1The northern Aquitaine Basin—and more broadly the southwest of France—comprises unparalleled evidence for our understanding of the Palaeolithic, and in particular the Middle Palaeolithic (e.g., Bordes and Bourgon, 1951; Geneste, 1985; Jaubert, 2011). Furthermore, numerous ongoing field operations and research projects in this geographic zone continue to produce significant data. Despite considerable variability, recent studies tend to highlight strong regional unity, characterized by the succession or coexistence of “cultural entities,” each defined by specific technical traditions and spatial usage practices—at least in part—adapted to the environmental variations of the Last Glacial (Vieillevigne et al., 2008; Faivre et al., 2014; Ortega et al., 2022). However, these studies focus primarily on the Périgord. Consequently, the chrono‑cultural structuring of the northern Aquitaine Basin and adjacent areas (Pyrenean Piedmont, southern Massif Central), their links with neighbouring areas, as well as the potential territorial boundaries of the different Neanderthal societies, are still difficult to discern. This article addresses this issue through the study of the Rescoundudou site, which, both in terms of its chronology (MIS 5a or 5c) and its isolated geographical position in the Aveyron Causses, stands out from the corpus of Middle Palaeolithic sites in northern Aquitaine.
2The main excavation area was stripped in the early years of fieldwork, revealing a rich lens of lithic and bone remains at least one metre thick (Jaubert, 1983). This level appears to correspond to a series of occupations in a propitious geomorphological setting with water accumulation—attractive to both humans and animals. Available seasonality data (Roussel, 2023) indicate occupation episodes centred on the warm seasons (spring, summer, early autumn).
3The flint material s.l. shows the coexistence of two conceptually different flaking systems, with the presence of dominant Levallois flaking associated with ‘laminar-lamellar’ flaking (fig. 1). Both operational schemes display similar dynamics, based on the production of recurrent series of blanks, as well as a two-stage process with the secondary selection of flakes as production matrices exploited according to similar principles to those observed for block reduction.
Fig. 1 – Rescoundudou, branching operational chain: reconstruction of the first operational stages.
4Levallois flaking is the dominant production system and is part of a long operational sequence initiated by the partial or total shaping of blocks by removing centripetal cortical flakes. This first stage served to establish the technical and volumetric criteria necessary for this type of flaking by hierarchising the core surfaces (striking platform and debitage), and to initiate the production criteria by creating guiding ridges for elongated flakes.
5The initial production sequence is uni/bipolar recurrent, aiming to obtain flakes that are often longer than they are wide and allowing for self-maintenance of the distal convexities. This production phase may be interspersed with short sequences maintaining the technical criteria by detaching predetermining products (core edge flakes or éclats débordants). The striking platforms are carefully prepared, as evidenced by the number of facetted butts. The operational dynamics continue in a recurring pattern, with successive shifts of the exploited poles on the core periphery, which then presents centripetal removal scars. As core volume decreases, the range of products is more varied and includes flakes bearing unipolar, convergent unipolar, cross-bipolar and centripetal negative scars. The debitage sequences continue intensively, as shown by the high degree of depletion and the reduced dimensions of the few cores in the series.
6A second operational stage involves the selection and exploitation of flake-cores or matrices following similar production dynamics to Levallois block flaking (fig. 2). Two flaking modes are used, either exclusively or combined, depending on the face worked. The lower face method (Kombewa type) is the most common, but the upper face method (Kostienki type) is also well represented. Both methods are managed in a similar way by initiating removals that take advantage of pre-existing convexities on the flake-matrices (guiding ridges, lower face of the flake) accompanied by careful and localised preparation of the striking platforms by direct and/or inverse truncation. Production is carried out by a series of recurrent removals from one or two poles, but the sequences are significantly shorter than for Levallois production due to the small size of the initial flake-core volumes. Most examples follow independent flake-core management (mainly for the ‘Kombewa’ type), but some show matrix management with independent successive phases (recycling tools as production matrices). Again, these elements show branching Levallois production chains, frequently described in Mousterian contexts (Bourguignon et al., 2004; Faivre, 2008; Mathias and Bourguignon, 2020).
Fig. 2 – Rescoundudou, branching operational chain: reconstruction of the second operational stages.
7Alongside Levallois flaking, we have identified the use of an autonomous ‘laminar-lamellar’ reduction scheme at Rescoundudou. It can be distinguished from Levallois flaking on the basis of the morpho-technical criteria of products, the discrete presence of specific technical products and its operational dynamics. Initial volume management relies mainly on favourable preexisting volumetric criteria (regular convexities or cortical dihedrals), requiring little adjustment (opening of a striking platform, partial modification of convexities) before the production phase.
8Production is carried out from a single pole in recurrent series following a semi-rotating volumetric scheme, with rare evidence of lateral and distal management of convexities during reduction. These criteria lead to the regular removal of laminar products with cortical sides or residual distal cortex, which are both predetermining (maintenance of convexities, creation of guiding ridges) and predetermined. Products without cortex show certain similarities with elongated unipolar Levallois products, but also present ‘laminar-lamellar’ characteristics: exterior flaking angles (angles de chasse) less than 80°, triangular or trapezoidal sections, predominantly curved profiles, smooth or linear butts, poorly defined or even absent bulbs, and frequent abrasion of striking platforms. These characteristics are solid indications of the use of tangential percussion with a soft (mineral) hammer associated with this operational scheme. Elements linking “laminar” and Levallois operational schemes have also been observed, and indicate the recycling of elongated removal production matrices into Levallois cores via a volume reconfiguration phase. These technical indicators echo certain lithic assemblages where the dichotomy between the ‘Levallois method’ and ‘volumetric’ laminar flaking is not easy to identify (Boëda et al., 1990; Révillion, 1993, 1995). In a way, the real autonomy of these two operational schemes is questionable: the absence of laminar cores and the exclusive use of local materials (without distinction from Levallois production) raise questions about the systematic recycling of laminar cores into Levallois cores. They may represent two interwoven operational schemes, which can only be distinguished at an early stage of the operational chain, depending on the possibilities inherent in the technical criteria of the available volumes.
9The coexistence of Levallois and laminar schemes is well documented in northern France during the Late Middle Palaeolithic (Révillion and Tuffreau dir., 1994) and is one of the criteria used to define the ‘North-Western Technocomplex’ (Depaepe, 2007; Locht and Chaussé, 2021). In north-western contexts, laminar knapping can involve significant technical investment (care taken in the shaping and maintenance of the block during knapping), but this is nevertheless largely limited. The technical processes involved in laminar production are known and mastered (ridges, tablets, latero-distal overshooting), but are only used occasionally (Locht and Chaussé, 2021). Thus, Levallois and laminar productions are generally easily distinguishable (Révillion, 1995) and are usually autonomous. In the south of France, the coexistence of Levallois and blade productions in the Late Middle Palaeolithic is also documented, but is much more discrete (Soressi, 2002; Bourguignon et al., 2004; Faivre, 2008; Blaser et al., 2012; Carmignani and Soressi, 2023).
10Lamellar or bladelet production, which has been increasingly documented in recent years in Middle Palaeolithic contexts (Slimak, 1999; Fernández et al., 2004; Faivre, 2008), can be observed on two scales. On the one hand, it occurs in the production of ‘laminar’ removals, where it plays a predetermining and predetermined role by helping to maintain the proximal convexities of the knapping surfaces. On the other hand, it is expressed through autonomous debitage on small-sized matrices (flakes or fragments of frost-fractured blocks). This autonomous debitage is characterised by initialisation and volume management phases involving the creation of lateral ridges designed to maintain distal convexities, and by a knapping process that involves both the narrow face (thickness) and the broad face of the matrix. As with Levallois production, this lamellar production shows a branching of the laminar-lamellar operational chain. Whether in terms of knapping methods or the range of products obtained, this organisation of the operational chain thus presents many similarities with the series from the upper part of the Combe-Grenal sequence (Faivre, 2008).
11Retouched tool manufacture is somewhat disparate in terms of the products issued from the different operational schemes and stages described above. Products from the second operational stage, whether they derive from Levallois or ‘laminar-lamellar’ schemes, are practically always left without retouch. Only one Kombewa-type flake is retouched. With regard to the first operational stages, once again a distinction can be made: retouched blanks are preferentially selected from predetermined and/or predetermining Levallois flakes, as well as from undifferentiated flakes, whether cortical or not. Laminar products are less selected for retouching and are mostly left unworked, like core edge flakes (éclats débordants).
12Tools, in the broadest sense of the term, comprise non-retouched and retouched products. With regard to potential unretouched tools, the techno-functional characteristics of the blanks reveal variations related to the operational scheme and stage. However, four main classes of products with similar, additional and/or complementary technical and functional characteristics can be distinguished (fig. 3), mainly based on the presence or absence of a steep edge opposite a cutting edge. Most products issued from ‘laminar-lamellar’ and Levallois production (uni-bipolar modes) have a structure composed of two regular subparallel cutting edges, occasionally converging at the distal end (especially for laminar-lamellar elements), depending on the elongation axis of the products. In Levallois production, the sometimes marked presence of a facetted butt creates a steep area perpendicular to the axis of the cutting edges. In both production schemes, a second type of structure can also be observed, consisting of one (laminar-lamellar pattern) or two (Levallois pattern) linear cutting edges opposite a steep part. The latter may correspond to a natural edge (cortically backed flake), a raw edge (core edge flake s.l.) or a backed butt (limited backed flakes). These different structures mainly seem to illustrate the search for an ‘edge’ type transformative unit, and to a lesser extent, a ‘point’ type, with cutting angles of less than 30°. As for potential associated uses (Claud, 2008; Costamagno et al., 2019), the available data point almost exclusively to longitudinal, or more rarely transverse, cutting actions on soft materials such as skin or fresh meat.
Fig. 3 – Summarized diagrams of the techno-functional characteristics of the potential toolkit (non-retouched and retouched tools) from Rescoundudou.
13Apart from a few denticulate-type pieces, retouched tools are largely dominated by tools from the scraper typological group (fig. 3), particularly simple lateral scrapers. Cutting edge retouch methods appear to be stable and result in the creation of transformative contacts with homogeneous morpho-functional characteristics. Simple lateral scrapers, whether the shaped edge is opposite a raw cutting edge or an asymmetrical edge (back or butt/back), have a regular outline (convex or straight) and a dihedral cutting edge angle of between 37° and 44°. These data point to a modification of the cutting edge with a view to creating transformative rather than prehensile contacts. For double or convergent scrapers, the delineation and/or angle of the cutting edge between the two modified edges is different: a transformative contact with identical characteristics to those of simple lateral scrapers is opposite a retouched cutting edge with a higher edge angle. These characteristics may reflect a diachronic difference in the use of the edges or distinct purposes for the two cutting edges. However, regardless of the type of tool, there is generally little retouch, often confined to the cutting edge, indicating limited modification of the initial blanks.
14The procurement territory appears to be mainly centred on the Causse Comtal (> 85% of the lithic material) for the collection of flint and quartz (fig. 4). This geotope also offers access to varied environments, conducive to the exploitation of different hunting grounds: plateaus, favourable to family groups of ungulates, and valleys or low forest levels, more exploited for the capture of isolated adults (Roussel, 2023). The acquisition of more distant materials mainly concerns the west and the Lot valley (a limited logistic area). This indicates close links between the site’s territory and the circulation of production matrices and raw and/or retouched products. A few isolated retouched products and/or retouching flakes provide evidence of links with western areas (flint artefacts from the Haut-Agenais) and northern areas (Aurillac Basin). Activities contemporary with the Rescoundudou occupations therefore seem to have taken place mainly in the interfluve between the Lot and Aveyron rivers, without extending beyond the Terrefort Rouergat to the west and the Causse Comtal to the east. The apparent absence of links with the south or east, the Grands Causses and, beyond the Tarn and its tributaries, the Larzac (southern Aveyron, Lozère), is noteworthy.
Fig. 4 – Potential territorial organisation of the Rescoundudou group(s).
The dotted arcs represent the apparent boundaries of the different areas, used directly (during the seasonal nomadic circuit and/or for logistic expeditions) or indirectly (links with other territories).
15These various elements seem to demonstrate that Rescoundudou was a residential camp (rather than a residential hunting camp) and not a logistic site (Fougère, 2011): varied economic activities, complete or almost complete operational chains for materials used in everyday life and limited logistics, as well as a significant site area. Indeed, it should be noted that the estimated surface area of the Middle Palaeolithic levels is around 200 to 250 m2. Similarly, faunal data (Roussel, 2023) seem to rule out the hypothesis of a short-term logistic camp: they indicate activities such as the exploitation of family groups of large ungulates and the prolonged acquisition of the same type of game over several months, characteristics that are not compatible with occasional logistic use. In addition, the presence of human remains, including those of young Neanderthals (Jaubert and Maureille, 2008), points to multiple age groups, a finding generally associated with residential groups. Current data cannot conclusively distinguish between two main site use hypotheses: occupations lasting several months or shorter stays of a few weeks, or frequent versus occasional visits. However, in terms of larger scale organisation, Rescoundudou could be positioned within a cycle of seasonal nomadism, a priori between the Périgord and Cantal regions, probably linked to the Lot valley. The hypothesis of a group territory centred around the Marcillac valley and the Decazeville basin, with links to other more distant territories, is also conceivable, despite some reservations. The occupation of the Aveyron area could be linked to the exploitation of the left bank of the Lot and the Causse Comtal, either occasionally or on a more permanent basis, probably related to the specific seasonal characteristics of these different biotopes.
16The northern Aquitaine basin has yielded unparalleled data for our knowledge of the Palaeolithic, and more particularly of the Middle Palaeolithic in Europe (fig. 5), and numerous ongoing field operations and multiple research projects (e.g., Faivre et al., 2017) continue to add to this stock of knowledge. From a strictly chronological point of view, few sites are a priori contemporary with Rescoundudou (MIS 5a or 5c) and, more broadly, with isotopic stage 5 (Jaubert, 2011). This gap in the record is often attributed to the differential preservation of records dating from this period. If we broaden our chronological window to include the beginning of isotopic stage 4, a certain technological consistency is observed in regional archaeosequences (relative to the diversification observed during MIS 4 and MIS 3; Jaubert, 2011; Faivre et al., 2017). The series are mainly dominated by Levallois flaking, mostly with recurrent methods (centripetal, unipolar or bipolar). This constitutes the cultural ‘background noise’ of this period in the Aquitaine Basin and more broadly in Western Europe (Locht et al., 2016; Faivre et al., 2017; Vaissié, 2021). Other operational chains are observed, but are mainly marginal (discoid, bifacial shaping, laminar). In this context, we can identify around ten sites (fig. 5) with an association at least partially similar to Rescoundudou, with the dual presence of Levallois and non-Levallois laminar knapping.
Fig. 5 – Location of the main Late Middle Palaeolithic sites (125 – 40 ka) in southwest France and comparative contexts mentioned in the text.
The colours indicate the main production methods used: grey = unipolar ‘semi-rotating’ flaking + discoid + bifacial; blue = Levallois + Quina + laminar; red = Levallois + Laminar + Discoid + Bifacial shaping; yellow = Levallois + Laminar + Bifacial shaping; green = Levallois + Laminar.
17Outside the strict confines of the Aquitaine Basin, but close to Rescoundudou, the Canalettes rock shelter, located on the Larzac (southern Aveyron; Meignen dir., 1993; Lebègue, 2012), is a relevant sub-contemporary site for comparative purposes. However, the laminar-lamellar component is proportionally less developed at the latter site than in the Aquitaine contexts or at Rescoundudou, while the centripetal mode clearly dominates Levallois production. The Canalettes site is thus more similar to a ‘Mediterranean cultural area’ (Lebègue, 2012). This regional group, comprising Languedoc, the middle and lower Rhône valley and the Ligurian-Provençal arc, displays relatively homogeneous techno-cultural characteristics during the Late Middle Palaeolithic (Porraz, 2005; Lebègue, 2012; Lebègue and Wengler, 2014; Moncel et al., 2014). Lithic production is dominated by the Levallois concept (mainly recurrent centripetal methods, but also unipolar/bipolar), and the coexistence of other production systems (laminar, discoid, etc.). In these contexts, laminar production is generally marginal (represented by a few isolated blades). However, a few contexts present well-defined and apparently autonomous laminar components: Maras and Baume Flandin in the middle Rhône valley (Moncel, 2005), Canalettes in the southern Grands Causses (Lebègue, 2012), Baume-Bonne and Bau de l’Aubesier for the Verdon and Vaucluse (Gagnepain et al., 2001; Carmignani and Soressi, 2023). These laminar productions are dated between the end of MIS 6 and the first half of MIS 4 (or even from MIS 7 to MIS 5 for Bau de l’Aubesier). Laminar flaking is unipolar with little investment in block volume, with production remaining close to the surface. Among these contexts, however, the Bau de l’Aubesier series stands out for its autonomous production of blades and bladelets following different operational schemes: unipolar volumetric production and surface production following unipolar, bipolar and convergent methods. The presence of curved flakes, crests and neo-crests attests to maintenance and reworking phases (arching and convexities) during debitage, both for laminar and lamellar production. The fairly high proportion of retouched laminar blanks in the Bau de l’Aubesier series (between 12 and 25% depending on the layers; Carmignani and Soressi, 2023) also stands out, although this remains marginal and seems to mainly concern the regularisation of product edges.
18The presence of raw materials from the Cantal region at Rescoundudou also points to links with the Massif Central and other similar contexts in terms of technical traditions (Vaissié, 2021). One example is the Champ-Grand site (Loire), which differs from the contexts mentioned above in that discoid rather than Levallois flaking is predominant. However, like at Rescoundudou, it comprises autonomous laminar and lamellar components (Slimak, 1999). Another context showing strong similarities is Baume-Vallée in Velay (Haute-Loire; Hermens and Laborde, 1965; Raynal, 1975; Vaissié, 2021), where the sequence covers a chronological interval between the end of MIS 5 and the beginning of MIS 4. Similarities appear in particular in the exploitation of procurement areas, with use of the same procurement sites in the Goul valley as those observed at Rescoundudou. The series are marked by the predominance of Levallois flaking (uni/bipolar and centripetal modes), but also comprise a less pronounced laminar component, which seems to result primarily from conditions conducive to the production of short series of elongated removals. Nevertheless, this site comprises highly diversified materials from the main neighbouring regional groups (Aquitaine Basin, Mediterranean Languedoc and Saône and Rhône valleys; fig. 6). These examples of shared technical traditions, supported by established links with raw material circulation, raise the question of the existence of possible social networks structuring the Neanderthal populations of this period (Vaissié and Faivre, 2024).
Fig. 6 – Location of the main sites and associated lithic techno-complexes for the initial phase of the Late Middle Palaeolithic (MIS 5 / early MIS 4) in Western Europe.
19Finally, the last complex with comparative elements to Rescoundudou is northern France, where sites presenting a Levallois – laminar association are much more common and better documented. These sites, associated with the North-western Technocomplex (Locht et al., 2016; Locht and Chaussé, 2021), display relatively homogeneous techno-cultural characteristics during the Early Weichselian. This techno-complex is characterized by the coexistence of several operational sequences: flake knapping, mainly Levallois (more rarely Discoid or non-diagnostic), point knapping (following convergent unipolar patterns), and prismatic laminar knapping. Unlike the contexts observed in southern France, laminar flaking in northern sites generally shows significant proportions of complete, integrated operational sequences in the series. Production sequences are often long, combining various technical processes for core maintenance and the continuation of flaking (creation of crests, curved flakes, etc.), although their frequency varies. The choice of an adequate initial volume remains, however, a central criterion for initiating this type of flaking. The distinct geomorphological contexts (open-air sites) may be one of the factors accounting for such differences. Lastly, the almost exclusive use of local materials, due to the abundance of available resources, reflects rational economic behaviour adapted to environmental constraints.
20The re-evaluation of the Rescoundudou site and, in particular, the study of a representative lithic series, have provided new data on this isolated Aveyron context, both in terms of raw material circulation and techno-cultural characteristics. Based on current knowledge, the various occupations, attributed to the beginning of the Last Glacial Period (MIS 5c or 5a), are part of the Middle Palaeolithic cultural landscape in northern Aquitaine, including the beginning of MIS 4. The lithic assemblages are dominated by Levallois flaking, using recurrent mainly unipolar and bipolar techniques, associated with laminar-lamellar production schemes, some of which are autonomous. This technological association is documented in certain contexts in the Aquitaine Basin (Garris 2 and Combe-Grenal in particular), which are very similar to Rescoundudou in terms of technical traditions (both from a general point of view and in terms of the more subtle characteristics of the operational schemes). Rescoundudou is nevertheless original in terms of the percussion techniques used for laminar production, which seem to indicate the use of tangential percussion and/or a soft hammer (probably mineral). This does not apply to Levallois production on quartz.
21The raw material procurement territory, oriented towards the west and the Lot valley, supports these observations. This territory suggests a strong link between the Rouergat domain and the Mousterian macro-area of the Périgord. Conversely, the absence of materials from eastern or southern regions could define the boundaries of a possible cultural territory. These various data link the Rescoundudou site to a Middle Palaeolithic macro-territory in northern Aquitaine. It appears to form the easternmost limit of this territory and is a chronological milestone for the beginning of the late Middle Palaeolithic (MIS 5). However, from a geological and geographical point of view, the site is located in the south of the Massif Central and the hydrography of the Lot basin may have played a significant role in defining land use strategies.
