The systematic excavation, directed by A. Leroi-Gourhan between 1948 and 1963 at Arcy-sur-Cure, brought to light an important middle to upper Palaeolithic sequence. Ongoing research directed by F. David and M. Hardy since 1998 considerably enriches our knowledge of the cultural evolution at Arcy-sur-Cure. It is now possible to propose a global sequence from several caves and shelters located along the Cure River (Fig. 1, Fig. 2).

A first complex, interpreted as the Charentian, is present in layers IVb7 to IVb3 of the Grotte de l’Hyène (Girard, 1978), layers J and I of the Grotte du Bison (Girard, 1982; Lhomme, in Hardy et al., 2014), layers XIV and XIII of the Grotte du Renne (Girard, 1980) and probably all the layers of the Galerie Schoepflin (Hardy et al., 2013 and Leroi-Gourhan, 1956) and the Rotonde of the Galerie Schoepflin or RGS (Hardy et al., 2013 and Rocca, 2006). A second complex, the Denticulate Mousterian, was recognized in layers IVb2 and IVb1 of the Grotte de l’Hyène (Girard, 1978, Thiébaut, 2005 and Thiébaut, 2007), in layers H and G of the Grotte du Bison (Girard, 1982, Lhomme et al., 2005 and Thiébaut, 2005), and in layers XI and XII of the Grotte du Renne (Girard, 1980). Over the past ten years, the archaeological complexes from Renne and Bison have been extensively dated (Bailey and Hublin, 2006, Higham et al., 2010, Hublin et al., 2012 and Lhomme et al., 2005). Dates from layer XI of Renne by the Max Planck Institute both place this level before 44,000 BP according to the Bayesian model and before 40,900 ± 750 (EVA-85, Hublin et al., 2012) with standard measurements (Fig. 2).

The third complex, the Levallois Mousterian, is represented in the Grotte du Bison, layer E/F (Lhomme et al., 2005). The date for layer E of Bison is more recent at 38,400 ± 1,600 BP (OxA1001(Ly-1294)) and layer F could be slightly older with an age of 40,200 ± 1,500 BP (GrA20477(Ly-1915) (Lhomme et al., 2005). On this basis, the last two Mousterian levels of Bison (layers E and F) would be more recent than layer XI of Renne (Fig. 2).

Then, Châtelperronian occupations are present in layers X, IX and VIII of the Grotte du Renne and in level D of the Grotte du Bison (Connet, 2002, Connet, 2007 and David et al., 2001). Finally, the Aurignacian is represented in level VII (Schmider, 2002), before a Gravettian industry in levels VI and V (Klaric, 2003)

Layer XI of Grotte du Renne was excavated between 1956 and 1960 by A. Leroi-Gourhan (Leroi-Gourhan, 1962 and Leroi-Gourhan Arl. and Leroi-Gourhan A, 1964). It is situated between a Discoid–Denticulate layer (XII), and the first Châtelperronian layer (Xc).

However, it seems, as confirmed by the nearby Grotte du Bison, that an erosional event occurred between the last Mousterian and the Châtelperronian occupations (David et al., 2005) (Fig. 2). On the basis of the available data for the Grotte du Renne, no detailed analysis or sedimentary dynamics can be undertaken. In contrast, the recent excavation of the Grotte du Bison provides interesting data (David et al., 2005), particularly due to the proximity of the two entrances and the identification of parallel stratigraphic sequences, which may be continuous from one porch to another for layer VIII (David et al., 2001 and David et al., 2005). Thus, layers H to E at Bison (middle Palaeolithic) are of the same type as layers XII to XI at Renne. The H to D complex presents stratigraphic discontinuities due to erosive phases (David et al., 2001 and David et al., 2005) and it is possible that the same episodes may have affected Renne. Moreover, the preferential orientation of some types of remains at the base of layer X seems to be related to taphonomic processes such as solifluxion, and some authors suggest more extensive disturbances (Bertran et al., 2010). Finally, as stated above, the inter-layer refits in the Châtelperronian from Renne demonstrate vertical movements (Hublin et al., 2012), but also possible problems of identification of the archaeological complex during the excavation. The extent of these movements is impossible to identify due to the absence of altimetry levels for each object. Based on the available sections, these movements could sometimes exceed 20 cm. Refits were not tested for layer XI, but the sedimentary and archaeological continuity between the base of layer X and layer XI facilitated the migration of objects between layers mainly comprising stacks of slabs (David et al., 2001).

In layer XI, the excavated surface is almost 40 m², with an average density of 56 lithic remains per square meter. The base of the layer is irregular, following the pattern created by the collapsed limestone blocks exposed at the top of layer XII. Layer XI is about 20 cm thick over most of the excavated zone but can reach as much as 50 cm or as little as 10 cm in the median axis, along a channel (Fig. 3).

The lithic assemblage of layer XI was interpreted as a Denticulate Mousterian with rare Levallois production and was compared with layers IVb2 and IVb1 of the Grotte de l’Hyène (Girard, 1980). At the Grotte du Renne, the Levallois Mousterian occupation represented by layer E/F at Bison is not present.

The layer XI assemblage represents a total of 2258 lithic remains. The new study of this assemblage aimed to assess the integrity of the collection. A first observation detected certain heterogeneity within the collection. Some pieces (152) show very typical upper Palaeolithic (Châtelperronian and Aurignacian) features, such as volumetric blade and bladelet cores, backed points, burins, scarred pieces, truncated blades, blades shaped by soft percussion (Fig. 4). One third of these elements derive from the slope in front of the cave entrance. This part of the site presents numerous problems of layer identification. The examination of the stratigraphic and topographical data demonstrates that layer XI is not a flat surface and is in direct contact with the layers below and above it (Fig. 3, Fig. 5). It is therefore possible that some objects were attributed to the wrong layer during excavations. This hypothesis is supported by two other arguments. On the one hand, these pieces are mainly in flint, a much rarer raw material in layer XI. On the other hand, in typological or technological terms, they have absolutely nothing in common with the rest of the assemblage (Fig. 4). We therefore decided to take the precaution of excluding these pieces from this study. We also discarded the unlabeled pieces. Finally, after this whittling down, the lithic series studied here contains 2102 pieces, or 1862 pieces without the 240 natural chunks and blocks (Table 1).

The aim of this analysis is to assess the technical system, to reconstitute the reduction sequences, the production aims and the functional potential. We therefore used a classical technological approach (Boëda et al., 1990, Inizan et al., 1995 and Inizan et al., 1999).

We recorded the following data for the flakes: dimensions, morphology, knapping angle, order and direction of the removals on the dorsal surface. The objective of this study of the flakes is to identify their morpho-technical characteristics and their position in the production sequence.

For cores, we describe the final morphology of the flaked core, the knapping angle, the order and direction of removals. The objective of the core analysis is to identify the structure of production (Boëda, 2013) and the intended products.

Finally, for the retouched tools, we observed blank morphology and technological provenance, the transformation methods, as well as the consequences of retouch on the blank.

More than 80% of the raw materials are made up of coarse to fine-grained chert (Table 2). This Jurassic chert is available directly on the area near the site as blocks and slabs. Various colours and quality of flint are also present (16%), and one third of the pieces in this latter material are retouched tools and retouch flakes. The flint is from the Cretaceous formations south of the Paris Basin, and is available in secondary position on slopes and alluviums about 20–30 km from the site (Connet, 2002 and Schmider, 2002). The remaining raw materials are very rare (1%), and include quartz, quartzite, metaquartzite and chalcedony. They derive from the Morvan formation and are probably present in the alluviums of the Cure River. Raw material distribution from layer XI is comparable to that of the other Denticulate Mousterian assemblages of Arcy-sur-Cure (Girard, 1980, Girard, 1982, Lhomme et al., 2005 and Thiébaut, 2005).

The assemblage thus includes a total of 1862 items, excluding the probable intrusive pieces (Table 3). The production results from two main débitage concepts: discoid and additional flaking.

The discoid reduction sequence is the most represented concept in the collection. This concept is identified by the morpho-technical features of the products and the core characteristics. The direction of the removals on the dorsal surface of the flakes, as well as their knapping angles, are consistent with a discoid concept (sensu Boëda, 1991 and Boëda, 1993). Typical flake products are made up of four groups: débordant flakes, non-débordant flakes with a width superior to length, non-débordant elongated flakes and pseudo-Levallois points (Table 4, Fig. 6). The elongated flakes have wide butts and open knapping angles and are produced using hard hammer percussion set back from the overhang. The elongation ratio does not exceed 2 and is generally less than 1.5. These flakes cannot come from a Châtelperronian type production, which is characterized by tangential percussion with a soft hammer, for the production of objects with flat sections using unipolar, or even recurrent bipolar management. Débordant flakes are the most abundant blank type, whereas pseudo-Levallois points are poorly represented. The cores assigned to the discoid reduction sequence are generally exploited on one surface, frequently due to the heterogeneous quality of the raw material (Fig. 7). Therefore, even if most of them are not morphologically discoidal, the order and direction of removals, as well as the débitage angle are part of this concept and are features of discoid production.

The second débitage concept is included in the variability of the additional concept (Boëda, 2013). This means that production only concerns part of the initial block, unlike for discoid or Levallois production, where the entire block could potentially be exploited. In this type of debitage system, production generally uses the natural lateral and distal convexities on the block to obtain the intended flakes. An initialization phase can create these convexities, but they are limited to a small part of the block. The production sequences are short and if there are several sequences on the same block, there is no link between them.

Blanks show the same morpho-technical characteristics as for discoid production, but due to the direction of removals and the knapping angle, they cannot be attributed to the discoid concept. The characteristics of the raw material, generally heterogeneous chert, probably incited the knappers to use different production concepts on certain blocks, which nonetheless resulted in the production of similar blanks.

The implemented débitage resulted in wide variability in blank morphology, cutting-edges and potential prehensile parts. Most of the products are not retouched and probably represent the primary functional aim of the toolkit.

Retouched tools are mainly made up of notches and denticulates (Table 5, Fig. 6). The débordant flake is the main blank used for these tools, probably on account of a suitable prehensile part. The denticulates and notches are of different dimensions, ranging in size from micro-denticulates to large notches, generally opposite backs. Notches can be direct or inverse, and include retouched and Clactonian notches.

We also note the presence of some marginally retouched flakes and a few scrapers. These tools concern mostly non-débordant small to medium-sized flakes. Retouch is direct, of varied scope and morphology, generally with a low angle.

The results of the new analysis of layer XI highlight the middle Palaeolithic aspects of the lithic assemblage. Indeed, the production concept and tools are typically Mousterian. The dominance of discoid production, the absence of blade or bladelet production and of upper Palaeolithic tools, the importance of denticulates and notches among the retouched tools, do not represent the emergence of major technological changes. The presence of some elongated pieces, as well as backed débordant flakes, is consistent with the discoid concept and cannot be removed from their technological context to argue in favour of the advent of modern behaviour.

Layer XI of the Grotte du Renne is part of the broader middle Palaeolithic sequence of Arcy, attributed to the last part of the Mousterian, which can be divided into three techno-cultural complexes.

The study of the last Mousterian industry reveals a technical system rooted in the middle Palaeolithic. No elements indicate characteristics portending the Châtelperronian. The two complexes, Levallois/discoid/Mousterian and Châtelperronian, are easy to distinguish in terms of blank production modes and functional aims. In the Châtelperronian, flint represents 57% of the raw material, and fine-grained chert is preferred to the coarse and medium-grained varieties. The blade technical system is totally absent from the Mousterian sequence, while it is the main technical system during the Châtelperronian, like in other upper Palaeolithic assemblages (David et al., 2005 and Schmider, 2002). Châtelperronian flake producing cores are not very frequent (22% of the cores, 151 objects out of 674) and comprise 106 cores centripetal cores, 39 polyhedral/formless cores and 6 unipolar cores (Fig. 8). The centripetal debitage of plane to slightly convex surfaces is the best represented with 64 cores. The removal angles are about 60/70° for a production of only 2 to 3 convergent flakes with an elongated morphology. In 40% of the cases, flakes are used, and in particular flint flakes. For the Châtelperronian of Arcy, these cores can represent an opportunistic production of elongated flakes with a flat section, in addition to the production of volumetric laminar cores. 34 discoid cores are also present (5% of the cores), as well as several tools including notches, denticulates, side scrapers; but also several end scrapers or borers. In layer X, they are mainly on blocks, they are more frequently on flakes in layers IX and VIII where their proportion decreases considerably (6.5% in X, 2.2% in IX, 3% in VIII). Thus, although it is probable that some of these cores come from the Mousterian layers of Renne, a Châtelperronian provenance cannot be totally ruled out. The same observation can be made for the 39-polyhedral Châtelperronian cores; particularly those from layer X which present similar characteristics to the cores related to the additional debitage concept. The production of short flakes is known in Châtelperronian complexes and is present for example in open-air sites such as Vieux Coutet (Grigoletto et al., 2008), Bastet and Bidart (Bachellerie, 2011), where the sedimentary sequences do not present the same problems as in caves.

The comparison of the distribution of retouched tools throughout the Arcy sequence shows some clear tendencies (Table 6). All these elements indicate distinct technical traditions as regards tool conception, and probably for use and prehension. The Châtelperronian of Arcy is clearly different from the middle Palaeolithic as regards lithic production (blade technical system), but also on account of the presence of a bone industry and ornaments, which are absent from the Mousterian layers. These elements link the Châtelperronian to the upper Palaeolithic world (Bordes and Teyssandier, 2011). From a technical point of view, we observe that at the onset of the Châtelperronian at Arcy, the last Mousterian industry represents a technical system with no future.

While evidence of the final Mousterian is quite rare in northern France, the presence of the Discoid–Denticulate or/and Levallois Mousterian at the end of the middle Palaeolithic has also been documented at other sites in North-Central and southwestern France (Bodu et al., 2014, Delagnes and Rendu, 2011, Guibert et al., 2008, Jaubert et al., 2011, Koehler, 2011, Locht, 2003, Soriano et al., 2014, Thiébaut et al., 2009, Thiébaut et al., 2012 and Vieillevigne et al., 2008). In these sites, backed pieces, elongated flakes and bifaces are absent. Conversely, denticulates, notches and scrapers represent the main retouched tools, probably completed in functional terms by the unretouched blanks provided by Levallois and/or discoid blanks.

Although certain authors consider that the Châtelperronian derives from the MTA (Pelegrin and Soressi, 2007, Ruebens et al., 2015, Soressi, 2002 and Soressi, 2005), it seems to precede other techno-complexes, and in particular the Discoid–Denticulate (Aubry et al., 2011, Jaubert et al., 2011 and Vieillevigne et al., 2008). In addition, at Arcy-sur-Cure, as in other contexts, the Discoid–Denticulate does not seem to be the last expression of the Mousterian. Indeed, another industry, characterized by the Levallois concept, scrapers and points, similar to the “Levallois Mousterian with large scrapers” according to Jaubert et al. (2011), indicates Mousterian occupations (Lhomme et al., 2005). This complex may correspond to a very brief period, a final Mousterian pulsation. This contrasts with the Levallois–Charentian (Renne XIV and XIII, Bison J and I, Hyène IVb7 to IVb3, Schoepflin, RGS) and the Discoid–Denticulate (Renne XII and XI, Bison H and G, Hyène IVb2 and IVb1), as suggested by their significant chronological sequences.

Thus, from a strictly technical point of view, the lithic industry from the end of the Mousterian in the Arcy region, as already claimed by other authors (Thiébaut et al., 2009), cannot be considered to be the origin of the Châtelperronian.