A symbolic Neanderthal accumulation of large herbivore crania

This work examines the possible behaviour of Neanderthal groups at the Cueva Des-Cubierta (central Spain) via the analysis of the latter’s archaeological assemblage. Alongside evidence of Mousterian lithic industry, Level 3 of the cave infill was found to contain an assemblage of mammalian bone remains dominated by the crania of large ungulates, some associated with small hearths. The scarcity of post-cranial elements, teeth, mandibles and maxillae, along with evidence of anthropogenic modification of the crania (cut and percussion marks), indicates that the carcasses of the corresponding animals were initially processed outside the cave, and the crania were later brought inside. A second round of processing then took place, possibly related to the removal of the brain. The continued presence of crania throughout Level 3 indicates that this behaviour was recurrent during this level’s formation. This behaviour seems to have no subsistence-related purpose but to be more symbolic in its intent.

warmest month (July) is 28.3°C, and the average annual mean temperature is 10°C 144 .
The mean annual rainfall is 703 mm, distributed throughout the year with two months of drought (July and August) 144 . In the highest areas, the climate acquires montane characteristics. In the nearby station of Puerto de Navacerrada (alt. 1890 m), the mean minimum temperature of the coldest month (February) is -3.9°C, the mean maximum temperature of the warmest month (July) is 20.9°C, and the average annual mean temperature is 6°C. The mean annual rainfall is 1423 mm, again distributed throughout the year with same two months of drought 144 . The bottom of the valley its home to a willow (Salix spp.) forest that follows the river bed, surrounded by ash trees (Fraxinus angustifolia) and Pyrenean oaks (Quercus pyrenaica) in areas with siliceous soil.
Pyrenean oaks are replaced by gall oaks (Quercus faginea) in areas with calcareous soil.
Pyrenean oak woods become dominant with rising altitude, until giving way to pine forests (Pinus sylvestris var. iberica), which in turn give way to high mountain scrubland (locally known as piornales) and juniper landscape. Silicophilous and psychroxerophilic grasslands are found in the areas with the highest peaks 144 .

Background
The existence of cavities in the upper valley of the Lozoya River has been known of since at least the mid-19 th century 145 . However, only in the last quarter of the 20th century was their archaeological and paleontological potential not realized, with the fortuitous discovery of the Cueva del Camino cave, then called the Pinilla del Valle site 146 . This site was excavated during the 1980s by researchers from Madrid's Complutense University, led by Prof. Francisco Alférez. The cave had been completely filled with sediments, but at the time of its discovery some of this had disappeared, a consequence of the cave roof falling in due to erosion, and to disturbance caused by construction of the country path that gives the cave its name (camino means path or track in Spanish). The first excavations provided one of the richest collections of vertebrate remains ever recovered in the Iberian Pleistocene, including two human teeth [146][147][148][149][150][151][152][153][154][155] . It was initially interpreted that the site formed during a warm period in the late Middle Pleistocene, and that the human teeth were identified as anteneanderthal. The recovery of stone tools and thermally altered bone remains led to the suggestion that the faunal remains had accumulated during the occupation of the cavity by an anteneanderthal group.  156 ; original version can be found at https://arqueologiaprehistorica.es/arpi/#pdf-arpi-05/3/).
In 2009 a field campaign was organized to determine the extent of the breccia that contained the Microtus vaufreyi remains. For this, the edaphic deposits that smoothed out the slope of the hosting calvero (the local name for a small sloping hill) and which covered the geological substrate were removed, and it was confirmed that the breccia extended to the northeast along a narrow strip limited on both sides by parallel walls of Cretaceous dolomites. A few meters further, the breccia gave way to other silty and clayey sediments that were still confined on both sides by walls of Cretaceous dolomites.
After a few meters more, the dolomite walls turned to the south, then to the east, and again to the south, all along defining a narrow strip (maximum width 4 m) filled with sediments of different nature (carbonatic breccias, chaotic accumulations of speleothems, clays, sands). The presence of speleothems attached to the dolomite walls revealed the entire conduit to be an old gallery of a cave filled with sediments, the roof of which had collapsed owing to erosion, leaving a roofless cave 172 . Subsequent excavations in a sector of these sedimentary fills led to the recovery of the materials described in the present work.

Geological context
The upper valley of the Lozoya River is of tectonic origin. It is a pop-down or sunken block limited by two pop-ups or raised blocks, the structuring of which took place during the Alpine Orogeny 173 . Reverse faults put the different blocks in contact 174 . The raised blocks are made of igneous or metamorphic siliceous rocks from the Hercynian basement.
In the sunken block the sedimentary cover is preserved and includes materials deposited from the Late Cretaceous to the late stages of the Neogene. Quaternary deposits of different nature are recognized (terraces, moraines, peat bogs, alluvial fans) 175 lie over the materials of the raised and the sunken blocks. The Late Cretaceous sedimentary series begins with detrital deposits (sands, gravels and clays) of Turonian-Coniacian age 175 .
Over these lie marls, limestones and dolomites of Late Cenomanian-Lower Turonian age 175 . Limestone and dolomite outcrops are mainly concentrated on the right bank of the Lozoya River, where they emerge in the form of calveros devoid of trees. These carbonatic rocks are extensively karstified 176 , sometimes giving rise to large cavities such as the Cueva del Cabo del Río cave 177 , and others that have been located underground following geophysical procedures 178,179 . The Cueva Des-Cubierta cave is located in a calvero -known as the Calvero de la Higuera -on the right bank of the Lozoya River. This calvero's karst system develops in the 35 m thick Cretaceous limestones and dolomites that outcrop in the area. In these carbonate rocks, a multi-level cave system appeared, with sub-horizontal conduits, controlled by the structure and the nesting of the base level of the Lozoya River, which determined the evolution of the Quaternary landscape. Twenty four terrace levels have been identified in The Calvero de la Higuera multi-level karst lies between 1111 m (the top of the cave) and 1090 m above sea level (the cavities are at the edge of the reservoir). The Cueva Des-Cubierta cave lies in the uppermost known level. It is a roofless cave; its roof was dismantled by erosion at the surface of the Calvero de la Higuera (maximum altitude 1118 m) during the second half of the Late Pleistocene. Its name is a play on words, in Spanish descubierta means both discovered and uncovered, and it alludes both to the fact that it was discovered by the current research team, and that it is an uncovered or roofless cave ( Supplementary Figures 3 and 4).
It consists of a zigzag-shaped subhorizontal conduit under strong structural control, 80 m long by 2-4 m wide and about 3.5 m maximum (known) depth. It is located towards the southeastern boundary of the Calvero de la Higuera, and from its southern extreme it initially runs to the north. The first section is where the sector known as "La Monumental" is located (Supplementary Figure 3). It then zigzags to the west, then to the north, and finally to the southwest until reaching the Navalmaíllo Rockshelter and the Cueva de la Buena Pinta cave at lower altitudes.

Stratigraphy
The excavation of the "La Monumental" sector has not been completed nor has the bottom of the gallery been reached. The accessible stratigraphic series begins with Level 7 (Supplementary Figure 5), so far the oldest and deepest excavated unit of the sequence and which has only been exposed in the centremost, widest area of the cave. It consists of a light reddish-brown, sandy and silty clay with lenses of sands with scarce granules.
The base of the unit has not been exposed, whereas the top of the unit is partially eroded.
The speleothem is complete and laterally continuous (although partially fractured in both the northern and southern sections) but gravitationally collapsed due to the erosion of the underlying Level 7 in the widest centremost part of the cave.
Level 6 is a light reddish brown sandy and silty clay deposited over the collapsed Level 7 speleothem fragments. Over an eastward dipping erosional surface a gour surface has formed due to the partial cementation of the clays.
Level 5 is a 30-45 cm thick light reddish brown sandy and silty clay with lenses of sands and granules, deposited over the above gour surface.
S2 is a thin (1-4 cm) speleothem precipitated over S1 in places where it is in situ and over Level 5 in areas where S1 has collapsed. S2 is overlain by the growth of several stalagmites in the centremost, widest areas of the cave. Level 4 is a thin 15-20 cm clast-supported gravel deposit of flat, rectangular limestone and angular dolostone cobbles and boulders with a scant carbonatic clayey silt matrix.
Level 3 is a pale yellowish brown clast-supported gravel deposit, composed of limestone and angular to sub-angular dolostone cobbles and boulders with a scant carbonatic silty matrix. Towards the north and south of the cave, Level 3 rests conformably over S1 covering Level 7, whereas in the centre of the cave it lies on top of Level 4. Level 3 is up to 2 m thick in the centremost area and progressively becomes thinner outwards until it pinches out towards the North and South of the cave. In the centremost area of the cavity, the uppermost 30 cm appear edaphically modified.
Level 2 lies conformably over Level 3. Level 2 is a hardened pale reddish brown clastsupported gravel deposit cemented with carbonate, with rounded and sub-rounded heterometric limestone and dolostone pebbles and cobbles in an allochthonous silty sand matrix of calcareous, quartz, feldspar and mica composition alongside a few igneous rock fragments. Thin, laminated centimetric speleothem fragments are commonly found embedded in this level. Level 2 reaches its maximum thickness (55 cm) in the northernmost section, and pinches out southward.
Level 1 is an intensely cemented brown clast-supported gravel deposit with differently sized limestone and dolostone pebbles, cobbles and boulders, and with a silty sand carbonatic and quartz-feldspathic allochthonous matrix. It is up to 40 cm thick and liesconformably over Level 2, pinching out southward.
Level H is a 10-80 cm thick Holocene brown to dark brown Ap soil horizon up to 75 cm thick, unconformably overlying Levels 3, 2 and 1. This level includes taphonomically reworked fossils from these three underlying levels.

Site formation and chronology
The currently available data cannot confirm when the Cueva Des-Cubierta cave formed.
Its origin is older than 500 ka since a stalagmite from the interior of the cavity has been dated to 513.4 ± 72 ka by the U-Th method. In addition, a molar of the rodent Mimomys intermedius (= M. savini, see 181 , for a discussion on these taxa's synonymy) has been recovered from one of the stratigraphic levels in the southernmost sector of the site. This species is an important biochronological marker in Western Europe, having disappeared occurred more than 600 ka ago, i.e., before MIS15 182 . The presence of this species, even as a reworked element, indicates that at some point there were sediments in the cavity that were deposited before the extinction of this species. These data suggest that the cavity already existed in the early Middle Pleistocene or late Early Pleistocene.
The U-Th dating of speleothem samples from different points of the Cueva Des-Cubierta cave outside of the La Monumental sector also suggests speleothems developed inside the cavity between MIS11 and the base of the MIS8.
In the La Monumental sector, Level 7 (comprised of reworked, water-laid sandy and clayey deposits) is so far the oldest deposit recorded. Since the oldest dating of the overlying S1 is 231 ± 7.6 ka (beginning of MIS7; Supplementary Tables 30 and 31), a minimum age of MIS8 is inferred for this unit. However, the base of S1 has not been dated and could be much older. S1 (which is 10 to 15 cm-thick) overlays the clayey and sandy deposits of Level 7. Five U-Th dates for the site place the formation of this speleothem as spanning the entirety of MIS7 (Supplementary Tables 30 and 31). Intense erosion, probably due to an important interglacial-glacial transition, affected the underlying Level 7 and partially wore S1 away. This led to the gravitational collapse of the speleothem in the wider cave areas.
Over these collapsed flowstone fragments, sandy and silty clay (Level 6) was deposited.
These sediments were later eroded, and over this eastward dipping erosional surface, a gour surface developed due to the partial cementation of superficial clays. Overlying this gour surface, the sandy and silty clay of Level 5 is deposited. Small and large vertebrate remains confirm the karstic system was connected to the exterior, although the autochthonous nature of the sediments and fossil remains indicate this connection was either limited or far from this area.
S2 is precipitated over these deposits. Associated with this flowstone, several stalagmites developed. A 234 U/ 230 Th dating of one of these stalagmites gave an age of 135.7 ± 1.9 ka (Supplementary Tables 30 and 31) within MIS6a. Conformably overlying the flowstone and stalagmites, flat, rectangular limestone and angular dolostone cobbles and boulders (Level 4) were deposited, probably due to the effect of frost weathering the walls and roof of the cave.
Level 3 was deposited over S2 and Level 4. The cobble and boulder morphology and the scant autochthonous carbonatic matrix reflect the initial stages of the cave roof's collapse; there is little or no allochthonous input. The pollen and micromammal assemblages suggest this level was deposited during a cold period. The maximum age of this level is defined by the dating of the underlying S2, which places it at the end of MIS6. An age for Level 3 within MIS5 is unlikely because conditions were predominantly warm during this interval, while the pollen and micro mammal records indicate cold conditions. Level 3 was therefore probably deposited during MIS4 or the earlier half of MIS3, at the end of which the Neanderthals became extinct. A Neanderthal connection with Level 3 is clearly documented by the typically Mousterian industry recovered. Level 1 lies conformably over Level 2, and is interpreted as a continuation of the denudation and erosion of the cave, with important allochthonous input. Level H represents a Holocene soil unconformably overlying Levels 3, 2 and 1. Some human remains have been identified in the Cueva Des-Cubierta cave. The lack of any repeated element, and the inferred ontogenetic age of all the remains being the same, is compatible with their belonging to a single individual. Five upper deciduous teeth were recovered in situ in the cemented sediments of Level 2, squares G'47 and H'47. Two jaw fragments that refit each other were recovered from Level H (Holocene) near the contact with the underlying Pleistocene sediments in square G'41. In that square, Level H rests on Level 3. These mandibular remains were likely moved from their original position in Level 2, which outcrops a few meters to the north, by post-depositional processes. Finally, a deciduous lower canine was recovered in Level 3 in the adjacent square H'41, one meter below one of the mandibular fragments. This element fits into the corresponding alveolus of the aforementioned mandible. After its detachment from the jaw it probably reached its derived position due to percolation through the gaps between the blocks of Level 3.

Preservation of the Cueva Des-Cubierta 1 individual
The Cueva Des-Cubierta 1 mandible is comprised of two fragments preserving the corpus from the right di2 to the left dm2 (Supplementary Figure 6). The roots of the left deciduous molars, canine and lateral incisor are partially preserved in the alveolar sockets, and the partial crown of the heavily eroded left dc1 is also preserved. Moreover, 6 isolated deciduous teeth have been recovered: 5 deciduous upper teeth (from left side dM1 to dI2 and one dI1 and one dM2 from the right side) and a lower right deciduous canine.

Description of the mandible
Virtual reconstruction and mirror-imaging of the Des-Cubierta mandible allows us to confirm that it has a squared-shaped dental arcade (Supplementary Figure 7). The external symphysis of the Des-Cubierta mandible is retreating and lacks the features related to the presence of a bony chin. The angle of the anterior symphysis in the Des-Cubierta mandible is low, whether this is measured relative to the alveolar plane (67.6°) or the On the external face of the lateral corpus, a single, large mental foramen is located under the dm1/dm2 septum and is placed in the lower half of the body. Neandertals generally show a more posteriorly placed mental foramen with respect to the tooth row, even in the youngest individuals 184,185 . A similar placement below the dm1/dm2 is seen in 37.5% of Neandertals of a similar developmental stage as Des-Cubierta 1, while this placement in modern humans is much less frequent (7.4%) 184 . In addition, the mental foramen is invariably placed in the lower half of the mandibular corpus in adult Neandertal mandibles, and this appears to be a derived feature in this taxon 186 . Internally, the mylohyoid line is pronounced and slightly inclined.   suggesting a moderate shovel-shape. The lingual basal eminence is also moderatelydeveloped. On the lingual surface, a small localized tubercle is present at the base and is placed slightly mesially. This could represent an expression of a canine mesial ridge (grade 2 for canine mesial ridge, Martinón-Torres et al. 19 ). In lateral view, the labial surface is slightly convex from the cervix to the incisal edge. From the occlusal aspect, the labial convexity is moderate. In sum, several of the morphological traits in the Des-Cubierta deciduous dentition are frequently seen in Neandertals (e.g. Roc de Marsal, Chateauneuf, Pech del Azé), but are rarer in modern humans 189,199,200 . In particular, the degree of shoveling in the incisors and the shape of the crown outline in the molars are similar to the expression of these features in Neandertal permanent teeth. While the phylogenetic polarity of individual dental traits is often difficult to establish, the combination of traits seen in the Des-Cubierta dentition is more commonly found in Neandertals 19,185,201-203 .

Small vertebrates
Since 2009, more than 10 tons of sediment from Levels 2, 3 and 5 of the Cueva Des-Cubierta cave have been washed and sieved. The picking out process has not yet concluded, although work with the >2 mm fraction is more advanced since such pieces can be seen with the naked eye. Thus, only qualitative data (presence/absence and relative abundance) are presented here. Levels 2, 3 and 5 are relatively rich in microvertebrates, and a large sample (more than 2000 identifiable remains) has been recovered in which at least 30 taxa have been recognised among amphibians (n = 4), reptiles (n = 4) and small mammals (n = 22) (Supplementary Table 3). Some species were recorded in all three levels, while others were found in only one. Level 2 returned the poorest sample and the smallest number of taxa identified. This is owed to its reduced extension and the cemented nature of its sediments, making the extraction of microvertebrate remains more difficult (and sometimes impossible).  with a relatively narrower crown, a long metastyle and crown edge more convex buccally) were not present for analysis. P4 is the dental element of greatest diagnostic value for distinguishing Panthera spelaea spelaea from the Middle Pleistocene cave lion (P. spelaea fossilis), given the differences in the relationship between the length of the paracone and metastyle [210][211][212] . The lower dentition (p3, p4 and m1) also provides diagnostic information 212  The dimensions of the P3 specimen from Level 2 are much smaller than those of the Middle Pleistocene lions from Azé, and group better with the rest of the P. spelaea spelaea samples as the smallest in the range (Supplementary Figure 14). These remains are thus proposed to belong to Panthera spelaea cf. spelaea Goldfuss 1810, which is in agreement with the proposed chronology for the site.
In the analyzed levels of the Cueva Des-Cubierta cave, dholes (Cuon alpinus Pallas 1811) are well represented, including 21 remains belonging to an adult and a juvenile (a cub with milk dentition) in Level 2, three elements in Level 3 (one being an adult mandible with its dentition), and a metapodial bone and a talus probably belonging to C. alpinus in

Quantification of the large-mammal assemblages of Levels 3 and 2
A total of 3576 large mammal remains were identified in Level 3. Of these, 1314 were indeterminate remains, 1613 were identified anatomically and taxonomically, and 649 anatomically. Based on their size, the latter were attributed to the large, medium or small animal categories (Supplementary Table 5 At Level 3, the skeletal representation of the bovines is characterized by the great abundance of crania (Supplementary Tables 7 and 10 Table 5 Table 7). According to these data, and to the relative abundance (%MAU) (Supplementary Figure 15) of the anatomical elements, the representation of the anatomical elements is relatively homogeneous. A MNI of 2 was calculated from the lunate remains recovered.
Stephanorhinus hemitoechus is the second best represented taxon in terms of NISP, with the same minimum number of individuals as bovines (MNI = 2) (Supplementary Table   8). The NME is 56, for which those of the axial skeleton reach an NME of 13, the forelimbs 11, the posterior extremities 14, and the phalanges and sesamoids 18.
According to these data and the %MAU of the anatomical elements, the representation of anatomical elements is again relatively homogeneous. An MNI of 2 was determined from the calcaneus elements recovered, one of them unfused and the other fused. Thus, the first belonged to a young individual and the other to an adult. Table 7       forest. The identification of aquatic plants in (almost exclusively) Level 2 suggests that times existed when it was wetter or water was more consistently available.

2-
The nitrophilic species throughout the sequence -continuous in Level 2, less common in Level 5, and sporadic in Level 3 -might be understood in terms of the abundance of grazing animals.

3-
Level 3 appears to represent a dry period, while Level 5 shows changes in the available water, and Level 2 appears as quite wet. Level 3 appears to have been a time of more extreme cold and dryness than Levels 5 and 2.

4-Within this climatic framework, the presence of Asphodelus (a pyrophyte) and
carbonicolous NPMs (types 55 C and 7 A), which are found at the top of Level 5 and throughout Level 3, but are only eventual in Level 2, might indicate a relationship between cold and fire.

5-
The behaviour of the vegetation and the available dating information place Level 5 in the isotopic stage of MIS6, Level 3 within MIS4 or a cold period at the beginning of MIS3, and Level 2 within a more temperate period at the beginning of MIS3.
As seen for the other Calvero de la Higuera sites, quartz was the main knapped material associated with the Cueva Des-Cubierta cave (76.6% in Level 2, 72.8% in Level 3), followed by chert, porphyry, rock crystal, and other materials.  Tables 11 and   12). Quartz tends to break and shatter irregularly, and conchoidal fractures that allow the shape of final product to be predicted, do not always occur. The Neanderthal toolmakers that used this cave therefore required a good knowledge of the surrounding landscape; they needed to know where the best quality nodules that would allow the production of larger, better flakes, could be acquired. Collecting the raw material used for the large anvils would have demanded more planning and a greater expenditure of energy since it all had to be transported uphill to the cave (Supplementary Figure 18). Indeed, the economy of raw material use at the site implies strategic decisions had to be taken regarding the acquisition of different materials for use in specific contexts and activities.
Certainly, at the other Calvero de la Higuera hill sites (such as the Navalmaíllo Rock Shelter where different subsistence activities took place) 162,163 , the lithic tools that have been found are smaller, suggesting a different resource collection strategy was followed.
Quartz cores were the knapped materials most commonly recovered, and usually reflected a mid-to-late state of exploitation (determined taking into account the amount of remaining cortex) (Extended Data Fig. 1 and Supplementary Figure 19 and 20.1-4). The mean dimensions of Level 2 cores were 55 x 44 x 32 mm, while for those from Level 3 they were 59 x 46 x 31 mm.  The re-use of broken hammerstones as cores was common (Extended Data Fig. 2). The cores were generally knapped in an expeditious manner, taking advantage of the special features of each blank. The dominant knapping methods used were centripetal and orthogonal, both in a bifacial and unifacial manner (Extended Data Fig. 1 and 2

Evidence of Fire
Fire-induced thermoalteration of bone remains, dolomitic clasts, lithic industry, charcoals and sedimentary matrix was evident. Nearly 38% of the archaeological remains from Level 2 had been affected by fire, as had 34.0% of those from Level 3 (Supplementary   Tables 13 and 14).
Supplementary Table 13. Evidence of fire damage on archaeological remains (Levels 2 and 3). The right columns for each type of remain shows the number of remains (NR) and associated percentage representation (%).   Table 13).

Bone remains Clast remains Lithic industry Charcoals
In general, the lithic tools found were likely burnt after their final configuration; only one piece was affected by fire before it was knapped.
At Level 2 the charcoal fragments are more abundant (NR = 915), larger and better preserved than at Level 3 (NR = 338). These differences may be due to post-depositional processes (Supplementary Table 13). The cranial and post-cranial parts of the skeleton were thermoaltered in similar proportion (with slightly more cranial elements so altered) (Supplementary Table 20  In the southern part of the cave, an area of greater temperatures and fragmentation was identified. In addition, a speleothem soil at the base of Level 3 showing signs of direct burning clearly indicates combustion occurred inside the cave (Extended Data Fig. 3).

Presence of hearths
During the excavation of Level 3, several concentrations of burnt bones, charcoal, rubefacted carbonate rocks and thermally altered lithic industry were found, which are here proposed to correspond to hearths. However, the type of sediment in which they were found (formed mainly by large clasts with a small amount of quickly friable matrix) can leave them difficult to appreciate and excavate.
The most evident hearth was found in square J'36 over speleothem S1. Several rubefacted carbonate rocks (with reddish and greyish colours) were located in this square, spatially

Post-depositional modifications
Although fragmented, the assemblage is generally well preserved. However, post-

Modifications of anthropic origin
Anthropic cut marks and fracturing were both documented, and both for Level 2 and 3.
Most of these modifications were identified on postcranial skeletal remains (mainly the appendicular section) of large adult animals (Supplementary Figure 40). However, modifications of anthropic origin were also seen on elements of medium-sized animals, e.g., a humerus of Cervus elaphus that showed anthropic fracturing (Supplementary   Tables 23 and 24).
In Level 3, two cranial remains (one from Stephanorhinus hemitoechus, the other from Bison priscus), for which the cortical bone surface was better preserved, showed clear evidence of human activity (Extended data Fig. 4 and 5). The rhinoceros cranium showed both cut marks and fractures of anthropic origin. On the bison cranium, cut marks were identified in the nasal area (Supplementary Table 25).

Supplementary Table 23. Anthropogenic modifications identified for the faunal remains of Level 3.
The columns show number of specimens (NSP) by anatomical elements and the size of the captured animals.

Anthropic breakage Cut marks
Very large-sized Large-sized Medium-sized Very Large-sized Large-sized

Carnivore activity
The remains recovered from Levels 2 and 3 suggest that the activity of carnivores inside the cavity was practically non-existent. The associated modifications identified included only the rounding and collapse of some of the fracture edges of the bones, and the presence of small grooves and depressions. These reflect the nibbling or furrowing activity of a small carnivore with little modifying capacity. .

Experimental butchering of cattle heads and comparison with the Level 3 cranial remains
The experimental butchering of the cow heads showed the best way to extract the eyes was to first break and remove the maxillae, and then the zygomatic bones. The removal of the eyes may therefore explain why the crania lack these elements. The morphology of the resulting cranium is very similar to that of the bovine crania from Level 3 (Supplementary Figure 41). The general absence of maxillae and zygomatic bones in the archaeological assemblage would indicate that these bones were removed before the crania were taken into the cave.
It is more difficult to understand how Neanderthals might have accessed the brain. The assemblage is mainly composed of crania that preserve the horns, but with a poorly conserved neurocranium, the result of post-depositional fracturing. During the experimental butchering of the cow heads, the easiest way to access the brain after removing the skin and muscle was to hit the back of the occipital bone with a hammerstone. In agreement, for most of the bovine crania from Level 3, the occipital bones were incomplete or even absent, although at this assemblage they are represented with a varying degree of integrity as isolated elements separate from the rest of the neurocranium. That might be related to accessing the brain by direct percussion on the posterior part of the occipital bone.

Summary of taphonomic results
The taphonomic analysis of the faunal remains from Level 3 revealed evidence suggestive of an anthropogenic assemblage. However, the general characteristics of the assemblage are clearly uncommon.
First, the %NISP with cut and percussion marks was low. However, this is clearly related to the presence of badly preserved cortical surfaces 227 , a consequence of the postdepositional processes that affected the assemblage. The cortical surfaces were thus disturbed, and many anthropic alterations of those surfaces could have been lost. The low ratio of anthropic marks cannot be thus understood as identifying an assemblage unaffected by human activity.
The %MAU values recorded prohibit the assemblage being interpreted as the result of carnivore activity. Carnivore dens (especially hyaena dens) commonly contain fragments of herbivore crania, but post-cranial bones and teeth are also commonly present 26,27,228-235 . In Level 3, however, post-cranial elements, the mandibles and maxillae were clearly underrepresented. It might still be argued, however, that the assemblage was produced by the action of other abiotic or biotic processes. Indeed, a fossil assemblage can result from animals falling victim to a natural trap. However, in such cases the resulting accumulation would include complete animals 24, [236][237][238] . Assemblages can also be produced by the action of a watercourse (of either high or low energy). However, the %MAU values recorded again rule this out for the examined site. Crania without maxillae and teeth, and the absence of post-cranial elements, are not characteristic of water-driven accumulations 134, [239][240][241] .
The other agents that might modify bone surfaces include ungulates and micromammals (which might gnaw at the bones for minerals or other nutrients), plants, insects, and even natural disarticulation [242][243][244][245][246][247][248][249] . However none of these agents can create accumulations of bones, much less with a clear over-representation of certain anatomical elements as in Level 3.
The differential preservation of bones can also be ruled out. Differences in bone density These crania were taken into the cave after most of the food provided by the captured animals was consumed elsewhere. As suggested in the main text, these crania may have been kept as hunting trophies and thus had a symbolic use (e.g., to project power, authority or hunting prowess, etc.). The cave is thus a new type of archaeological site.

Radiocarbon dating
Several attempts were made to date the bone remains from Level 3, but none yielded enough collagen to allow this. Even when using one of the large bovine petrous bones, in which organic matter is usually better preserved 254,255 , no dating could be made.
A minimum dating of c. 43 ka cal BP was obtained for a charcoal fragment from square H'45 in Level 2. Its maximum age, however, might lie beyond the range of the method used. Details of the provenience of this sample are provided in Supplementary Table 27.
The result of this dating, and its calibration, are shown in Supplementary Table 28.
Supplementary Figure 42 graphically presents the results of this calibration.
The age of the above charcoal sample sets a minimum age for Level 3 (which stratigraphically lies below Level 2) at >43 ka.  The blue strip indicates the IntCal20 calibration curve. The grey area represents the probability density function of the calibrated age. Supplementary Table 29 shows the locations at which the S1 and S2 speleothem samples  Table 29 and Figure   46), and obtained a date for each.   and 2 had to be deposited later, within the Late Pleistocene.

Supplementary
The dates obtained by the two laboratories for S1, which stratigraphically lies below S2, places it in MIS7 (spanning 243-191 ka 257 ). According to the results returned by the CENIEH, the S1 sample from square I'35 gave dates from 231 ka (bottom) to 185 ka (top), while the sample from square K'41, where S1 is located immediately below stalagmite S2, was dated to 223 ka (Supplementary Tables 30 and 31). The latter result agrees with the stratigraphically lower position of the S1 flowstone with respect to the S2 stalagmite.