1Our knowledge of the Middle Palaeolithic is very disparate throughout France. This can be explained by the history of research institutions as well as by the heterogeneous distribution of caves, rock shelters and Pleistocene sedimentary formations likely to have conserved occupations from this period. In regions where researchers have been less attentive, more emphasis is often placed on the absence of sites with ample stratigraphies considered as reference sites rather than on the absence of Pleistocene occupations. The site of Roc-en-Pail, in Chalonnes-sur-Loire, discovered in the nineteenth century and excavated in the 1940s and 1950s by Dr Gruet, is an example of this. Despite the fact that this site contains the longest Palaeolithic stratigraphy in the Lower Loire and yielded the only fossil human remains in a Middle Palaeolithic context for the region (Gruet 1950), it is still very poorly known and cannot be compared with sites in south-western France generally considered to be reference sites. Our objective here is to critically review the bibliographical sources, as well as the excavation records relating to Dr Gruet’s research, in order to reconstruct the history of research on one hand and to establish a starting point for new field research initiated in 2014 on the other hand.
2The fortuitous discovery of fossil fauna in association with the caverns, caves and fissures of the Devonian limestone terrain in the Chalonnes region began in 1841. A critical reading of nineteenth century publications suggests that the discoveries of fauna and lithic industry at Roc-en-Pail were made in two, or even three distinct contexts. The first discoveries of fossil fauna occurred as early as 1863, and over the ensuing decades in the area of a quarry in a small hill, from which limestone was extracted to supply lime kilns. Currently, all that remains of the original landform, which must have been about 20 m above the Holocene filling of the alluvial plain, is the southern schistose slope. Considerable accumulations of bones were exhumed from the sandy-clay fillings of the Devonian limestone karst cavities, which were intersected during limestone quarrying. It is reported that more than 40 cartloads of bones were transported to the spoil from one cavity alone. Only L. Davy collected very rare Mousterian knapped flints in this context. This is therefore a palaeontological site with evidence of human occupation. In light of the composition of this fauna, reported in the literature at the time, it appears to date from the Middle Pleistocene, and is clearly different from the faunal assemblages of the Roc-en-Pail site itself. Other discoveries seem to be linked to the construction of a quarry road, which uncovered a cavity containing scattered flint tools. The opening of the cavity was blocked by a thick layer of sand and clay covering a large accumulation of bones (Biaille 1904). The last context is that of the Roc-en-Pail site, the subject of this study. After the winter of 1870-1871 at the latest, Dr Farge and C. Port discovered a flint flake in stratigraphy in a trench excavated the previous winter on the edge of the quarry (Farge 1872). They enlarged the trench and collected about twenty knapped flints, attributed by Farge to the “Moustiers” type, as well as numerous bones. During a second exploration, where E. d’Acy accompanied Farge, other flints were extracted, some of which we found in the d’Acy collection at the Musée d’Archéologie Nationale (Saint-Germain-en-Laye). The characteristics (typology, technology, surface condition) of these flints are comparable to those of the most recent Mousterian industry in Dr Gruet’s stratigraphy, which he designates as a Denticulate Mousterian.
3The exploitation of the quarry had already ceased by the very beginning of the twentieth century, shortly after the last publications, and the site of Roc-en-Pail was soon forgotten, to such an extent that research conducted by O. Desmazières and G. Denizot in August 1929 was unsuccessful. Subsequent surveys by Dr Gruet between 1934 and 1938, and again in 1942, located remains on the surface of a plot of land under cultivation. He dug out a first test pit in 1943 which intersected the top of the archaeological deposits. He then excavated from 1944 to 1956, with some interruptions in the 1950s. The bedrock was reached at several points in his excavation at a depth of almost 5 m. The excavation was backfilled in 1956, but Dr Gruet resumed excavations in 1969 with the aim of presenting the site during an INQUA congress field trip to the region. The long 1949-1956 trench was cleared and the eastern section was regularised and drawn. On this occasion, before the site was backfilled again in November 1970, samples for palynological analysis by L. Visset (1980) were taken in several stages. At the end of his research, Dr Gruet described a thick stratigraphy, which he considered to be the filling of a rock shelter destroyed by quarry workers’ pickaxes. This stratigraphy comprised seven complexes called ReP1 to ReP7, from the top to the base, which he considered as archaeological layers (Gruet 1984). With the exception of layer ReP1, which yielded lithic industry from the Upper Palaeolithic, related to the “Perigordian” with Noailles burins, and to the Aurignacian, all of them contain Middle Palaeolithic industries associated with fauna (table 3). In particular, we observe the succession of three Ferrassian Mousterian layers, a Quina Mousterian layer, and finally a denticulate Mousterian layer, like in the Périgord. Dr Gruet’s unpublished typological counts, transmitted by the latter, were then used for several studies in this domain, in particular those of F. Bordes (1953). In 1978, the Roc-en-Pail site was classified as a Historic Monument. Subsequently, a few occasional studies were carried out (e.g., David and Fosse 1999), but with no discussion of the stratigraphy established by Dr Gruet.
Tabl. 3 – Summary of information on lithic industries and fauna in Gruet (1984, fig. 2). The names of the archaeological units used in the Gruet collection have been added in the first column.
4Publications relating to Roc-en-Pail are too partial or too summarized to draw up a complete state of knowledge. Therefore, we compared the various sources, publications, excavation archives (section drawings, photos) and other accessible documents, to draw up a complete inventory of the collection from Dr Gruet’s excavation. This collection is stored at the Museum of Natural Sciences in Angers, and consists of more than 50,000 objects. The analysis of the documents and the collection sheds light on the temporality and the construction process of the archaeostratigraphy finally published by Dr Gruet (1984), as well as on how the excavation progressed, and particularly on its method. The archaeostratigraphic record appears to have been in its infancy during work in the 1940s at Roc-en-Pail. Initially, and potentially until 1947, all the remains discovered in the upper part of the sequence were grouped together in a single archaeological unit known as the "Upper Mousterian", which he compared to the Advanced Mousterian of La Quina. In the 1940s-1950s field notebook, differences in calligraphy (type of pencil) observed on the schematic records of sections between the geological descriptors noted in blue ink (the ink used for drawing the sections), and the main designations of the archaeostratigraphic units, noted in heavy pencil on a section probably dating from 1956, suggest that the archaeostratigraphy was partly constructed, or at least recorded, a posteriori. These elements indicate, that at best, the remains were stored with a provisional identification of their stratigraphic provenance, and at worst, that no distinction was made between the different levels of provenance in the early years of excavation, even though, as early as 1947, Gruet reached what he identified as the bedrock in place at the northern end of the excavation. This appears to be backed up by the examination of the Gruet collection. Dr Gruet did not record the position of the remains discovered during the excavation and henceforth their position is based at most on the combination of the stratigraphic attribution to one of the seven units ReP1 to ReP7 and the mention of the year of excavation which refers to an area of several square metres. In reality, the combination of both sets of information is very rare. For a significant proportion of the remains (22.6%) the stratigraphic provenance is unknown, reworked, surface-related or ambiguous. On the basis of typological and technological criteria, we have shown that the ReP3 assemblage (Mousterian Quina) consists of a mixture of a predominant Quina component, and a Levallois component. The proportion of the latter varies greatly in the current storage containers, suggesting that they may correspond to sectors of the excavation with varying degrees of mixing of the two archaeological layers (during excavation or subsequently). There are therefore significant biases in the Gruet collection. These have been confirmed and clarified by recent excavations, and have enabled us to discern which data could be inferred from this ancient collection.
5New excavations of the Roc-en-Pail site, under the direction of S. Soriano in 2014 and from 2016 to 2018, have located at least part of Dr Gruet’s excavation area (fig. 9), and the eastern edge of the 1969 trench was reopened over a length of about twelve metres. However, due to land-related and technical constraints, only the upper part of the deposits (about 2 m) could be documented. Some preliminary results allow us to review Dr Gruet’s interpretations and stratigraphy. Geomorphological observations first of all made it possible to rule out the idea that the Roc-en-Pail deposits were the filling of a rock shelter. The occupations were in fact conserved in deposits accumulated at the foot of the limestone hill. These deposits were firstly fluvial formations, a remnant of a low terrace of the Loire-Layon system, then thick slope formations derived from weathered limestone and shale and residues of Cenomanian sands. The layers of this formation were not organised in a continuous and parallel manner, but they resulted rather from the succession of several cycles of deposition and erosion, probably due to paleoclimatic and geomorphological factors. Samples for geoarchaeological analysis and OSL dating are currently under study, and will enable us to clarify the chronology and accumulation process of these deposits. Certain characteristics of the sediments (texture, structure) and the organisation of the layers clearly indicate that periglacial dynamics affected the organisation of the remains and their state of preservation to varying degrees. The archaeostratigraphy therefore appears more complex than Dr Gruet had envisaged, but the presence of a Quina Mousterian and a Denticulate Mousterian was confirmed. The excavation showed that fauna and microfauna were conserved in all the archaeological layers. Our initial observations show that Roc-en-Pail could potentially become a reference sequence for the Middle Palaeolithic of the Pays de la Loire, and even for central-western France. For this to happen, the specific characteristics of the site must be taken into account, namely its complex stratigraphy linked to a very dynamic sedimentary context, through a global taphonomic approach to the deposits.
Fig. 9 – Position of the site on the cadastral map and indication of the areas where surface remains were discovered by Dr Gruet (CAD S. Soriano).
