Introduction

The use of bone as tools is a behavior unique to our lineage. The earliest evidence for bone technology comes from South and East African sites, some of which were dated to 2.4 Myr1,2,3,4,5,6,7,8,9,10. The recent discovery of a large assemblage found in situ within a single layer of the Oldupai (originally misnamed as Olduvai) T69-Complex suggests that, by 1.5 Myr, some hominin groups implemented a standardized technological strategy that entailed the selection of specific skeletal elements from hippos and elephants, their systematic modification by direct percussion to produce a recurring shape and the use of these tools in percussive and compressive activities5. Our understanding of the role played by the earliest bone technologies in Pleistocene cultural systems remains nonetheless fragmentary. In Europe, for instance, discoveries made over the last decades leave no doubts that Neanderthal and pre-Neanderthal groups fully understood the properties of osseous materials and frequently exploited them to make tools. This assertion is supported by evidence for the use of bone or antler to detach flakes from cores, to shape and retouch the cutting edges of stone tools, to produce bone tools with cutting edges shaped by knapping, and to transform more pliable materials11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35. Among this latter category, the most frequent functional hypothesis to explain the presence of recurrent sheen and rounding on the tip of rudimentary objects pertains to their suggested use as smoothers for hide processing activities11,20,21,22,23,24,28.

Initially, the discovery of hide processing tools was interpreted as indicative of an innovation at the end of the Neanderthal trajectory where ribs were targeted for this task owing to their standardized distal morphology20. Similar finds at Châtelperronian (45–40 ka cal. BP) sites from France suggested a continuity, both biological between the authors of the last Middle Paleolithic industries and the Middle to Upper Paleolithic transition industries, and cultural, between the Mousterian of Acheulean Tradition – Type B (MTA-B) and the Châtelperronian36 lithic technocomplexes. However, research conducted in the last decade now demonstrates this innovation is likely much older than previously thought and not limited to Southwestern Europe. Several unmodified bone fragments bearing sheen and rounding at one end and found at sites located in Germany, Spain, Italy, Central Asia, Altai, China and North Africa from contexts dated to 380 to 45 ka were interpreted as expedient tools used for hide working11,28,37,38,39,40,41,42,43.

Unmodified bone tools have traditionally received limited focus by archaeologists. Historically, skepticism inherited from the reinterpretations44,45,46,47 of the Alpine Mousterian48 and the so-called Osteodontokeratic Culture49 encouraged the development of taphonomic approaches in archaeology46,47,50,51,52,53 but delayed formal investigations in the potential role played by rudimentary osseous technologies in the implementation of subsistence activities. Ethnographic data nonetheless suggests that unmodified bone fragments may be used for a wide variety of tasks e.g., 54, including butchery and carcass processing e.g., 55 as well as acquiring and transforming plant materials e.g., 56. Yet, archaeologists interested in documenting early bone technology face two major challenges. First, recognizing minimally modified tools is a challenging task. Although experimental archaeology allowed the definition of clear criteria to identify bone fragments intentionally modified by knapping1,57,58,59,60, detecting unmodified bone fragments used as tools in large faunal assemblages primarily relies on the recognition of clear use wear patterns. Unfortunately, during the transition from the biosphere to the lithosphere, taphonomic processes may alter, even erase, use wear patterns, and at times, produce confounding surface modifications similar to anthropogenic modificationse.g., 32,61,62,63,64,65,66,67,68,69. Thus, distinguishing between natural and anthropogenic bone surface modifications therefore constitutes a first challenge that must be overcome. To surmount this limitation, one must compare the surface modifications potentially originating from use with those likely resulting from taphonomic alterations on the same object and/or on associated faunal remains. Second, establishing the function of these bone tools in reproducible and reliable ways is not an easy task. To achieve this goal, experimental archaeology allows the creation of comparative samples in controlled settings that may then be used to interpret use wear patterns preserved on archaeological specimens70. Historically, the focus given to butchery and carcass processing activities e.g., 71,72,73,74,75 has biased our interpretations on the role played by unmodified or knapped bone technology in past subsistence. Indeed, items bearing smoothed, rounded, glossy surfaces are almost systematically interpreted as tools used in hide processing activities11,20,21,22,23,24,36. Yet, hide processing can be achieved in multiple ways, e.g., on fresh, dry or rehydrated skin, with or without abrasive, and each variant results in use wear patterns that differ in nature and extent, e.g., 75,76,77, and see 78 for a synthesis and an experimental demonstration. Furthermore, experimental data demonstrates that the use of bone fragments in other carcass processing activities may produce wear patterns qualitatively similar to those resulting from hide processing28,37,59,75,78,79. To overcome this interpretative limitation, efforts must be pursued to increase the number of activities represented in comparative samples and to develop the field of quantitative use wear study to reproducibly test the functional hypotheses put forward via qualitative comparisons.

To that end, the recent analysis of the beveled bone tools from Sibudu exemplifies the pertinence of combining qualitative and quantitative approaches in use wear analysis to overcome these two challenges. Rather than serving as hide processing tools, it was shown that these objects were used to debark trees and, occasionally, dig in humus-rich sediments between 80 and 60 ka80. Similarly, Ma et al.78 demonstrated that it was possible to accurately set apart carcass processing activities and several hide processing variants by combining qualitative and quantitative approaches. A similar approach allowed to document the use of bi-pointed object as wool weaving implements and the use of bone awls in textile working activities at the Late Neolithic site of Cueva del Torro, Spain81, and identify threshing sledges used for processing large quantities of cereals during the Early and Middle Neolithic in Greece82. Here, building on these methodological advances that rely on quantitative surface texture analysis, we document a shaft fragment of a reindeer femur bearing, at one end, a highly smoothed area. The object was found at the Abri du Maras, layer 5.1, a context dated between 132 and 114 ka, which corresponds to the Marine Isotope Stage (MIS) 5 (Fig. 1). We apply discriminant analyses of surface textural data (1) to confirm that the smoothed area originates from the use of the tool rather than from taphonomic alterations and (2) to establish the function of this unique object through comparisons with experimental specimens used in a diversity of tasks. Our results show that the tool was likely used to flay carcasses, i.e., an activity that aims to detach the skin from the body of an animal without damaging it. Our results demonstrate the potential of quantitative use wear studies to overcome limitations inherent to uncovering the role of rudimentary bone tools in past cultural systems. In doing so, they provide a new outlook at Neandertal subsistence strategies by providing a means to investigate the transformation of perishable materials during the Paleolithic.

Fig. 1
figure 1

Modified from Richard et al.83 (c).

Location, excavated area and stratigraphy of the Abri du Maras. Location of Abri du Maras in the middle Rhône Valley, France (a). Grid of the excavation with the location of the portions excavated by R. Gilles (light grey), J. Combier (medium grey) and M.H. Moncel (dark grey). The red square indicates the location of the bone tool (b). Stratigraphic sequence, age estimates per layer and location of the bone tool. A mean age was calculated when several ages were obtained using the same method on the same layer (right). Individual ages (one age per method available) are shown on the left (in italic: maximum age). Note that within a layer, the ages are not necessarily represented following stratigraphic order. The OSL, IRSL and ESR/U-series ages are presented at 1 σ; radiocarbon (14C) and U-series ages at 2 σ.

Full size image

Archaeological context

The Abri du Maras is a site located in the Ardèche Gorges on the southeastern margins of the Massif Central (Fig. 1). The Abri du Maras is a largely collapsed rock shelter. The first excavation at the site occurred in the 1950s and 1960s by R. Gilles and J. Combier84. Since 2006, new excavations of an 88 m2 area revealed an up to 7 m-thick multi-layered stratigraphy subdivided into six units numbered from 1 (top) to 6 (bottom) based on geological and sedimentary observations. Two main phases of roof collapse of this vast cavity are recorded between layers 2 and 3 on the one hand, and layers 4 and 5 upper on the other hand. Most of the archaeological material was recovered in layers 4, 5, 5 upper and 6. Layer 4 was subdivided into two main occupation phases, i.e., 4.1 and 4.2, based on vertical and horizontal variations in the density of archaeological remains identified during post-excavation spatial analyses. The upper part of layer 5 (layer 5 upper) was subdivided into three main occupation phases for the same reasons, i.e., 5.1, 5.2 and 5.385,86. The object reported here was unearthed during the 2017 excavation of level 5.1 in square L6 (x: 70; y: 47; z: 399) and was attributed the accession number Mar’17-L6-5-1539 (Fig. 2). Recent multi-method dating and associated Bayesian age modeling of layer 5.1 suggests that its formation occurred between 132 − 114 ka (optically stimulated luminescence, OSL, on quartz)87 and 114 − 105 ka (electron spin resonance combined with uranium-series, ESR/U-series, on tooth enamel)88, which makes it coeval with MIS 5.

Fig. 2
figure 2

The bone tool MAR’17-L6-5-1539 found at the Abri du Maras. General (a) and close-up views (b-g) of the worn area on the cortical (b, d-e) and medullar (c, f-g) surfaces observed in low-magnification using a Leica Z6 APOA stereomicroscope with incident light on the object (b, c) and transmitted light on resin replicas (d-g). The general views were photographed with a Sony A6400 equipped with a Sony E 30-mm F3.5 macro lens. Scales = 1 cm (a), 5 mm (b, c) and 1 mm (d-g).

Full size image

For layer 5-upper, lithic technology primarily relies on flint sourced from local and semi-local outcrops near the site, with additional material coming from more distant southern sources. The lithic assemblage suggests a long-term occupation, characterized by diverse core reduction methods and high frequency of retouched flakes compared to the upper part of the sequence. The numerous tiny retouch flakes attest to a high consumption of flakes of various sizes at the site, including some that were knapped elsewhere. Notably, the assemblage includes semi-Quina tools and Quina products85, which raises questions about the definition of a Rhodanian Quina facies. In situ core reduction strategies encompass discoid, orthogonal, multidirectional and Levallois methods, along with cores on flake. Evidence of Alnus roots working is documented in the layer89.

From a zooarchaeological perspective, the occupation of level 5.1 at Abri du Maras presents characteristics typically associated with residential camps. The presence of Rangifer tarandus (74%), Cervus elaphus (11.3%), Equus ferus (10.7%), Bison priscus (2.8%), Capreolus capreolus (0.6%) and Sus scrofa (0.1%) suggests an open landscape with forest patches which likely provided Neanderthal groups with a wide range of opportunities to carry out their hunting and foraging activities90. Although isolated teeth constitute the majority of identified remains by taxon, the skeletal element representation for the dominating species, i.e., Cervids, reveals an abundance of appendicular elements with post-cranial axial elements being less frequent. Mortality profiles indicate a preference for hunting prime adult individuals across all species. Seasonal evidence suggests reindeer were hunted in summer while horse hunting occurred throughout summer and into early autumn90.

Bone surface preservation on specimens recovered from layer 5.1 is generally good, with more than three quarter of the assemblage (78.6%) displaying intact surfaces over two-thirds or more of the cortical bone. Root etching is the most frequent post-depositional alteration affecting more than a quarter (28%) of the remains. Small fissures and desquamations were recorded on 8% of the specimens, which indicates the subaerial exposure of the faunal assemblage for a moderate period prior to its burying. Black stains of manganese oxide in the form of dendrites are common (20%). Orange spots caused by iron oxide precipitation are also recorded albeit at a lower rate (3.8%). Carnivore modifications are scarce (0.7%).

Anthropogenic modifications are frequent on the faunal remains recovered in level 5.1: 32.8% display signs of exposure to heat, 8.3% bear cut marks and 8.3% of the fragments retain percussion marks likely produced when fracturing the bone to access marrow. Green fractures are present on most bone fragments which lends further support to marrow extraction activities taking place at the site. A total of 18 bone fragments exhibits well defined areas on their cortical surfaces with overlapping pits and linear marks indicating their use as retouchers for shaping lithic cutting edges90. The abundance and distribution of cut marks suggest that hominins had primary access to animal carcasses. Furthermore, numerous reindeer metapodials bear cut marks that appear to be related to the removal of the skin and tendons. Overall, the faunal assemblage indicates that Abri du Maras functioned as a long-term seasonal campsite, primarily occupied during the summer, and occasionally into early autumn. High-caloric portions of prime adult ungulate carcasses were transported over relatively short distances and processed at the site.

Results

The MAR’17-L6-5-1539 specimen (Fig. 2a) is a shaft fragment of a reindeer left femur measuring 133.3 mm in length, 26.9 mm in width, 13.7 mm in thickness and 4.0 mm in maximum cortical thickness. It preserves the antero-lateral face of the midshaft and one-third of its total circumference. All fracture planes are smooth; they are fairly abrupt on one side and curved with an oblique angle on the other. On the cortical surface of the anterior aspect, an isolated flake removal scar is interpreted as resulting from the breakage of the bone to access marrow. Together, these features indicate the femur was broken while still being fresh. Localized ferro-manganese oxide aggregations in the form of dendrites and coloration by precipitations are visible on the bone surfaces (Fig. 2a). Evidence of root-etching is recorded on the cortical and medullar aspect of the object (Fig. 2a). Incipient fissures oriented longitudinally along the femur’s main axis (Fig. 2a) suggest a moderate weathering of the object resulting from subaerial exposure, i.e., stage 150. These taphonomic alterations do not together prevent the recognition of a rounded, worn surface located at the distal tip of the object.

Qualitatively, unmodified surfaces (Supplementary Figures S1, S2, panels starting with N) display features consistent with a slightly weathered, i.e., type 1, bone surfaces. Their microtopographic features remain largely intact with seldom evidence of rounding of the peaks and plateaus. They present a grainy aspect and fissures likely caused by weathering are present. The vascular canal openings show slightly crenulated edges and are filled with small-sized sediment particles. These surfaces differ from those that are rounded and worn at the tip of the object. The latter are homogenous and seldom preserve the microtophography of unmodified bone surfaces. The peaks and plateaus are smooth and substantially reflective with numerous preferentially oriented micro-striations (see below).

Quantitative surface texture comparison confirms that the worn area differs significantly from taphonomically altered surfaces on the same specimen regardless of the surface type, i.e., cortical, medullar or fracture planes (Supplementary Figures S3–S10; Supplementary Tables S1–S7). This result is particularly evident when considering the root-mean-square height (Sq), the skewness of height distribution (Ssk), the inverse areal material ratio (Smc), the surface complexity (Ymax_S), and the mean depth of furrows (Mea_Dep_Furr) that systematically yielded lower values for the worn surfaces compared to the taphonomically altered ones. While the depth of furrows is lower for the worn surfaces, their density is higher, which attests to a distinct life-history for the distal end relative to the remainder of the object. Using the 14 surface texture variables identified as discriminative and not highly correlated (Supplementary Text S2; Supplementary Table S1), the discriminant analyses accurately distinguish between “worn” and “taphonomic” surfaces more than 90% of the time for the cortical and medullar surface (cortical surface: LDA accuracy = 91.30%, LDA kappa = 79.46%, CVA accuracy = 92.41%, CVA kappa = 82.06%; medullar surface: LDA accuracy = 100%, LDA kappa = 100%, CVA accuracy = 94.87%, CVA kappa = 89.17% (Supplementary Figs. S11a–d; Supplementary Tables S2–S5). Discrimination between “worn” and “taphonomically altered” fracture planes yields lower accuracy results albeit remaining above 60% (LDA accuracy = 81.25%, LDA kappa = 61.29%, CVA accuracy = 66.67%, CVA kappa = 33.61%) (Supplementary Figs. S11e, f; Supplementary Tables S6, S7). Thus, the qualitative differences observed between the “worn” area and the rest of the object combined with the variation in quantitative surface texture data support an anthropogenic rather than a taphonomic origin for the alteration.

Through the naked eye, the worn end displays a bifacial flattening of its surface in the form of a bevel (Figs. 2b, c). Deformations and compaction of the cortical bone are visible on the cortical and medullar aspect of the object as well as on the fracture planes. The edges are rounded and worn, and their smooth surfaces gives them a matte, uninterrupted sheen. On the cortical surface, this wear pattern extends up to 21.4 mm from the apex while it is slightly less invasive on the medullar surface with a maximal extent of 19.7 mm. The small difference in wear extent recorded between the cortical and medullar surface suggests that both faces of the distal end were simultaneously in contact with the worked material. Two micro-flake removal scars are present near the center of the apex (Figs. 2b, c,d, f). The one on the medullar aspect displays a patina that differs from the adjacent surfaces and the rest of the object (Fig. 2c), which suggests a recent damage. Conversely, the one on the cortical surface presents a patina similar to adjacent areas which suggest this damage likely occurred prior to the abandonment of the object at the archaeological site (Fig. 2b). No evidence of distal and proximal compression or flaking typical of intermediate pieces, e.g., [91,92, were detected.

At low magnification, the working edge is highly smoothed and display a homogeneous surface, i.e., with minimal differences between the high and low points, without striations (Fig. 2c). The vascular canal openings usually found on intact bone surface are no longer visible. On the cortical aspect of the object (Figs. 2d, e), the density of micro-striations increases the further we move away from the apex. While this pattern may appear counterintuitive, it suggests the abrasive properties of the worked material differed between its primary and secondary contact with the tool’s working edge. The micro-striations are variable in width and display smooth outlines with few or no hertzian cones. The finest ones are preferentially oriented along the main axis of the object while the larger and deeper ones are obliquely oriented relative to this axis. Irrespective of their size, the striations overlap one another without clear recurring patterns in their relative arrangement. Several micro-striations overlap the flake removal scar located at the center of the apex on the cortical surface and extend on the adjacent cortical surface by a few millimeters. They display similar orientation and morphology as those located on the adjacent rounded primary contact surface of the tool. While the origin of the scar remains ambiguous–it could either result from use, transportation, trampling, etc.–, the wear pattern on it suggests the object continued to be used for the same activity after the damage occurred. On the medullar aspect of the object (Figs. 2f, g), subparallel micro-striations extend from the right lateral edge along the oblique fracture plane. Their size is relatively homogeneous although coarser striations become visible the further we move away from the apex.

At high magnification, smoothed surfaces display evidence of intense flattening and homogenization (Fig. 3a; Supplementary Figures S1–S2). This pattern is particularly evident with the partial filling of the bone microstructure. Variations in texture (Fig. 3b–d), especially the rugged edges between dales and plateaus on the fracture planes, likely results from weathering. Nonetheless, micro-striations are preserved on flat areas where the bone structure is more homogeneous (Fig. 3a-d; Supplementary Figures S1–S2). As we move away from the apex, the micro-striations become more visible despites the alteration of their contours by taphonomic processes (Supplementary Figures S1–S2). Their width varies between 3 and 12 μm and their depth rarely exceeds 1.5 μm. The range of their length is difficult to establish because the majority are interrupted by taphonomic alterations. Nonetheless, the best-preserved striations often extend beyond the field of view, which entails length surpassing 320 μm. In cross-section, they display an open U-shaped profile, with rare instances of V-shaped profile. On both the cortical and medullar aspect of the tool’s mid-shaft, fine micro-striations running in parallel and perpendicularly to the tool’s main axis are visible on flattened peaks and plateaus (Fig. 3e, f; Supplementary Figures S1–S2). They are found along some scattered, shallow, rounded depressions that can neither be attributed to the material structure nor to taphonomic processes. Owing to their location on the object and their morphology, these micro-striations and depressions appear consistent with wear resulting from handling described in the literature93,94,95,96. However, further experiments are required to generate a comparative sample that would allow to quantitatively test this hypothesis and establish their precise origin. Overall, the observations at low and high magnification–heavy rounding of the edges and flattening of the cortical and medullar surfaces, obliteration of the original bone structure, potential traces of handling–suggest the tool was in a prolonged contact with soft pliable material64,75,77,97,98.

Fig. 3
figure 3

Photo-simulations of the use wear patterns documented on the bone tool from the Abri du Maras. Sample of use wear patterns observed at high magnification via confocal microscopy on the fracture planes (a) and smoothed edges (b) as well as the cortical (c) and medullar (d) surfaces. Possible handling traces were also recorded on the cortical (e) and medullar (f) surfaces. See Supplementary Figures S1, S2 for the exact location of the acquisitions on the object.

Full size image

Comparison of the wear pattern recorded on the Abri du Maras specimen with those that developed in controlled experimental settings allow to rule out several activities as its potential origin (Supplementary Text S4; Supplementary Table S17). The archaeological pattern differs markedly from those generated by tasks that involve abrasives–such as digging in soil, processing plants or treating hides with ash, sand, or ochre–which produce dense, deep and uniformly oriented striations that cover the tool surface (Fig. 4h-l, n). On the Abri du Maras specimen, the striations are sparsely distributed and significantly shallower indicating that the tool was unlikely used in such abrasive contexts (Fig. 4a-c). Although the observed wear resembles that generated while flaying (Fig. 4f), experimental use of bone tools on fresh or rehydrated hides without abrasive typically results in both striations and some isolated pits78, the latter being absent on the archaeological specimen (Fig. 4e, i).

Quantitative surface texture analyses were implemented to establish more precisely the function of the tool. On the experimental specimens, 21 variables were identified as being apt to discriminate between the various activities included in the comparative sample (Supplementary Table S1). They include the auto-correlation length (Sal), the texture aspect ratio (Str), the texture direction (Std), the mean hill area (Sha), the hill count (Shn), the mean hill roughness (Shrn), the standard deviation of hill roughness (Shrnq), the standard deviation of dale roughness (Sdrnq), the length-scale Y-max (Ymax_L), the new length-scale anisotropy (New_epLsar), the heterogeneity of the fractal complexity (HAsfc81 and HAsfc), the mean density of furrow (Mea_Den_Furr), and the first three directions of the texture (Dir_1st, Dir_2nd and Dir_3rd) (Supplementary Figures S12–S27). The LDA results suggest the use wear pattern present on the object corresponds to those produced experimentally in activities that entailed a prolonged contact with soft animal tissues (Supplementary Figure S28a, b; Supplementary Tables S8–S10). To refine the functional interpretation, both LDA and CVA point to carcass flaying as the most probable activity (Fig. 5, Supplementary Figures S28c-f–29, Supplementary Tables S11–S16). This classification is based on a high posterior probability (Fig. 5; Supplementary Figures S28c, d; Supplementary Tables S11–S13), supporting a robust statistical match with experimental flaying traces. While we acknowledge the possibility that the tool may have been used in an untested activity, the strength of the discriminant classification, combined with the distinctive absence of abrasive-related features, makes flaying the most parsimonious explanation. The few discrepancies between the archaeological specimen and the experimental sample likely reflect variability in wear caused by the contact with fresh meat and the effects of post-depositional processes.

Fig. 4
figure 4

3D rendering of natural and worn surface texture on experimental bone tools and the Abri du Maras specimen. Comparison between the use wear pattern observed on the Abri du Maras specimen (a-c), the natural surface of a reindeer femur curated in the PACEA comparative anatomy collection (d), and the experimental acquisitions curated in the ExOsTechBank (e-n). The experimental acquisitions were done on bone tools used for fleshing rehydrated deer (Cervus nippon) hide (e), flaying carcass (f), cutting fresh meat (g), scrapping rehydrated cow (Bos taurus) with ash (h), scrapping fresh hide with marrow (i), ochre (j), and sand (k), scrapping dry cow hide with sand (l), debarking pine branches (m), and digging in humic sediment (n).

Full size image
Fig. 5
figure 5

Discriminant analyses results. Linear discriminant (LDA) (a) and canonical variate (CVA) (b) analyses results distinguishing different activities involving soft animal tissues (Flaying: flaying carcass; FreMeat: cutting fresh meat) with the projection of surface texture data acquired on the Abri du Maras specimen (black stars). See Tables S11-S13 for the corresponding confusion matrices and model performance statistics.

Full size image

Discussion

The tool discovered at Abri du Maras, level 5.1, provides further evidence of Neandertals using unmodified bone fragments during the MIS5. Our discovery opens a new perspective on the subsistence strategy implemented by the human groups that occupied the Rhône Valley and on the cultural adaptations they resorted to in their daily and seasonal activities. Textural and discriminant analyses indicate that this object was most likely used in flaying activities that produced a use wear comparable to our experimental sample. Interestingly, the function of the Abri du Maras tool is not linked to hide processing, with which similar objects, e.g., the objects interpreted as smoothers (fr. lissoirs), have been traditionally associated. The identification of this object is entirely consistent with ethnographic and archaeological data. Comparable ethnographic examples from the Algonquian Nehiyawak and Nakawēk nations of North America reveal that elongated unmodified bone fragments were sought after for flaying carcasses because they could be inserted between the skin and the meat to efficiently detach these soft animal tissues without piercing or cutting the hide in the process55, a risk that would increase if lithic tools were used to perform this task.

From an archaeological perspective, the presence of an object used for flaying carcasses is entirely compatible with the interpretation of the Abri du Maras being a long-term occupation site visited during the good season. Throughout the summer and the beginning of the fall, the Neanderthal visitors could intercept the migratory species that were present in the vicinity of the site. The targeting of prime adult individuals and processing of caloric-rich body parts at the site, especially breaking the bones to access marrow, would likely have produce a wealth of bone fragments of various size and morphology from which to choose an ideal specimen that had the affording characteristics to flay carcasses. While the selection of a reindeer femur shaft fragment as tool is likely explained by their relative abundance in the faunal assemblage90, we cannot rule out that this choice of raw material as tool to remove the skin of carcasses, including a majority of reindeer, may reflect particular symbolic aspects inherent to the relations between Neanderthals’ technology, their subsistence strategy, and the environmental contexts they exploited99,100,101,102.

The preferential introduction of specific body parts at the site entails that the initial phases of the butchery and carcass processing activities didn’t take place at the site but rather in its vicinity. The highly smoothed edges of the object suggest that the flaying tool was used over an extended period. As such, it is reasonable to hypothesize that it was likely transported on several occasions in the toolkit carried by an individual who participated in daily hunting or lithic procurement trips through the season of occupation at the site. This scenario not only implies that Neanderthals understood the technological potential of bone and took advantage of it. It also highlights their capacity to plan their technological needs ahead and select lightweighted items that could efficiently perform the tasks that would arise following a successful kill.

This discovery has also important implications for the evolution of clothing and the production of bags or other leather items, particularly among Neanderthals. The use of a bone tool to avoid perforating the hide suggests that Neanderthals took care to preserve impermeable skins, likely to produce waterproof leather items, including clothing. Additionally, this latter practice implies that such garments were made from large or complete hides, as perforations would have been less problematic for clothing constructed from smaller pieces, where undamaged sections could be selectively used99,103. The emphasis on maintaining intact hides further supports the idea that Neanderthals developed strategies to maximize the functional properties of their garments. Furthermore, at the same site, though in a more recent layer, there is evidence of a complex twisted thread made from tendons and plant fibers by Neanderthals104,105. This suggests an ability to manufacture robust clothing or bags, reinforcing the notion that Neanderthals not only processed hides with care but also had the technological means to assemble durable and functional leather items.

Finally, our results highlight the fact that osseous materials are particularly apt to keep a record of object-matter interactions, especially when these interactions are made in the context of anthropogenic subsistence activities. The methodology we implemented relies on combining two complementary approaches, i.e., the traditional qualitative use wear documentation and the emerging quantitative use wear analyses relying on surface texture data. These two approaches go hand in hand. While qualitative observation may help identifying use wear patterns and propose a first hypothesis on the nature of the material the bone tool encountered, the quantitative methods are crucial to confidently assess that the purported use wear indeed differ from taphonomic alterations, and to narrow down the type of human action that most likely led to its development. In the future, it would be worth reassessing the numerous objects interpreted as smoothers with the complementary methods presented here to test whether they were indeed used for processing hides or served to fulfil other tasks.

Materials and methods

Fieldwork at the Abri du Maras follows standard archaeological excavation protocols including the three-dimensional recording of stone tools, large mammal bones, visible features, etc. Smaller finds are bagged by 1-m2 unit of provenience. Sediments are dry-sieved using a 2-mm mesh screen. The bone tool documented here was identified by one of us (JMH) during the zooarcheological analysis of the faunal assemblage recovered from level 5.1. Taxonomic and skeletal element identification was carried out by comparing the bone fragment with faunal remains from the zoological reference collection curated at the Muséum National d’Histoire Naturelle, Paris, France. Morphometric data, i.e., maximum length, width, and thickness, were collected using a digital caliper. Anthropogenic modifications were distinguished from natural ones based on criteria available in the literature46,47,50,51,52,69,106,107 and via quantitative comparison of surface texture data (see below). The object was photographed with a Sony A6400 equipped with a Sony E 30-mm F3.5 macro lens to supplement observations achieved through the naked eye. Low magnification (4x to 40x; field of view ranging between 14.84 × 11.07 mm to 2.36 × 3.17 mm) microscopic observations were conducted using a motorized Leica Z6 APOA equipped with a BFC420 digital camera linked to a LAS Montage and Leica Map DCM 3D computer software at the PACEA laboratory. These observations were done both on the original object in incident light and on transparent resin casts of the object in transmitted light. High-resolution microscopic observations and surface acquisitions were obtained using a MarSurf CM mobile confocal microscope driven by MarSurf MSW 8.6 software. This equipment was used with the aim of characterizing the roughness of worn and unworn areas and producing 3D renderings of the used surface. Acquisitions were done using a 50x magnification lens with a working distance of 10.6 mm (field of view: 323,4 × 323,0 μm). To ensure quality, surfaces with less than 95% measured points were systematically re-acquired. Post-acquisition treatment was carried out using the Mountain View 8.2 software and followed a procedure adapted from Ma et al.78 and Mazzucco et al.82. First, a data augmentation procedure was implemented to better capture the range of variation in surface texture. Consequently, each acquisition was subdivided into five overlapping areas measuring 200 × 200 μm respectively located at the four corners of the original acquisition without touching the edges and at its center. Then, using built-in operators, each sub-area was processed using standard protocol that entails levelling the surface (least square method), removing outliers (both isolated and close to the edge), removing points outside of 0.01% and 99.99% threshold for height distributions, filling-in non-measured points (interpolating values from neighbors), removing form (polynomial of fifth order) and applying metrological filter (Gaussian 25 μm) to distinguish between the waviness (S-F) and the roughness (S-L) of the object. Roughness parameters (ISO 25178), fractal parameters (SSFA), furrow analysis parameters and texture direction and isotropy parameters were extracted on the S-L surface (Supplementary Table S1 for definitions; Supplementary Text S1). Parameters expressed in percentages and ratios were transformed into linear variables centered around zero using logit transformation108,109 to allow their use in multivariate analysis (see below). Likewise, angle values were decomposed into their corresponding sine and cosine vectors for the same reason110,111 Surface texture images were produced using the ‘viridis’ color ramp and photo-simulation operator to help visualization for colorblind individuals.

A preliminary data exploration was done to remove variables with missing values (Supplementary Text S2). To ensure that the worn area resulted from the use of the bone fragment rather than from taphonomic alterations, they were first compared with unworn intact areas along the diaphysis on the cortical and medullar surfaces as well as on the fracture planes. For linear data, the Shapiro-Wilk test and Levene’s test were used to respectively assess the normal distribution and homogeneity of variances for each texture parameters. Depending on the distributional assumptions, one-way ANOVA with Tukey HSD pairwise comparisons or a Kruskal-Wallis with Bonferroni-corrected Dunn post-hoc tests were performed to assess the ability of the linear texture variables to capture differences between categories. For circular variables, uniformity tests included Rayleigh, Kuiper, Watson, and Rao spacing tests and group comparisons were performed using the Mardia-Watson-Wheeler and Watson-Williams tests with pairwise post-hoc comparisons adjusted via Holm correction. Parameters that displayed significant differences between worn and unworn surfaces (p < 0.05) were retained and those highly correlated were removed (threshold: R2 = 0.7). Principal component analysis (PCA), linear discriminant analysis (LDA) and canonical variate analysis (CVA) were performed to securely distinguish between worn and taphonomically altered areas based on their surface texture data. LDA was applied directly on the surface texture data–and on transformed data for circular, ratio and percentage variables–with stratified training/testing splits (respectively 70% and 30% of the sample size) and three separate 11-fold cross validations were used as resampling scheme. CVA was performed on the five first PCs of the PCA-transformed data with 101 rounds of cross-validation with a Jackknife resampling scheme to assess group separability. The model’s performance was evaluated using classification accuracy, confusion matrices, per-class statistics, and visualization of single-LD (or -CV) density plots or pairwise-LD (or -CV) scatterplots by groups with 95% confidence ellipses.

Second, to establish the most likely function of the object, worn surfaces were compared using the same procedure with an experimental sample curated in the ExOsTechBank, i.e., a database of worn surfaces recorded on archaeological, experimental and ethnographic bone tools which currently includes 10,772 surface acquisitions (28 October 2025). Supplementary Text S3 provides a summary of the activities represented in the ExOsTechBank and the acquisitions used in the present research. For the present study, six variants of hide working were selected: rehydrated deer hide without abrasive, dry hide with sand, fresh hide with ash, sand, ochre, and marrow. Additionally, experimental use wear patterns generated while flaying carcasses, cutting fresh meat, debarking pine branches, and digging humic soil were also considered (Supplementary Text S4). The comparative experimental sample selection was based on the initial qualitative description of the use wear patterns, the comparison of similar patterns reported in the literature e.g., 37,75,77,78,97,112, and the activities that are coherent with the archaeological context in which the specimen was found. The comparison between the archaeological and experimental samples followed a two-step process. A first attempt was done by comparing the use wear with broad categories of worked material, i.e., hide (irrespective of their state or the use of abrasive and tannins), animal soft tissues (flaying and cutting meat), miscellanea (debarking and digging in sediments). A second attempt aimed to refine the category of activity by comparing the use wear with those included in the two categories that were most often predicted. To avoid sample imbalance in the discriminant analyses, the training sample includes 70% of the smallest sample size for any of the groups; all other individuals are assigned to the testing sample. After the testing phase, several statistics were computed including the model’s accuracy, kappa, precision, recall and F1-score, etc. The discriminant analyses, i.e., both LDA and CVA, were iterated 1001 times, i.e., from the random creation of the training and testing subsets to the computation of the statistics. The iteration which yielded the best result was then used to predict the most likely function that could explain the use wear pattern identified on the Abri du Maras specimen. All analysis were performed in R-CRAN v4.4.1113. The code used and the confocal acquisition are available in open access on Zenodo (https://doi.org/10.5281/zenodo.17506415).