet meadows, as indicate the abundance of Microtus agrestis and Microtus (Terricola) sp. and the presence of Erinaceus cf. europaeus, Talpa sp., Sorex sp., Anguis fragilis and Coronella cf. austriaca. The presence of Apodemus sp., Clethrionomys glareolus, Rhinolophus ferrumequinum, Myotis sp., Bufo spinosus and Anguis fragilis indicates certain open forest coverage. Arvicola sapidus is indicative of flowing or stationary bodies of water surrounding the site, although the presence of Rhinechis scalaris and Pliomys coronensis may suggest the occurrence of somewhat drier open or bushland habitat. Wet conditions were also suggested by the presence of Haliaeetus albicilla, since this species indicates that there were deep waters near the cave. The absence of cold-adapted species among the micromammals should be stressed.

The rich Pleistocene fauna reflects the presence of varied habitats to satisfy the needs of all these species. The temperate and wet conditions, indicated by the large and small-vertebrate fauna, are consistent with the MIS 5 chronology indicated by OSL dating.

From a biochronological point of view, the best diagnostic species is Pliomys coronensis, an extinct vole whose most ancient record in Iberia are found in the middle Pleistocene layers of Galeria and Zarpazos caves (Cuenca-Bescós et al., 2010a and Galindo-Pellicena et al., 2011) with an age between 430–245 ka (Berger et al., 2008). P. coronensis is a relative abundant species in the Iberian sites from the latest middle Pleistocene to the first half of the late Pleistocene (ca. 40 ka) (Cuenca-Bescós et al., 2010a, López-García, 2011 and López-García et al., 2012), although we can find it occasionally in some Cantabrian sites until 13 ka (Cuenca-Bescós et al., 2010b). The relative abundance of P. coronensis in Valdavara 3 suggests and age of > 40 ka. The absence of the small mammals typical of the late Pleistocene, like Arvicola amphibius or Chionomys nivalis, well represented in the latest Pleistocene layer 4 of Valdavara 1 (López-García et al., 2011), suggests a latest middle to early late Pleistocene (ca. 200–60 ka) age for Valdavara 3.

The origin of Ursus arctos is not very well documented, but the species is cited in Heppenloch and Lunel-Viel (Argant, 1996), that is also from about 400 ka onward. The first appearance of Stephanorhinus hemitoechus is believed to be around 450 ka (Van der Made, 2010a and Van der Made and Grube, 2010). Stuart and Lister (2007) reported a latest radiometric date for this species of ca. 42 ka cal BP and believed that its extinction could be related to the last glacial maximum. The species is possibly still present in Valiña, dated to 31.6–34.8 ka (López-González, 2003). The large mammal fossils suggest an age of 450 to 31.6 ka for Valdavara 3.

The chronocultural attribution of the lithic assemblage is difficult due to the low number of artifacts. Knapping strategies seem clearly aimed at flake production. In addition, one of the cores and the recovery of débordant flakes suggest the use of the discoidal knapping, which is common in Middle Paleolithic assemblages, although it can also be found in other technological contexts. Moreover, no artifacts diagnostic of other technocomplexes (Acheulian, Upper Paleolithic) have been recovered. These features are consistent with a chronology in the time range of the Middle Paleolithic and fall within the Middle Paleolithic variability on northwestern Iberia. Although scarce, these assemblages are characterized by the use of immediate and local lithic resources from secondary deposits, mainly quartz and quartzite. Flint artefacts only appear in the Upper Paleolithic sites, as in the case of Valdavara Cave (de Lombera-Hermida and Rodríguez Rellán, 2010). From the technological point of view, the presence of Levallois technology is scarce, only identified when fine-grained quartzites are available, as in Cova Eirós (de Lombera-Hermida et al., 2014 and Rodríguez-Álvarez et al., 2011).

The age ranges indicated by the small and large mammals and lithic assemblage are consistent with the OSL dates of 103.414 ± 6956 and 112.837 ± 8903 years BP, which corresponds to MIS 5d.

Of particular interest is the long list of large mammals (ungulates and carnivores) from Valdavara 3. Compared to the recent large mammals of Spain (excluding recent introductions or re-introductions), Valdavara 3 lacks Dama, Capra, Felis, Lynx and several of the mustelids (Mustela, Martes, Meles). By comparison, the living fauna from Galicia lacks Cervus, Dama, Capra, Rupicapra and Lynx, meaning that present ungulate diversity is considerably impoverished. The lack of mustelids in Valdavara 3 is very likely due to an insufficient sample size. However, Valdavara 3 has a number of wild mammals, which are extinct (Ursus spelaeus, Stephanorhinus hemitoechus, Bison priscus) or extinct in Europe (Crocuta, Panthera pardus, P. spelaea) or reintroduced (Equus ferus). Compared to the temperate taxa of Europe of isotope stage 5, the following large mammals are lacking in Valdavara 3: Macaca, Elephas, Cuon, Ursus thibetanus, Hyaena, Equus hydruntinus, a large Equus the size of E. mosbachensis, Hippopotamus, Megaloceros, Bos primigenius, as well as the mustelids. Of course, not all European stage 5 species lived in the same place and, in reality, in most places only a selection of these species lived and interacted.

Which species lived together and which did not, depended on the local environment. The fossils found at a site indicate which species lived there, but there are many reasons why species, which lived in that place may not be represented in the fossil collection of that site. In addition to the examples from Galicia (Fig. 21), many well-known sites probably do not represent the true biodiversity, which existed there. The lower layers of Lezetxiki have a high taxonomical diversity, but cave bear is clearly dominant (Altuna, 1972) and represents more than 60% of the identified remains., Similar biases are found in Arlanpe Cave, Labeko Koba Cave and Ekain (Ríos-Garaizar et al., 2012 and Ríos-Garaizar et al., 2015) and Amutxate Cave, where only 39 of approximately 17,000 bones belong to mammals other than cave bears (Torres et al., 2007).

Aguirre (1989) provided lists of the mammalian species present in 38 sites and levels of the upper Pleistocene of Spain, Von Koenigswald and Heinrich (1999) gave faunal lists of central European localities, including many of Eemian age, and Arribas (2004) gave numerous Pleistocene and Holocene faunal lists, some 25 of them of Spanish sites of the Eemian or MIS5. The recently published fauna from Artazu VII in Basque Country is somewhat younger, but shares many of its elements with Valdavara 3, though there is no indication of human presence and the site is younger (dated to 98.4–88.5 ka; Suárez-Bilbao et al., 2016). Compared to those all those sites, Valdavara 3 is one of the richest in species. Most other sites have much shorter lists and it is not possible to know whether these lists represent a minor part of the fauna that lived there, or whether the fauna which lived there was much less diverse. It is possible that the faunal list of Valdavara 3, at least for the large mammals, approaches the real diversity, which was there.

The European fauna suffered a major extinction event during the end of the Pleistocene and the Holocene. Whether this event is due to the expansion of Homo sapiens into Europe, to climatic change, or to a combination of the two is debated, but the result is that the European fauna is considerably impoverished compared to any previous moment of the Quaternary. Valdavara 3 and other rich MIS5 fossil associations are the closest analogues of what the present European local faunas could have looked like without this extinction event.