Introduction

The subsistence of neandertal groups was based on significant mobility, which involved complex management of the distribution of resources across the territory and of seasonal differences throughout the year. As hunter-gatherers, their movements and camps were largely determined by the variable availability of resources necessary for the survival of the group1, although other social variables must also be considered2.

Territorial mobility patterns are often modelled from ethnographic studies3,4,5. Conventional models use two systems described since Binford as collector/forager or radiating/circulating mobility6. Collector/radiating mobility implies sites of specific activities that provide resources to residential sites. Forager/circulating mobility implies that the group moves in predictable seasonal rounds with a single settlement type7. These are two dichotomous models within a wider range of strategies8. A wide diversity of mobility behaviors has been proposed for neandertals such as farsighted circulating models in the Rhône Valley (France)9,10 or a circulating mobility strategy based on daily supplies in Southeastern France11. These models of mobility correlate with the types of stone tools and the techniques used to make them In South-western France, the mobility associated with a laminar Levallois system correspond to the circulating model while the groups associated with a Quina or discoid-denticulate system matched more with the radiating model6,12 as in the Mediterranean part of the Iberian Peninsula13,14.

In addition to spatial logistics, the temporal organization of resource acquisition is another important element of mobility strategies. This is particularly pronounced in temperate climates, where fluctuations in resource supply are highly dependent on annual seasonal cycles15,16. Seasonal logistics have been studied through the duration and seasonality of occupations. The main proxies come from the analysis of faunal remains, based on techniques such as dental wear, cementochronology and the timing of capture of juvenile prey indicated by tooth eruption patterns (see for example17,18,19,20,21,22,23).

Iberian Peninsula, and in particular northern Atlantic Iberia, was intensively occupied by neandertals and numerous sites are known dated between MIS 5 and MIS 3, ca. 130–29 ky bp (see24 for synthesis). However, only a few studies are directed to neandertal organization of annual cycles and territories in this region (see19,22,25). Northern Atlantic Iberia region is located in the southernmost part of the Eurosiberian zone, limiting with the Mediterranean zone. During the cold periods of the Pleistocene, the Iberian Peninsula acted as a refuge area, allowing the preservation of temperate taxa, while cold-adapted species were largely predominant in the Eurosiberian zone26. Northern Atlantic Iberia was an ecotonal area where a mixed fauna coexisted, including both taxa adapted to temperate climate and those suited to cold habitats27,28,29,30,31. In some sites, typical temperate forest species such as Capreolus capreolus and Sus scrofa can be found in association with cold adapted species as Mammuthus primigenius or Rangifer tarandus27,28,29,32. Thus, neandertal groups were able to select their prey from a vast and diverse herbivore community, with no modern equivalent.

This study investigates the level D of the site of Axlor (Basque Country, Spain) which was accumulated during the cooling period from MIS 5a to MIS 424, ca. 85–59 ky bp. Its origin is predominantly anthropogenic33,34,35,36, so the herbivores present in the level D of Axlor correspond to the prey that were hunted by Neandertal groups. We aim to (1) describe the faunal spectrum of the assemblage, (2) identify the dietary patterns and the ecological niche occupied by each herbivore, as well as the prevailing environmental conditions, (3) assess the duration and seasonality of prey mortality events and human occupation(s) during level D accumulation in addition to the seasonal scheduling of hunting activities.

Axlor, level D

The Axlor cave (Dima, Basque Country) is located in the northern Atlantic Iberia, in Western Europe’s middle latitudes (43° 07.3′ N, 02° 43.7′ W). The cave opens to the NW, on a steep cliff 315 m above sea level and 30 km from the present coast. The surrounding area is a karst environment characterized by steeply sloping terrain and enclosed valleys (Fig. 1).

Fig. 1
Fig. 1
Full size image

Geographical location and illustration of the level D of Axlor Cave. Upper left: position of the study area in southwestern Europe (map created with Inskape v.1.4, https://inkscape.org/release/1.4). Upper right: map of the northwestern Iberian Peninsula showing the North Atlantic basin and the boundaries of the Eurosiberian region, in blue; map created with Inskape v.1.4, https://inkscape.org/release/1.4; color area from Iberian Peninsula. Relieve. Hypsometry and shading (hillshade), layer WMS: relieve_hipso_P_Iberica Cartography Service of the Autonomous University of Madrid, https://guiadigital.uam.es/SCUAM/documentacion/relieve_espanya.php, Creative Commons Attribution Share Alike 4.0. Bottom left: location of the Axlor cave in the Indusi karst (seen from the north). Bottom right: surface of Axlor level D under excavation.

The site contains a Middle Paleolithic deposit spanning 50,000 years. The sequence comprises seven levels35,37,38 and has been dated to between circa 100,000 and 50,000 years bp by optical stimulated luminescence-OSL single grain24.

Level D is a deposit between 10 and 20 cm thick. The sedimentary matrix consists of 80% silt and clay. It is a low energy, predominantly fine-grained deposit with a mode and median in the 0.125–0.08 mm class. Sedimentological information indicates the presence of limited but continuous water flow regimes24.

The level is 70.4 ± 5.4 ka old24 and shows a heavy anthropization, with a very dense accumulation of archaeological material. More than 140.000 bone fragments larger than 4 mm and around 11,000 remains of knapped stone tools larger than 4 mm have been recovered. The level contains abundant remains of vegetation carbonised by combustion.

The fauna remains correspond almost exclusively to ungulates33 (deer, goat and bison being the most frequent animals). There are only three remains of carnivorous mammals (two isolated teeth and a mandibula bearing 3 teeth): a leopard cub tooth (Panthera pardus) This represents less than 0.5% of the total number of teeth identified at the level. No postcranial remains of carnivores have been found. The impact of carnivores on the collection is very limited. It can be summarised in a few rare pits and punctures, and no digested remains have been identified. No coprolites have been recovered either. On the contrary, the faunal remains show intense anthropic processing that is reflected in the intentional and systematic fracturing of all the long bones and in the butchery marks present on the cortical surfaces of the bones. The use of long bones is not only for food, but also for technical purposes. Some of these bones have been prepared and used as retouchers for stone tools. 186 retouchers have been recovered34, mainly made from diaphyseal splinters of metapodials and tibiae of deer and bison. These bone tools were used to retouch and maintain the stone tools, in particular the Quina -type scrapers35,36,37,38,39. These bone tools have an average length of 9 cm and weigh 30 g. Microscopic traits of the impacts show that the retouchers were used when bones were fresh, as they retain the mass and elasticity of undegraded tissue34 (Fig. 2).

Fig. 2
Fig. 2
Full size image

Above, surface of level D during excavation. Note the dense accumulation of fractured bones. In the centre and below, retouchers on bone splinters.

The lithic assemblage is unequivocally Quina36,37,40. Lithic materials were imported into the site as retouched tools, including thick sidescrapers of the Quina type, which constituted over 50% of the tools. These tools were intensively curated and exploited as cores for a new generation of tools. Secondary production was retouched in situ, yet there is no evidence of blank production36,38,39. Flint of exogenous origin makes up most of the lithic raw material used, accounting for 70% of the total. This flint comes from outcrops located more than 30 km to the north and south. Additionally, there is limited use of rocks from the local area. These include shale, quartzite, and quartz, which can be found between 1 and 20 km away35,36,38,39. Bone tools are plentiful and are primarily retouchers in diaphyseal splinters of metapods and tibiae of bison and red deer. These are fragments of an average length of around 10 cm and weighing 30 g. They are used for retouching and resharpening Quina-type sidescrapers37,38,40.

Materials and methods

In this study, we selected only the teeth for which a taxonomic attribution was possible. We analyzed 677 teeth covering 9 species in order to assess taxonomy, age and wear. We performed all the statistical analyses with R version 4.3–RStudio version 4.3.3.

The archaeological material analyzed in this work, from the Axlor site, is deposited in the Arkeologi Museoa in Bilbao (Spain).

Taxonomic attribution

Morphologically very similar taxa can coexist, and a precise paleontological analysis is necessary to distinguish, in particular, the aurochs from the bison, the horse from the hydruntine, and the steppe rhinoceros from the woolly rhinoceros.

This study is based on the morphological analysis of teeth. For the Bovinae, it uses mainly the criteria published by several authors41,42,43,44. For the equids, we integrated already described criteria45,46,47,48, as we did for Rhinocerotidae49,50,51,52 as comparison.

Dietary reconstruction of the herbivores

All the available permanent teeth with a taxonomic attribution were selected to reconstruct the herbivore diet using dental mesowear and dental microwear53,54. The two types of analysis allow for the consideration of different periods in the life of individuals. Dental mesowear reconstructs the diet of an individual during the last years of their life, while it is the diet of the last days for dental microwear reconstructs the diet. The combination of these two proxies provides different levels of resolution. Dental mesowear averages the annual diet and provides information about the regional composition of vegetation55. The data provided by dental microwear are influenced by seasonal changes and relate to a more local environment23,55,56.

Both methods allow herbivores to be classified into three dietary categories: browsers, mixed-feeder and grazers. Browsers focus on dicots corresponding to woody plants and herbs, grazers feed mostly on graminoid plants, and mixed-feeders alternate between the two types of plants. For example, in the Kruger National Park (South Africa), browser include between 0 up to 15% of grass into their diet and grazer, between 84 and 100% of grass. Mixed-feeder can include between 4 and 76% of grass into their diet57.

Dental mesowear

Dental mesowear is a method used to assess the dietary wear on herbivore teeth, which reflects the physical properties of the food that affects the teeth during chewing53. Dust and grit consumed during feeding can influence the mesowear pattern58, but high abrasion levels are more likely caused by the phytoliths present in grasses, thus linking mesowear to diet59,60. Rounded or blunted cusps indicate an abrasive diet due to tooth-on-food contact while sharp buccal cusps are associated with low-abrasion diets and result from attrition (tooth-on-tooth contact). Mesowear cover an undefined long period into an individual’s life. It could signify a lifetime accumulation61 or reflect dietary habits from several years before the animal’s death62,63,64.

Following the recommendations from Fortelius and Solounias53, the paracone of the second molars from well-preserved teeth was selected (47.5% of the sample). When second molars were unavailable, first (12.5% of the sample) or third molars (27.5% of the sample) were chosen as well as fourth premolars (12.5% of the sample). Teeth with taphonomic defect, partially or extremely worn were excluded from the analysis53,65. Horse and rhinoceros teeth were discarded from this analysis due to their bad preservation state (horse) and of their too limited wear due to age (rhinoceros).

The shape of the cusps was examined macroscopically using the “ruler method” developed by Mihlbachler et al.66 and later modified by Rivals et al.67. Cusp shape and relief were classified into seven categories: 0 = high and sharp cusp; 1 = high, narrow-angled, and slightly rounded cusp; 2 = high, wide-angled, and rounded cusp; 3 = medium–high and very rounded cusp; 4 = low and round cusp; 5 = almost flat cusp; 6 = blunt with no relief. The average mesowear value for a sample is expressed as a “mesowear score” (MWS). Current browsers typically score between 0 and 2, mixed feeders between 1 and 2.5, and grazers between 2 and 5.5.

Dental microwear

Dental microwear correspond to the examination of microscopic marks on tooth enamel caused by chewing68. These micro-wear patterns can be influenced by several factors, primarily the presence of phytoliths in plants, which reflect the animal’s diet68,69. However, external particles like dust and grit can also leave traces70,71,72. Dental microwear provides a snapshot of the animal’s diet over a short time, ranging from the last few days to about a month73,74,75.

For this study, the microwear analysis followed the protocol and classification developed by Semprebon et al.76 and Solounias and Semprebon54. The occlusal surface of each tooth was cleaned using 96% ethanol. High-resolution silicone (vinylpolysiloxane) was then used to create molds, and casts were made with clear epoxy resin. These transparent casts were examined under a stereomicroscope at 35 × magnification. Observations were limited to a standardized 0.16 mm2 area using an ocular reticule. The micro-traces were classified into pits (small and large), scratches (fine, coarse, and hypercoarse), and gouges. Pits are round or slightly round scars, while scratches are elongated marks with straight and parallel sides. Small pits are shallow and appear bright, while large pits and gouges are deeper and darker due to reduced light refraction. Large pits are roughly twice the size of small pits and less regular in shape, and gouges are even larger (about three times the size of small pits) with highly irregular edges. Scratch texture was categorized as either fine (narrow, shallow scratches with low refractivity), coarse (wide, deep scratches with high refractivity), or a combination, depending on the tooth surface (Fig. 3). Each tooth received a univariate Scratch Width Score (SWS) : mostly fine scratches scored ‘0,’ teeth with a mix of fine and coarse scratches scored ‘1,’ and those with mostly coarse scratches scored ‘2.’

Fig. 3
Fig. 3
Full size image

Illustration of dental microwear features of ungulates from Axlor l.D at magnification × 35 using a numerical microscope Hirox HRX-01 housed in the LAMPEA, Aix-en-Provence, France. A: Right lower second molar of cf. B. priscus (H9-4-8-1037) with numerous fine and coarse scratches, B: Left lower second molar of Ca. pyrenaica (H8-1-23-33) with limited number of pits and scratches, C: Right lower first molar of E. caballus (G8-1-1-20) with numerous fine and coarse scratches, D: Right lower second molar of C. elaphus (J9-1-12-469) with limited number of fine scratches.

Second premolars were excluded from the analysis since they do not capture the same dietary signal as other teeth77,78. Additionally, teeth with poorly preserved enamel or those showing taphonomic damage were removed following the criteria established in previous studies79,80,81,82.

Death events duration and period

Estimation of the death events duration

The duration of the death event was estimated for the adult individuals from two variables from the dental microwear analysis that evaluate the scratch density variability: the standard deviation (SD) and the coefficient of variation (CV) using a tool developed by Rivals et al.69. Since dental microwear reflects seasonal dietary changes, the variability in micro-scratches enabled us to assess whether the accumulation was seasonal or not. While it’s not possible to determine the exact season, these measures can be used to evaluate the relative duration of the accumulation. By combining both variability metrics, assemblages can be classified into one of the following categories: (a) seasonal events or shorter timeframes (low CV and SD), (b) accumulations spanning more than one season (i.e., two or more contiguous seasons) (moderate CV and SD), or (c) events occurring in different, non-contiguous seasons (high CV and SD). This analysis was run for each species considered here as paleontological cohorts whose member may have died in three hypothetical different scenarios: at the same season, in longest continuous event or in discontinuous events. The scale color gradient (from white to black) indicates the probability value associated with the position of a cohort on the heatmap. For smaller samples that may not accurately reflect the SD and CV values of the larger population, a joint bootstrapped function of CV and SD (n = 500, with replacement) was applied using R code from Domínguez-Rodrigo et al.83.

Estimation of the death events period

The timing of the eruption and replacement of the deciduous dentition by the permanent one (tooth eruption and degree of wear of the occlusal surfaces), which is well understood in current species, allows to determine age, with varying degrees of precision, of an individual at the time of death. We estimated then the age of red deer84,85, roe deer86, Caprinae87,88,89,90, horses91,92,93 and bison94 based on previous works for each taxon.

As there is a certain variability between the reference populations and the fossils and that the fragmentation of the collection makes it difficult to estimate ages accurately, we use a common methodology has been applied to ungulate teeth. It allows dental remains to be assigned to relative age classes based on 7 stages of wear, which range from the development of the tooth to total wear of the crown95,96. From this approach, we calculated a minimum number of individuals per combination (MNIc).

In temperate regions, regardless of climate type, ungulates wild species have a well-defined calving period (most often in spring). The foaling period for Przewalski’s horse populations is currently confined to April through June, with a peak in births occurring in May–June97,98. For caprines, particularly the wild goat, births also occur mainly in May and June87. Calving in bison primarily takes place in May99,100. This period is similar for the roe deer, with the majority of births occurring in May101. For red deer, the peak of births occurs between mid-May and mid-June102,103. Following the principle of actualism (or uniformitarism), we determine the season of death for young individuals whose age could be estimated within a few months.

Results

Composition of the Axlor l.D faunal spectrum

A total of 802 dental remains were analyzed and taxonomically and anatomically determined (see Fig. 4). Two Caprinae are present: Capra pyrenaica and Rupicapra pyrenaica along with two species of Cervidae: Cervus elaphus and Capreolus capreolus. Bovinae are also well represented and perissodactyls are also present with Equinae and Rhinocerotidae.

Fig. 4
Fig. 4
Full size image

Illustration of bovid, equid and rhinocerotid teeth from Axlor l.D. I: cf. Bison priscus upper second left molar (J10-4-8-6), II: cf. Bison priscus lower third right molar (I9-3-4-23), III: cf. Bison priscus lower fourth left premolar (G9-4–4-35), IV: Equus caballus upper first right molar (G8-1-1-20), V: Equus caballus lower first or second left molar (K10-3-11-16), VI: Stephanorhinus cf. hemitoechus upper second left molar (K15-3-12-26), VII: cf. Bison priscus upper third left molar (K13-1–4-30). (A): occlusal views, (B): buccal views, (C): lingual views, (D): mesial view, (E): distal view. 1: parastyle, 2: mesostyle, 3: metastyle, 4: entostyle, 5: protocone, 6: “pli caballine”, 1’: parastylid, 2’: metaconid, 3’: entoconid, 4’: entostylid, 5’: hypoconulid, 6’: metastylid, 7’: vestibular groove, 8’: pli ptychostylid.

The Bovinae upper premolars are characterized by a metastyle not well-separated from the mesostyle in general which is commonly observed on bison teeth43,104. On the two fourth premolar, however, it is as marked as the parastyle. The upper molars have a square and vestibulo-lingually wide occlusal contour common in “bisontine” teeth43,44,105,106 (Fig. 4I-A). On the lingual view, the mesial and distal edges are convergent from the collar up to the top of the teeth (Fig. 4I-B, VII-B) while in Bos, the edges are more commonly sub-parallel42,43,107. The enamel islet between the protocone and the hypocone, common in Bos specimen and rare in Bison42,43,108 (Fig. 4I-A) is present only on one tooth (n°K15-3–2-127). In the distal view, the entostyle in short and not very prominent which is generally describe for Bison specimens41,43,109 (Fig. 4I-C; VII-C). Except one teeth (n°K15-3-2-127), they are swollen above the collar which give the impression of convergent vestibulo-lingual edges from the collar up to the top on the distal view which is commonly described in bison teeth42,43,44,110 (Fig. 4VII-D-E). A percent of 94.4% the lower premolars, have a type II morphology sensu Slott-Moller43: the metaconid is wide and fused with the entoconid along its entire height which is commonly found on bison specimens43 (Fig. 4III-A). On the slightly worn lower molars, the ectostylid is often short (62.5% of the teeth) like this is the case in Bison111. On the lingual view, the bottom of the lingual groove of 94.6% of the teeth is concave and shallow whereas it is often deeper and with sharper edges in Bos specimens43. Like for the upper molar, 63.63% of the lower molar are generally swollen above the collar. On the third lower molar (Fig. 4II), the entostylid is not very well marked (Fig. 4II-A) and is curved distally and the width of hypoconulid is constant from the top of the tooth to the collar on 75% of the teeth as it is observed in bison43.

Bovinae dental traits are very variable and only trends rather than diagnostic characters have been found to distinguish species or even genera112,113,114. Nevertheless, the convergence of the more common traits in one species or the other provides taxonomic hints44. In the case of Axlor l. D, the majority of the teeth display “bisontine” features. Thus, the Bovinae will be ascribed to cf. Bison priscus, awaiting for an investigation over the bone remains.

The well enough preserved Equidae upper tooth remains from Axlor l.D corresponds to 2 s premolars, 1 third premolar and 1 first molar. The interstylar portion of the paracone is always convex in the P2 and always concave in the other teeth whereas it is normally convex in hydruntine P3-445,46. The interstylar portion of the metacone is always concave. The parastyle is always board and is simple on the P2 and double on the P3 and M1 (Fig. 4IV-A). The mesostyle is always board and double on the premolars while it is narrow and simple on the M1 (Fig. 4IV-B) as it is most commonly the case on caballoid horses compare to hydruntine45. The metastyle is always simple and narrow. The fossette inside the paracone are not very wrinkled with a mean of three folds on its distal portion for all the teeth. The “pli caballine” is always present and well-marked (Fig. 4IV-A).

On the lower teeth, lingual groove is always flared while it is more often narrow in hydruntine teeth45,46,48. The metastylid is mostly triangular. The vestibular groove never crosses the isthmus except on the molars while it is usually longer in hydruntine cheek teeth47,48 (Fig. 4V-A). The pli ptychostylid is developed on every tooth except 1 M1 and 1 M3. The pli protostylid is always absent except on 1 P2 and 1 P3-4.

The morphological features of the Axlor equid are typical of caballoid horse and it can therefore, be ascribed to Equus caballus.

Only one upper second molar fragment of Rhinocerotidae has been discovered in Axlor l. D that correspond to the vestibular side of the tooth. The mesostyle is not well expressed and the paracone fold is marked and the parastyle is prominent. The posterior part of the ectoloph is almost flat (Fig. 4VI-A). The enamel is moderately rough (Fig. 4VI-B). All these characters are consistent with an attribution to Stephanorhinus cf. hemitoechus (cf. Lacombat49; Pandolfi et al.50,51; Pandolfi and Tagliacozzo52).

The main represented species is Ca. pyrenaica (NR: 50.52%; MNI: 31.75%) followed by C. elaphus (NR: 24.08%; MNI: 28.57%), cf. Bison priscus (NR: 14.18%; MNI: 14.28%) and E. caballus (NR: 5.91%; MNI: 9.52%). Other species are scarcer. Carnivores’ dental remains were identified by Pedro Castaños Ugarte and not studied here but only counted as they are very scarce (NR: 0.74%; MNI: 3.17%) (Table 1).

Table 1 Species representation in the level D of Axlor according to dental remains.
Full size table

Mortality periods and duration of Axlor l. D ungulates

The estimation of mortality event duration using dental microwear was only possible for Ca. pyrenaica, R. pyrenaica, cf. B. priscus, C. elaphus and E. caballus for which sufficient material was available. Since dental microwear was performed on definitive teeth, it reflects only the mortality duration for adults’ specimens. The horse fall within the plein yellow area associated with an error of 0.1 and therefore its position cannot be interpreted. The chamois is located in the left inferior part of the graph consistent with an uni-seasonal death period. The wild goat and the red deer fall within the second area of the heatmap reflecting continuous long-term death events. The bison fall within the third area of the heatmap corresponding to two separate short death events (Fig. 5).

Fig. 5
Fig. 5
Full size image

Boundary lines with the error probability (heat map) based on standard deviations (SD) and coefficient of variation (CV) values of microwear data used for the classification of Axlor l.D samples into short events (region A), long-term events (region B) or two separate short events (region C).

Periods of mortality of juveniles’ individuals were established using age and birth period estimations. The results of seasonality (Fig. 6) highlight the hunting episodes and therefore the site’s frequentation by human groups.

Fig. 6
Fig. 6
Full size image

Seasonality of Axlor’s D layer according to tooth eruption of bovines, horse, red deer, wild goat, roe deer and panther (in MNIc).

A first period is indicated by three remains of bovines corresponding to an individual approximately 6 months old (6–7 months) that would have been killed during the autumn: October or November. Four dental remains might belong to the same subadult individual, its death period is imprecise, spanning from winter to spring (January-May) for two remains, and summer (May–August) for the other two. If those teeth belong indeed to the same individual, its death period likely correspond to May–June (spring).

Horse seasonality was estimated based only on the teeth of a single individual, aged between 15 and 20 months at the time of death, which would be between August and December (summer/autumn).

The dental remains of young wild goat show progressive degrees of eruption/wear, indicating several episodes of acquisition of individuals in the first months of their lives. Three periods emerge: May–June, July–August, and September-December. Two individuals died at the end of spring/beginning of summer, and three at the end of summer/autumn. One individual indicates mid-summer, but this could be a bias due to differential wear (less significant or more advanced than the others). Thus, this species spans all along the warm season with two preferred periods identified for this taxon.

For roe deer, the dental remains of juveniles are scarce. Two deciduous premolars indicate the presence of an individual that died in the first month of its life (0–1 month), between May and June.

The red deer is the taxon that yielded the most remains of young individuals (9), all dying during the same period, mainly between July and September. Based on the presence of two first molar (lower and upper left) at the end of eruption, one of these individuals may indicate a slightly shifted period, between August and October. Overall, the season of acquisition for the red deer is strongly confined to mid-summer, extending to early autumn. Thus, the period during which most juvenile remains overlap is between late spring and mid-autumn.

Similarities between juveniles (dental eruption and wear patterns) and adults’ death periods (heat map) in Axlor l.D are strong. In spring, both Ca. pyrenaica and cf. B. priscus were preyed upon while C. elaphus, Ca. pyrenaica and maybe E. caballus were hunted during the summer. Autumn is the season when the highest species diversity is recorded with cf. B. priscus, C. elaphus, Ca. pyrenaica and maybe E. caballus. Herbivores death periods in Axlor l. D cover the warm season and winter seems to be avoided.

Dietary features of Axlor l. D herbivores

Dental mesowear analysis was possible only for R. rupicapra, Ca. pyrenaica, C. capreolus, C. elaphus and cf. B. priscus. The first four species display a very low MWS (R. rupicapra = 0.6, Ca. pyrenaica = 0.67, C. capreolus = 0, C. elaphus = 0.8) (Table 2). They range among the browsers (Fig. 7). The cf. B. priscus however, display a higher MWS (= 3.45) similar with the values of present-days grazers (Table 2; Fig. 7). There are no differences between the two groups determined by dental microwear (cf. infra).

Table 2 Summary of dental meso- and microwear data for the herbivores from Axlor l. D.
Full size table
Fig. 7
Fig. 7
Full size image

Mesowear score (MWSs) of the ungulates from Axlor l.D compared with the values of present-day ungulates53,54,115,116.

The most numerous dietary categories are the browsers and mixed-feeder specimens with a browsing trend (MWS = 0 and 1) and correspond to 66% of the herbivores MWS distribution (Fig. 8). The grazers (MWS = 4, 5 and 6) are very scarce with only 12% of the specimen belonging to that category. The mixed-feeder that include a moderate quantity of abrasive plants in their diet are also scarce.

Fig. 8
Fig. 8
Full size image

Distribution of the mesowear scores (MWSs) of the ungulates from Axlor l.D.

Few days to weeks prior their death, some of the herbivores’ species from Axlor l.D have a very number of scratches and pits such as C. elaphus, Ca. pyrenaica and R. rupicapra and they grouped within the browser data of Soulounias and Semprebon54 (Fig. 9). Their SWS is low and the large pits can be present (Table 2). Those parameters are consistent with a leaf browsing diet as described by Semprebon et al.117. Equus caballus and S. cf. hemitoechus are characterized by a high number of scratches associated with a low number of pits and they grouped mostly within the grazer data of Soulounias and Semprebon54 (Fig. 9). They have no large pits and the SWS is low to moderate consistent with grazers117. The diet of the cf. B. priscus is very variable with individuals ranging from a low to a high number of scratches associated with a low quantity of pits (Fig. 9). Since Axlor l.D cf. B. priscus population range from the grazer to the browser, its mean display among the mixed-feeder. However, the distribution of the individuals forms two groups with the first one is very homogeneous and characterized by a very low number of both pits and scratches. The second one is heterogeneous and displays a slightly higher number of pits and a number of scratches ranging in between the browser to the grazer areas (Fig. 9). The differences between the two groups are significant (Mann–Whitney test: pit, p-value = 0.003; scratches, p-value = 0.003).

Fig. 9
Fig. 9
Full size image

Bivariate plot of the raw values (circles) and of the mean number of pits and scratches (squares) of the herbivores from Axlor l.D. The error bars correspond to the standard deviation (± 1 SD). The ellipses correspond to the Gaussian confidence ellipse (p = 0.95) on the centroids of current grazers and browsers54.

Discussion

Dietary traits and regional to local vegetation

Dental mesowear is a pluri-annual signal of the dietary characteristics of the ungulates61,64 that can be indicative of the regional environmental settings23,55. In Axlor l.D the majority of the specimens are browsers or mixed-feeder with a browsing trend; the mixed-feeders and the grazers being rather scarce (Fig. 8). The MWS of Ca. pyrenaica, R. pyrenaica, C. capreolus and C. elaphus are low and homogenous and only cf. B. priscus seem to have incorporate abrasive plants on a regular basis (Fig. 7). Similar dietary segregation between the large bovid and the cervids and small bovids has already been described in this region during MIS329. Equus caballus and S. cf. hemitoechus are taxa very often considered as predominantly grazers118,119,120,121. The poor preservation of their teeth may have altered the results and contributed to reducing the representation of this dietary category. Nevertheless, their abundance within level D of Axlor is reduced, particularly in comparison to Ca. pyrenaica and C. elaphus. Therefore, it can be considered that the taphonomic alterations that impacted these taxa had only a moderate effect on the interpretation of the results. The prevalence of browsers might indicate that the regional environment exploited by the predator responsible of the accumulation was dominated by dicots: trees, bushes, or herbaceous plants. Monocots were also present but probably quite rare.

Axlor l. D yield an age of 70.4 ± 5.3 ka and was attributed to the transition between MIS 5a and MIS 424. Pollen studies indicate that this period is marked by the presence of grasses and heathers with patchy evergreen–boreal forest122,123. More precisely, due to the climatic instability of the period, cold and warmer events (GI 19 and GS 19) alternated with the cooling episodes marked by the development of the heathland and the warmer, marked by the increase of the oak forest124,125. The homogeneity in dental mesowear pattern marked by the predominance of browsers can reflect the regional prevalence of dicots, either heathlands or oak forest. The monocots were nevertheless abundant enough for cf. B. priscus to preferentially feed on them.

Dental microwear reflect only the warm season (spring, summer, autumn), since the herbivores did not die in winter. The majority of the specimens are browsers (Ca. pyrenaica, R. pyrenaica, C. elaphus and C. capreolus) and few of them incorporate abrasive plants (Fig. 9). Only some specimens of cf. B. priscus, E. caballus and S. cf. hemitoechus are classified among the mixed-feeders and grazers. The dietary distribution between dental mesowear and microwear is similar. Therefore, the local environment may have been broadly similar to the regional one: dominated by heathlands or oak forest with patches of grasses.

Seasonality of the Neandertal occupations in Axlor l.D

Axlox l.D faunal assemblage is largely dominated by herbivore (Table 1) and its accumulation can be interpreted as of mainly anthropogenic origin33,34,35,36,40. The layer is thick and encompass a palimpsest of several occupation events over the course of an undefined but probably long period. The reconstruction of mortality patterns indicates that the preys were hunted throughout the warm season, but through different ways.

Rupicapra pyreanaica and C. capreolus are only represented by three individuals each. While not being well-represented in general in middle Paleolithic sites, the two taxa may be part of the faunal spectrum hunted by Neandertals126,127. In Axlor l. D, it was not possible to determine the moment and duration of their accumulation and the preservation of the C. capreolus teeth was not good enough to allow observation. Nonetheless, the microwear features of R. pyrenaica fall within the variability of C. elaphus and Ca. pyrenaica. They may have been opportunistically preyed upon in similar habitats.

Stephanorhinus cf. hemiteochus is only represented by one teeth fragment. While in some sites, Neandertal rhinoceros hunting is a strong possibility128,129,130,131, its presence can also reflect, for example, the collect of isolated teeth129. At least six individuals of E. caballus are present in the level D of Axlor. It was not possible to determine the moment and duration of their accumulation. At the time of their death, S. cf. hemitoechus and E. caballus were able to focus on graminoid plants indicating the presence of enough resources for two large herbivores in the local environment.

Capra pyrenaica is the most abundant species at Axlor l.D and is the only one that has been hunted in spring, summer, and autumn. Currently, the mean annual home range of Ca. pyrenaica is of 22.6 ± 25.8 km2 for males and 6.1 ± 5.3 km2 for females132. Their seasonal movements correspond to altitudinal migrations, where populations move in the spring to reach high-altitude meadows and descend during autumn to avoid winter’s heavy snow cover133. Their movements are thus mainly conditioned by the availability of food resources132,134,135. Currently mostly confined to the mountains, the wild goat often occupied lower altitudes during the Pleistocene136, and its migratory behavior likely evolved depending of its habitats. In the vicinity of Axlor, at the time of its death, Ca. pyrenaica incorporate very limited to limited amount of abrasive plant into its diet but always remaining in the browser’s range. Thus, the relatively homogeneous environmental conditions may be among the factors which enabled Capra pyrenaica presence and availability to Neandertal predation at least throughout the entire warm season.

Cervus elaphus is the second most represented species and was preyed upon at least in summer and autumn. Currently, its annual home ranges vary from 1 km2137 to 94.8138. Some populations can be sedentary, and others, migratory with groups occupying generally low elevations in winter and high ones in summer139,140. The fact that they were not hunted during spring may therefore reflect their absence near the site during that season (due to altitudinal seasonal movements, for example). Alternatively, the Neandertal groups may have specifically avoided this species at this season. The dietary signal of C. elaphus overlap greatly with the one of Ca. pyrenaica which may indicate that they have foraged on the same plants as sympatric species at least, around Axlor like it can happen nowadays141,142.

At least 9 cf. B. priscus were found in the level D of Axlor. Depending on the extent to which the Neandertals exploited the carcass, this species could have been the main contributor to the meat supply in the human diet (look for example Uzunidis et al.23). Currently, the eco-ethology of bison is strongly constrained by human activities. Nowadays, most European bison populations are sedentary but can move seasonally up to 5 km beyond their core territory143. Only the American bison of Yellowstone still undergo a true migration over a distance of about a hundred kilometers144,145. In this area, migration begins in early spring (March) when the bison head toward higher-altitude areas and may continue for part of the summer. In autumn, the herds descend into the valleys to avoid winter snowfalls144,146. During the Pleistocene, the few populations whose migratory behavior has been investigated appear to have been sedentary147,148,149,150. The reconstruction of the season and of the duration of the death events indicate that cf. B. priscus was preyed upon during spring and autumn while summer was avoided. Such predation agenda would be compatible with migratory bison herd roaming in the Cantabrian-Mediterranean altitudinal corridors. Thus, at the transition between MIS 5 and MIS 4, it is during their passages near Axlor in spring and autumn, that Neanderthal groups may have hunted few individuals. This hypothesis remains, however, to be tested using other tools such as strontium and oxygen isotopic analysis. Alternatively, summer corresponds to the time of year when bison gather in large mixed herds for mating99,151. During the rut period, the males become overly aggressive and their fat reserve drop99,150,152. Neandertal hunters may have deliberately avoided bison cohorts at this period. Autumn may have been favored to preyed upon bison, which are then at their optimal weight and have their new winter coat at this time99,152,153.

Two dietary groups have been observed at the period of the death of the individuals: the first one being very homogeneous and characterized by a very low number of both pits and scratches, the second one being heterogeneous with a slightly higher number of pits and a number of scratches ranging in between the browser to the grazer areas. The group 1 diet might correspond to the autumn season when bison tend to eat more dicots while the group 2 might match the spring diet, which is richer in monocots154,155. However, the diet of bison is highly flexible, and comparative studies in similar environments would be necessary to confirm this point.

The reconstruction of the periods of deaths of the herbivore at Axlor l.D reveals that Neandertals were present throughout the warm period (Figs. 5 and 6). Depending on the season, different preys were hunted, reflecting either a choice by the human-predators or the unavailability of some species. Thus, the wild goat was hunted through the whole warm season, complemented by the horse and the red deer in summer and autumn and the bison in spring and autumn with punctual predation on chamois, roe deer and rhinoceros.

Neandertal groups occupied Axlor Level D during the warm season, either for long stays or for short but repeated stays. Similar patterns characterized by several consecutive season occupations were recorded in the Mousterian sites from Cantabrian region. In Covalejos level 8 the herbivores were hunted throughout the whole year and from winter to summer in levels 7 and 422. In El Castillo level 20, the predation occurred between autumn to spring19. Contemporaneous sites from the same region, such as Cueva Morín and El Pendo, however, display a more restricted occupation patterns from late autumn to winter19.

In Axlor l.D, the stability of the environment may have allowed at least Ca. pyrenaica to maintain itself in the vicinity of the site while the two main other represented species cf. B. priscus and C. elaphus were not hunted respectively in summer and spring reflecting their seasonal movements or the hunters’ choice. During autumn, the prey diversity reaches its peak which may reflect the geographical position of Axlor, as a junction point in the seasonal movements of herbivores, and consequentially for their predators.

Lithic tools were imported to the site in the form of flakes or as already shaped quina or semi-quina sidescrapers36,39. Intensive curation was carried out on these tools in the settlement, as evidenced by the presence of numerous bone retouchers and resharpening flakes37,40. The functional evidence of this type of toolkit (Quina sidescrapers and Quina resharpening flakes), in west Europe suggest that the activities carried out were not limited to the capture of animals, but included broader processing of carcasses and other types of activities on other materials, such as wood39,156,157,158. These activities allow us to infer that the occupations of Axlor level D reflect an intermediate type of settlement between specialized camps and residential camps8, similar to what has been suggested for other Quina sites159.

Conclusions

Neandertals’ subsistence strategy for animal-derived resources acquisition in Level D of Axlor mainly relied on the predation of Ca. pyrenaica, C. elaphus, and cf. B. priscus, occasionally supplemented by the acquisition of E. caballus, R. pyrenaica, and C. capreolus. Their regional and local habitat were similar and largely dominated by dicots, corresponding to a heathland or a forest with small patches of grasses. Capra pyrenaica was hunted throughout the warm season. In summer and early autumn, C. elaphus was probably hunted in the same habitats. The hunting of large herbivores, such as cf. B. priscus, occurred in spring and autumn. The absence of predation on certain species during specific seasons may reflect either their seasonal movements or the selective choices of hunters. Specifically, the hunting of bison in spring and autumn might suggest that, at that time, Axlor was located along a migration route for the species. Alternatively, sedentary bison might have been deliberately avoided by hunters during the summer, perhaps due to the rutting season. This study raises the question of bison behavior during the Middle Paleolithic in North Atlantic Iberia: were they migratory or sedentary?

Neandertal groups occupied Axlor Level D during extended periods or short, repeated stays during the warm season. The specific diversity (e.g., B. priscus, E. caballus, C. elaphus, Ca. pyrenaica) documented for autumn suggests that, particularly during this period, the site was likely situated at the intersection of the seasonal movements of these ungulates, making it particularly attractive to predators. The site of Axlor may thus have been part of a complex territorial management strategy by Neandertal groups in North Atlantic Iberia, where its geographic position facilitated faunal resource procurement such as meat, bones or skin during the warm season.