Level VIII (CP) yielded 6122 remains, of which 2503 (40.9%) have been identified (Table 1). The remaining 59.1% are elements assigned to size groups, with a predominance of medium-sized elements. About ungulates, according to %NISP, the most numerous are chamois (21.9%), Iberian ibex (10.8%) and deer (8.6%). The presence of equids, bovines and roe deer is scant. The carnivore group is even more heterogeneous, comprising wolf/dhole (6.2%), bear (5.6%), leopard (4.4%), fox (2.1%), hyena (0.9%) and wild cat (< 0.1%), as well as a collection of coprolites (11.1%). The presence of birds (16.2%) and other small animals (0.6%) has also been recorded. The MNI confirms chamois (27.4%) as the predominant taxon, while deer (16.1%) exceeds Iberian ibex, which shares third position with bear (14.5%).

The assemblage from level VI (G) is smaller, with 708 remains, of which 247 (34.9%) have been identified (Table 1). The remaining 65.1% are unidentified elements with a predominance of medium-sized elements. About ungulates, according to %NISP, the most numerous are chamois (40.5%), caprine (16.2%), Iberian ibex (5.7%) and deer (5.3%). The carnivore group is less abundant and comprises wolf/dhole (4.8%), fox (2.8%), leopard (1.6%) and bear (0.4%), in addition to four coprolites (1.6%). Birds are well-represented (15%).

In level VIII chamois is the species with the highest MNI (Table 2), eight immature specimens with deciduous teeth and nine with worn permanent teeth. In the case of Iberian ibex, seven individuals were identified, predominantly animals with permanent teeth, in addition to a very young specimen and another young one. In the case of deer, the teeth show six animals, predominantly very young and young; however, the repetition of certain bones indicates the presence of another five large individuals (adults/subadults). There are one young individual and one adult each of roe deer, bovine and equid show. In level VIII there are three hyena individuals (immature, adult-old and senile). Nine bear individuals were identified, two very young, one subadult, two adult, and four senile. This age structure coincides with the dominance of fused bones, except for a distal radius from one immature individual. Their graphic representation shows an attritional profile, where the youngest and most senile are overrepresented in relation to the live population. About other carnivores, a 2.5-year-old leopard, two wolves (adult and senile), one subadult dhole and three adult foxes were identified (Table 2).

Seven chamois were distinguished in level VI, predominantly individuals with functional permanent teeth and one very young individual. The three Iberian ibexes correspond to two subadults and one adult. Deer are represented by two individuals, one young and one adult (Table 2). The scant carnivore remains from level VI yielded one adult individual of Canis, Canis/Cuon, Vulpes, and Ursus, and one young Panthera.

According to %Surv, in terms of the main ungulates in level VIII, chamois, Iberian ibex and deer (Fig. 2; Tab. A1), all the anatomical elements are present, though with a bias towards skull, axial skeleton and small bones. On the other hand, we have determined isolated teeth, sternum or sesamoid bones in Caprinae (Tab. A1) and ribs and vertebrae in medium-sized indeterminate animals. In caprines, the limbs are better represented than in deer, especially the stylopodium and zeugopodium. In the case of carnivores from level VIII (Fig. 2; Tab. A2), there is considerable bias in all the anatomical groups, except for the leopard, of which there is a semi-complete specimen. The taxonomically and anatomically indeterminate fragments mostly correspond to long bones.

About the elements from level VI, it is difficult to assess their representation due to the small size of the assemblage (Tab. A3, A4). All the anatomical groups are present in the chamois, particularly the skull, based on the abundance of isolated teeth, radius, femur and pelvis; there are few axial elements, as is the case in Capra and Caprinae. This situation is the same for carnivores.

There is a clear dichotomy in the levels of fragmentation of caprines (30–50%) and deer (> 70%) from level VIII. In both cases, the values of fresh fractures are far higher than fractures on dry bone. In the ungulates from level VI, the above-described pattern for deer and caprines is repeated. The carnivores from level VIII present very few fragmented bones (10–20%), with higher values for bears (40%) and foxes (50%). The fractures in carnivores are mostly post-depositional, except in the case of bears, where the fractures mostly occurred on fresh bone. The carnivore remains from level VI are quite complete (Fig. 3; Table 3).

Fracture morphotypes are described for long bones in the main ungulates and bear from level VIII (Fig. 4 and Fig. 5). Diaphyseal fragments with and without complete circumference, distal fragments with complete circumference, and proximal fragments with complete and incomplete circumference are represented for caprines. For deer, however, a greater proportion of diaphyses with incomplete circumference is observed and there are no cylinders. The bear remains show diaphyses and articular areas with the complete circumference. In general, a pattern of greater fragmentation is observed in ungulates and bears whereas greater preservation is noted in other carnivores.

In level VIII 403 remains were identified, with 603 modifications (Table 4), mostly on ungulates (192 remains/320 modifications) and indeterminate species (173/218), primarily medium-sized fragments, and to a lesser extent on carnivores (38/65). Modifications due to the action of carnivores are the most abundant (480), followed by anthropogenic (92) and indeterminate (31) modifications. The latter correspond mainly to notches, which may have been produced by humans or other predators.

Carnivore marks mostly appear on ungulate bones (52.9%) (Fig. 6), although also on those of carnivores (13.3%) and indeterminate remains (33.8%), mainly medium-sized. Marks of this type are related to dental action (76.7%) and digestion (23.3%). Chamois bones are the most affected by carnivore bite marks, particularly elements of the stylopodium and girdles, where the largest muscle bundles are found: humerus (18.1%), femur (12.7%), scapula (11.4%) and pelvis (12%); lower values for vertebrae (6.6% cervical, 1.2% thoracic and 2.4% sacrum), tibia (6.6%), ulna (6%), metapodial (4.8%), talus (3%), mandible (1.8%), radius, calcaneus and third phalanges (all three with 0.6%) (Table 5). As regards chamois bones, the most notable modifications are pits, punctures, scores and crenulated edges, which account for 72.9% (Table 4). On a total of 66 chamois remains, most of them show one (36.4%) or two (27.3%) modifications per bone; 16.7% present three, 12.1% four, 4.5% five and 3% between 6 and 8 modifications. The action of carnivores also appears on other ungulates, although on a lower number (Caprinae, Capra, Cervus, and Capreolus). In Iberian ibex bite marks are particularly found on the stylopodium, zygopodium, mandibles and cervical vertebrae; on 15 bones, most show one (40%) or two (53.3%) modifications, and only 6.7% show three on the same remains. On a total of 11 deer bones with bite marks, most have one modification (72.7%), 18.2% have two and 9.1% have three. There are few bite marks on deer and they appear particularly on the vertebrae (40% on thoracic and 13.3% on cervical vertebrae).

The bear is the carnivore with the greatest number of modifications (Fig. 7). Particularly notable modifications are pits and crenulated edges associated in some cases with furrowing and with the disappearance of articular areas (scooping out), as well as scores, but there are no digested remains (Table 4). The most bitten bones are ribs (40.7%) with 35.2% of the marks, femur and ulna with 11.1% of the marks and thoracic vertebrae and humeri with 9.3%. The other elements present lower values: scapula, second phalanx and sacrum (3.7% each), calcaneus, sternum, first phalanx, fourth metatarsal, patella, cervical and lumbar vertebrae (1.9% each) (Table 5). In this regard, the bite marks appear to be concentrated on the axial area and at the top of the limbs. Of a total of 30 bones with bite marks, most present one (50%), two (30%) and three (13.3%) modifications per element, and few show four (3.3%) and six (3.3%) modifications.

Several deer, caprine, medium-sized animal and large bovid bones were identified with signs of digestion in the form of polishing, porosity and bone loss (Table 4). In the medium-sized group there is a predominance of long and cancellous bone fragments, in addition to elements of the axial region. Compact and small bones, such as phalanges and tarsal bones, are the most affected in deer. Among caprines, though, a greater diversity of digested remains is observed, from long bone fragments to skull fragments or smaller elements, such as patellae and phalanges (Table 6). Based on the length of the digested remains, in deer we observed mostly fragments measuring 31–40 mm (46.1%) and > 40 mm (38.5%), with fewer measuring 21–30 mm (15.4%) and none measuring < 20 mm; in chamois, mostly fragments measuring 21–30 mm (50%) and 11–20 mm (40%), with just one element > 40 mm (10%) and none measuring < 20 mm; in the medium-sized remains, predominantly fragments measuring 21–30 mm (40.3%) and 11–20 mm (38.7%), lower values for those measuring 31–40 mm (12.9%) and > 40 mm (8.1%), and again, none measuring < 20 mm. The intensity of digestion is stronger in deer, with considerable bone loss, than that observed in caprines (Fig. 6 and Fig. 8).

Marks of anthropic origin, especially cut marks and some percussion notches, appear on ungulate and indeterminate medium-sized remains, and on one carnivore bone (Fig. 9; Table 7). Additionally, 24 long bone fragments were identified whose morphotype corresponds to percussion flaking. The ratio between anthropogenic modifications and those caused by carnivores shows different values for deer (49.2), Iberian ibex (9.1) and chamois (7.3) and suggests greater human intervention on deer and a more significant role of carnivores in the accumulation of caprine bones. Anthropic marks on deer are especially notable on the stylopodium (16.1% of marks on humerus and femur), zygopodium (12.9% on tibia and 6.5% on radius/ulna), second phalanges (12.9%) and metacarpal (6.5%), with values of around 3% for the other elements, except the axial skeleton, which shows a greater concentration of marks made by carnivores. Marks of human origin appear on the long bones of the hindlimb and mandibles in chamois. In the fragments of medium-sized animals the marks are predominantly on the axial skeleton.

In level VI, 53 marks were identified on 36 ungulate and indeterminate remains, mainly on chamois and medium-sized animal bones. There is very little evidence of anthropogenic modifications, and those caused by carnivores are mainly from digestion and pits (Table 4).

The relationship between %Surv and the density of each skeletal element was significant for deer (r = 0.7687; P = 0.000074895), but not for caprines. In this regard, the differential destruction of less dense elements in deer may be related to hyenas scavenging on their remains.

The main diagenetic alterations observed are calcareous concretions, which represent 42.7% (VIII) and 27.9% (VI). There are also manganese stains of varying importance: 12.2% (VIII) and 15.1% (VI). Pending the sedimentological results, both processes could be related to phases of high humidity and even of flooded areas inside the cave, which are not favourable for human habitation.

Humans and carnivores were involved in the accumulation and alteration of bone remains in the basal levels of Llonin, but they acted differently on each taxon. We will therefore describe the principal modifications caused by the three main agents: Neanderthals, hyenas and leopards.

Anthropogenic modifications appear mainly on deer remains. This, together with the high degree of fragmentation and the morphotypes of long bone fractures, mainly with incomplete circumferences, points to the predominantly human accumulations of these ungulates. On the other hand, the location of the cut marks, their orientation and features inform us about the butchering of these cervids: oblique and longitudinal incisions and scrapes on the diaphyses of the stylopodium, zygopodium and metapodium are related to flesh removal and/or cleaning of the periosteum; oblique incisions on the proximal ulna are due to disarticulation processes, while those observed on the phalanges, calcaneus and skull resulted from skinning. These marks suggest the complete processing of deer that ended with fracturing the bone and accessing the marrow, as shown by the percussion impacts. This evidence appears on adult and immature deer remains, so they would not have been selected by age. According to this and the skeletal pattern observed, Neanderthals hunted deer and transported the entire animal to the cave for processing and consumption, which has also been established at other Middle Palaeolithic Cantabrian sites (Yravedra, 2013 and Yravedra et al., 2016). The palaeoenvironmental studies carried out on basal levels of Llonin and analyses of plant macroremains available for the region corroborate the existence of open forests with the presence of species that thrived in mild Atlantic or central European climates and montane or even subalpine zones during the Mousterian (Costa et al., 1997), an ideal habitat for deer (Carranza, 2011).

Human evidence on caprines is scarcer, as we consider the leopard to be the main agent responsible for the accumulation of these herbivores and we believe they would have also been transported to the cave whole. The abundance of chamois and Iberian ibex reflects the rocky, mountainous terrain in the immediate surroundings of the cave (Granados et al., 2001 and Pérez-Barbería et al., 2010).

Carnivore bones modified by humans appear more frequently during the Upper Palaeolithic, mainly on small animals, but there is increasing evidence of confrontational processes between Neanderthals and carnivores. At Llonin we have documented flesh removal marks on a wolf/dhole radius. There are already some known references in the Iberian Peninsula regarding anthropic processing of canids during the Middle Palaeolithic (Blasco and Fernández, 2012, Díez, 2006, Gómez-Olivencia et al., 2018, Morales et al., 2012, Pérez Ripoll et al., 2010 and Rodríguez-Hidalgo et al., 2015), or other carnivores such as the hyena (Rodríguez-Hidalgo, 2010), lion (Blasco and Fernández, 2012 and Blasco et al., 2010), leopard (Arribas, 1997, Sanchis et al., 2015 and Walker et al., 2012), lynx (Caparrós et al., 2012, Gómez-Olivencia et al., 2018 and Sala et al., 2012), wild cat (Gabucio et al., 2014), brown bear (Caparrós et al., 2012) or cave bear (Yravedra, personal comment). The question is whether Neanderthals hunted carnivores or made use of the remains of dead animals. In any case, there would have been a great deal of competition between them (Brugal and Fosse, 2004).

The rate of carnivorism according to the MNI is 34.5% in level VIII and 23.5% in level VI. The great diversity of species of both herbivores and carnivores is typical of carnivore dens (Binford, 1981 and Klein and Cruz-Uribe, 1984), and hyenas in particular can feed on almost all available resources (Kruuk, 1972, Kuhn et al., 2010 and Sutcliffe, 1970). In this regard, the dimensions of digested deer remains and in some cases their extreme alteration seem to suggest the involvement of the hyena, possibly as a carrion-eating agent that would have fed on the remains left by humans (Cruz-Uribe, 1991, Fosse, 1997, Fosse et al., 2012, Haynes, 1983, Horwitz, 1990 and Sutcliffe, 1970), and they differ from the characteristics of wolf coprocenoses (Esteban-Nadal et al., 2010). In theory, this scavenging activity should provide a bone record with a high number of modifications (Domínguez-Rodrigo and Pickering, 2010), but that is not the case. This could be the result of the intensity of processing and human consumption of the prey, which would leave few consumable remains for scavengers, or due to lower intensity of their consumption by hyenas.

Hyenas could also be responsible for the modifications found on bear remains. No evidence has been found of human intervention on them, but there are carnivore bite marks associated with an extreme modification pattern consisting of destruction of articular areas, fracture morphotypes with a predominance of complete circumferences and an attritional mortality profile (Diedrich and Žák, 2006 and Fosse, 1997). Likewise, the pits and punctures found on bear bones are large, in the range of those made by present-day hyenas (Fig. 11). No digested bear bones have been found, but this appears to be a variable criterion (Domínguez-Rodrigo and Pickering, 2010), although these remains could be contained in the coprolites recovered or could have been destroyed or deposited in other places. It is likely that hyenas would have been attracted by the carcasses of bears that died during hibernation, leading to two possible scenarios: •

the bears occupied Llonin, or

Bearing in mind how small the NISP is compared with the MNI (141/9) and the considerable anatomical bias, the second hypothesis seems more likely, which means differential transport (Fourvel et al., 2014). Likewise, there is no other evidence at Llonin of the use of the cave by bears (beds, trampling or scratch marks on the walls) resulting from these animals moving around during hibernation (Diedrich, 2011 and Sauqué et al., 2015). Scavenging activity by these bears on dead individuals has been identified at monospecific Ursus spelaeus sites, but the remains present larger marks than those made by hyenas and many complete elements are preserved (Pinto-Llona and Andrews, 2004, Pinto Llona et al., 2005 and Rabal-Garcés et al., 2011). Therefore, the accumulation of bears at Llonin is more likely linked to scavenging activity by hyenas, which entails significant alteration of the remains (Diedrich, 2013, Diedrich, 2018 and Fourvel et al., 2014). Both bears and deer exceed the weight range of the prey hunted by hyenas (Hayward, 2006) and the bite marks on these animals are located mainly on the axial skeleton, which is more consistent with scavenging. In level VIII there are also some Canis/Cuon, Vulpes, Crocuta and Panthera bones with bite marks, which could have been scavenged by hyenas and other carnivores.

Regarding the leopard, we believe that it is the main agent responsible for the accumulation of caprine bones. Unlike the bear and deer remains scavenged by hyenas, the bite marks on chamois bones are mostly located on the stylopodium and girdles. Moreover, the destruction pattern is not as strong as the one found on bear remains, the dimensions of the tooth marks are smaller than those made by hyenas and similar to those caused by smaller carnivores such as leopards (Fig. 11). In this regard, the capacity of these felines to accumulate and modify caprine bone remains has been established at different sites in the Iberian Peninsula, such as Boquete de Zafarraya (Barroso et al., 2006, Caparrós et al., 2012 and Geraads, 1997), Amalda VII (Yravedra, 2006a), Los Rincones (Sauqué et al., 2014 and Sauqué et al., 2016), Racó del Duc (Sauqué and Sanchis, 2017) and S’Espasa (Sauqué et al., 2017a). In all cases they are caves situated in steep areas close to the habitat of rock-dwelling species such as Iberian ibex and chamois. In southern and inland Mediterranean sites, the Iberian ibex is the predominant ungulate, while in the Cantabrian region, as observed at Amalda VII and at Llonin, it is the chamois (Sauqué and Sanchis, 2017). Current studies on leopard behaviour conclude that its range of prey is 20–80 kg, although the optimum size would be 23 kg (Hayward et al., 2006 and Mills and Harvey, 2001), which perfectly matches the chamois and, to a lesser extent, the robust Cantabrian Iberian ibex. In sites of the Iberian Peninsula that have been interpreted as leopard dens, caprines have been found together with remains of the leopards themselves, although this is not always the case (Ruiter and Berger, 2000). The caprine bones found there show a small number of modifications caused by biting, between 5 and 20% (Sauqué et al., 2017a). In the case of Llonin, in level VIII carnivore-modified bones account for 15% (chamois) and 7.4% (Iberian ibex); in level VI, 11% (chamois) and 7.1% (Iberian ibex). This modification pattern with predominantly one or two marks per bone is consistent with that observed in the sites described as leopard dens and is less similar to the actions of hyenas (e.g., Sauqué et al., 2017a and Sauqué et al., 2015, b). Likewise, the degree of fragmentation is not very significant, e.g., at S’Espasa it is 64.7% and at Racó del Duc 52%, and at Llonin VIII it is 53.4% for chamois and 31.1% for Iberian ibex. The modification pattern of long bones from caprines in leopard dens is moderate and there are many complete bones (Sauqué and Sanchis, 2017). However, the long bones found at Llonin show a very high level of fragmentation and practically no complete bones are preserved. This may be the consequence of a more complex taphonomic history where, in addition to human accumulation and processing, the scavenger activity of hyenas and perhaps canids may have contributed to a greater degree of destruction of these elements (Domínguez-Rodrigo et al., 2012). In any case, the dimensions of digested chamois bones and fragments of medium-sized animals at Llonin (Fig. 9) are greater than those found in actualistic studies on Canis lupus (Esteban-Nadal et al., 2010 and Fosse et al., 2012). In this regard, some digested caprine bones are quite long (12–15 cm), which we relate to the regurgitation of these elements by hyenas (Fosse et al., 2012).

Two medium-sized remains show both anthropogenic and carnivore modifications, but they do not overlap. They consist of a rib with incisions and bites on the articular region and a digested vertebra with incisions. It has only been possible to define the primary access of humans in the case of the vertebra, which would be later scavenged by carnivores.

Regarding the coprolites found in level VIII, most of them are related to hyenas. The hyena-like globular fossil coprolites clearly match the modern Crocuta scat range (Larkin et al., 2000) and the tube-like ones are clearly differentiated from those of the hyena because of their smaller diameter. Based on the macroscopic characteristics of these coprolites, they are closer to those of large felids, such as leopards (Panthera pardus), mountain lions (Puma concolor) and jaguars (Panthera onca). The four samples analysed under the microscope show similar features to those of modern jaguars (Panthera onca) and leopards (Panthera pardus) (Campmas et al., 2018 and Sanz et al., 2016). It is well known that fossils of extinct Panthera species, such as the European jaguar (Panthera gombaszoegensis) and the American lion (Panthera atrox), show characteristics of both the modern lion and jaguar (Janczewski et al., 1995). Therefore, following the macro- and micromorphological analyses we suggest that the origin of the tube-like coprolites is the Panthera sp. Moreover, the leopard record in the studied levels and their possible responsibility for caprine accumulations, together with the appearance of their coprolites, provide proof of their occupation of the cave and consumption of prey, representing the first coprogenic evidence of this species in a European archaeo-palaeontological context. Feline coprolites have been found at Cova del Coll Verdaguer but they are more compatible with the lynx (Sanz et al., 2016).

To confirm the hypotheses proposed here regarding the origin of the accumulations, especially in the case of the leopard/caprine binomial, it would be interesting to carry out a carbon and nitrogen (δ¹³C and δ¹⁵N) isotope study based on bone collagen in the future to learn about ancient trophic networks and, in particular, the relationship between predators and prey (Bocherens and Drucker, 2003 and Bocherens et al., 2011). It would also be interesting to apply new techniques (Aramendi et al., 2017, Arriaza et al., 2017, Yravedra et al., 2018 and Yravedra et al., 2017) to analyse and compare carnivore bites on caprine set bones from different sites as Llonin, Racó del Duc or Los Rincones, where the leopard seems to be the main modification agent.

The bone remains studied appear to be a cumulative palimpsest resulting from low rates of sedimentation. Initially, the cave served as a place of life and death and was alternately occupied by humans and carnivores (Fig. 12), a pattern observed at other Pleistocene sites of the Iberian Peninsula (e.g., Altuna, 1990, Blasco, 1995, Fernández Rodríguez, 2010, Martínez Valle, 1996, Rosell et al., 2010b, Yravedra, 2006a, Yravedra, 2010a, Yravedra, 2010b, Yravedra, 2013 and Zilhão et al., 2010). Llonin was occupied by hyenas, leopards and various species of canids, as well as by humans. The other species represented correspond to exogenous accumulations and are related to the predation and/or scavenging activities of hyenas.

Neanderthals inhabited Llonin at certain times, as shown by the described anthropic evidence found on some of the faunal remains and the lithic assemblage found here. Regarding the morphology of the cave, the concretions and manganese stains on some remains reveal the existence of high levels of humidity inside the cave, which would have made it unsuitable for human habitation but would have been ideal for cave-dwelling carnivores. This seems to be in line with a model of sporadic human occupations without the presence of hearths. As discussed in section 2, access to the cave by humans and carnivores would have been via a different entrance to the current one, which is now sealed off and at a lower level. Likewise, it is possible that human occupations were located in the outer part of the cave near the entrance, which would explain the lack of hearths inside it. The gallery would have been accessed via the current entrance; the lack of space in this area could also have influenced the nature of the occupations. We have been unable to obtain information regarding the seasonality of human occupations due to the high fragmentation of the assemblage and the lack of information on the age at death and sex of ungulates.

The presence of hyenas at Llonin is shown by both their remains and their action on herbivores and carnivores, especially deer and bears. The hyena assemblage from level VIII is scant and shows significant anatomical bias. In this regard, level VII (CP) presents stratigraphic problems but it has yielded many hyena remains, which could possibly complement those of the level VIII assemblage. Hyenas occupied the cave to feed, where they could transport parts of carcasses taken from other places or scavenge on the remains left by other agents. The appearance of coprolites at the base of the deposit confirms that hyenas occupied the inside of the cave to feed and they also defecated there (Diedrich and Žák, 2006 and Fosse et al., 1998). Present-day hyenas defecate to mark their dens and coprolites are found together with the remains of their prey (Fosse, 1999 and Fosse et al., 1998). The function as a breeding den cannot be confirmed due to the lack of young hyena remains, which are also normally associated with a large number of nibbling sticks that do not appear at Llonin (Diedrich and Žák, 2006), although some of the small, rounded coprolites may correspond to young hyenas. The hyenas may have used the old entrance as a breeding area, but we do not currently have access to that part. Cave occupation by cave hyenas as a breeding den or as a temporary feeding place has been documented at numerous sites in the Iberian Peninsula from the middle Pleistocene and more intensely during the upper Pleistocene (e.g., Arrizabalaga and Altuna, 2000, Arsuaga et al., 2010, Blasco, 1995, Blasco and Montes, 1997, Davis et al., 2010, Fernández Rodríguez, 2010, Huguet et al., 2010, Riquelme and Carrión, 2010, Rodríguez-Hidalgo et al., 2010 and Rosell et al., 2010a, b; Sala et al., 2012, Sauqué et al., 2017b and Zilhão et al., 2010). The maximum expansion of cave hyenas and their dens throughout Europe took place in the upper Pleistocene (Fosse et al., 1998).

Regarding the leopards and canids (wolves and foxes) found in level VIII, they occupied the cave to shelter and feed. In the case of leopards, this activity is evidenced by the appearance of their coprolites and the caprine accumulations, especially chamois. Some of these carnivores could have died in the cave, although the presence of a wolf/dhole bone with cut marks could be related to their acquisition by Neanderthals. In any case, it is difficult to establish the role played by canids at Llonin. With regard to wolves, their ability to accumulate remains in caves is limited and circumstantial, but a possible modification/scavenging action on remains (dispersing agent) accumulated by other agents cannot be ruled out and this could also be extended to foxes that occasionally can generate accumulations of ungulates (Domínguez-Rodrigo, 1994b, Krajcarz and Krajcarz, 2014, Yravedra et al., 2014, Yravedra et al., 2011 and Yravedra et al., 2012).

Lastly, it should be mentioned that although the bear mortality profiles recorded at Llonin coincide with those of death during hibernation, the degree to which the bones are affected and the anatomical bias suggest an exogenous origin and they could have been carried into the cave by hyenas or other carnivores.

Most Asturian sites from the Middle Palaeolithic have MIS 3 sequences (Fig. 1) and generally offer limited data on the origin of faunal assemblages. In the Galería del Osario (Ossuary Gallery) of El Sidrón the faunal remains are scant and belong to different species of carnivores and ungulates, where some carnivore-marked remains have been identified (Rosas et al., 2014). At La Viña, erosive processes have led to the decomposition of organic matter, so hardly any faunal remains are conserved from the Mousterian level (Rasilla and Santamaría, 2011-2012 and Santamaría, 2016). At La Güelga, some preliminary data have been published on the fauna from Mousterian levels. Eight thousand bones have been recovered in level 9 (D interior), of which deer and chamois account for 75%, with a smaller presence of roe deer and equids. The bone remains show anthropic signs in the form of cut marks, fractures and thermal alterations; moreover, this assemblage has been related to Mousterian lithic materials and combustion areas from a phase of intense anthropic occupation dating from 45 ky BP. The faunal assemblage from level 4b (D exterior) shows similar characteristics, with a predominance of deer (66%), chamois (21%) and the presence of other taxa such as the mammoth (Menéndez et al., 2009, Menéndez et al., 2014 and Menéndez et al., 2017). At Sopeña, the macrofauna found in Mousterian levels has been compiled into a list which includes several species of carnivores and ungulates (Pinto-Llona et al., 2012: 70), but the remains have not been the subject of a taphonomic study. Lastly, at El Conde only a list of species from the Mousterian level has been compiled, which was used to develop a palaeoenvironmental approximation, but again no taphonomic data are available (Arbizu et al., 2005).

Therefore, information on the human use of Mousterian fauna in Asturias is lacking. This is also the case in Galicia, where we have Mousterian levels 3 and 4 of Eirós Cave (Lombera-Hermida et al., 2014), but with no specific work about fauna. Consequently, to check the data obtained from the study of Llonin, we must look to Cantabria and the Basque Country. Even so, not all the sites in these areas provide specific data on fauna. On the one hand, assemblages such as those of Lezetxiki V and Axlor offer taxonomic studies but with only partial taphonomic approximations (Altuna, 1972 and Castaños, 2005). On the other hand, sites such as Hornos de la Peña (Yravedra, 2010b) and Otero (Yravedra and Gómez-Castanedo, 2010a) have Mousterian levels with faunal remains, but their results are not very representative due to some issues with the material collection. This is also the case with level VI of Lezetxiki, where the fact that 100% of the remains have been determined points to the selective collection of bone material (Martínez-Moreno, 2005). Despite this, in recent years sites such as El Esquilleu (Yravedra, 2006a, Yravedra, 2006b), El Ruso (Yravedra et al., 2010), Morín Cave (Yravedra and Gómez-Castanedo, 2010b), Covalejos (Castaños, 2005, Yravedra, 2013 and Yravedra et al., 2016), El Castillo (Landry and Burke, 2006), Amalda (Yravedra, 2006a, Yravedra, 2007 and Yravedra, 2010a) and Arlanpe (Arceredillo et al., 2013) have provided new data about Neanderthal subsistence models and the occupation of these caves by humans and carnivores. In all these assemblages with studies and updated reviews, notable similarities are observed. Firstly, occupations by Neanderthal groups appear to be short and seasonal, in cases where it has been possible to obtain data on seasonality, e.g., at El Ruso, El Esquilleu and El Castillo. The exception to this is level D of Arlanpe, where long occupations and intense processing of resources have been defined (Rios-Garaizar et al., 2015). The model of short occupations is based on the accumulation of water in the caves, which would have rendered the space less habitable, and on alternating occupations by different carnivores, which lasted until the beginning of the Upper Palaeolithic (Yravedra, 2013). These two issues are present in the levels studied at Llonin. In most cases the caves hold palimpsests of occupations, where both carnivores and humans contributed to and altered the bone assemblages to a varying degree at different sites and levels. In the levels for which anthropic occupations and accumulations have been described, there is a preponderance of large (aurochs and horse) and medium-sized (deer) species. Access to these prey by Neanderthals is primary, and their transport is based on size: differential (large) or complete (medium). The documentation of cut and percussion marks shows complete and intense processing, including skinning, viscera removal, disarticulation, flesh removal and fracturing to access the marrow. Thermal alterations do not seem to be very significant though, except for El Esquilleu (Yravedra et al., 2005 and Yravedra and Uzquiano, 2013), Covalejos (Yravedra et al., 2016) and Arlanpe (Arceredillo et al., 2013). However, their origin has only been established in the first case, linked to their use as fuel for fires.

Regarding small ungulates such as Iberian ibex, chamois and roe deer, they were accumulated by various different agents. At most sites Neanderthals do not play an important role in their procurement, although anthropic modifications have been identified on some remains at Amalda, El Ruso, Morín Cave and Covalejos. However, the number of human marks is small compared with the frequency of marks on deer or large ungulates. Nevertheless, this situation is slightly different at sites in mountainous locations such as El Esquilleu (Yravedra, 2006a, Yravedra, 2006b) and Valdegoba (Díez, 2006), where largely anthropic caprine assemblages have been described (Yravedra and Cobo-Sánchez, 2015). In levels VI–XIII of El Esquilleu there is a considerable presence of cut marks and percussion fractures, and the action of carnivores is less frequent than in higher levels (III–IV). In other assemblages, this information is not clearly defined, as is the case at El Castillo, where anthropic modifications on ungulate remains have been recorded, but the data are presented grouping together Middle and Upper Palaeolithic levels (Landry and Burke, 2006). At Arlanpe, Iberian ibex and chamois represent most of the fauna, and although a larger number of anthropic modifications (7.7%) has been documented than those made by carnivores (4.2%), their location, especially regarding the indeterminate remains, makes it impossible to clearly separate their origin (Arceredillo et al., 2013). In the case of Llonin, as with most other sites, caprine remains are in the minority, so it seems that they were not significant prey in the Neanderthal diet. Both at Llonin and at many Cantabrian sites and other non-Cantabrian sites such as Boquete de Zafarraya or Moros de Gabasa carnivores are the main agents responsible for the accumulation of caprine bones. In most assemblages, based on the size of the pits, the fragmentation pattern and the fact that specimens were transported whole, the caprine accumulations are attributed to medium-sized carnivores, which would sometimes be felids such as the lynx or leopard or canids such as the fox (Yravedra, 2010a and Yravedra et al., 2010, 2015). At most of these Cantabrian sites the presence of carnivores has been inferred from their own bones and dental modifications on prey, but there are few digested bones, coprolites and elements from very young individuals, which would be related to the more intense use of the caves as breeding dens. This contrasts with Llonin, where these features are present. The alternating use of the space by carnivores and Neanderthals is confirmed not only by the primary accumulations of small ungulates, but also by the identification of scavenging activities on the remains left by humans. In this regard, there are records of taxa accumulated by humans, such as deer with dental modifications and even bones with overlapping lithic and dental marks, e.g., at Amalda, Covalejos and El Ruso. The difference with Llonin lies in the fact that the secondary modifications on deer remains are linked to the action of hyenas, a larger carnivore than those described in relation to the aforementioned assemblages. This mixture of anthropic and hyenid activities is observed in other assemblages such as those of Abrigo de Navalmaillo (Huguet et al., 2010), Furninha (Brugal, 2010), and Teixoneres (Rosell et al., 2010b).

In summary, the archaeozoological context of the Cantabrian area provides some data about the subsistence models of Neanderthal groups, although they are generally scarce. Since we are largely dealing with short occupations, the anthropic accumulations are not very abundant and repeatedly focus on large and medium-sized prey. The consumption of smaller ungulates such as caprines or roe deer has only been recorded in certain specific cases. Consumption is intense, but the use of fire as a culinary technique cannot be established in any of the cases. Llonin does not provide any further information on this matter, as no burnt remains have been identified yet. What is clear is the complexity of the formation and alteration of these assemblages from the North of the Iberian Peninsula, where both Neanderthals and carnivores took prey to caves to feed on it, where different carnivores also scavenged on the remains left by humans and other carnivores. Sometimes it is therefore difficult to differentiate between occupations and the origin of the accumulations, although the fact that humans and carnivores chose different taxa at most sites, possibly as a result of interspecific competition for the available resources, could help us distinguish them.

This alternation between caprine predators seems to be common to other Middle Palaeolithic sites in the Mediterranean region of the Iberian Peninsula. In assemblages such as those of Cova Beneito and Cova Negra (Martínez Valle, 1996) or Abric del Pastor (Pérez Luis et al., 2017) we find similar mixed accumulations. However, there are also examples of the anthropic origin of these preys, e.g., in level IV of Abrigo de la Quebrada (Real et al., 2018 and Sanchis et al., 2013). It is interesting to reflect on why the anthropic caprine accumulations are not very significant in most Mousterian archaeological assemblages. In this regard, some authors (Yravedra, 2013 and Yravedra and Cobo-Sánchez, 2015) have proposed and analysed different hypotheses. The first is related to the difficulty that Neanderthals had hunting this type of prey due to lack of tools or knowledge of techniques (Martínez-Moreno, 2005). But this idea has been disproved by the discovery of assemblages of entirely anthropic origin, such as El Esquilleu and Valdegoba. Secondly, it has been suggested that the location of the site may be important. Although sites with anthropic caprine accumulations tend to be in mountainous inland areas, there are exceptions that invalidate this theory, such as Amalda and Llonin, where caprines are not the taxon most frequently accumulated by Neanderthals. Lastly, it has been suggested that it may have been a matter of decisions made by Neanderthal populations and their preferences in terms of hunting and eating. But this idea is difficult to analyse and quantify at present. Llonin raises this issue again, as the cave is situated in a mountainous area where chamois and Iberian ibex would have been available, but there would also have been valleys and forests nearby where deer could be found. The Neanderthal groups that occupied Llonin chose to consume mostly deer, although they occasionally hunted caprines. This may be linked to the seasonality of the occupations, but we cannot offer reliable data in this regard. It could also reflect an eating preference or competition with other predators such as leopards, which prey primarily on caprines.