The Moulin Quignon site in the suburbs of Abbeville was discovered by Boucher de Perthes during the XIXth century. This sand and gravel quarry is located in the High Terrace on the right bank of the Somme River close to its confluence with the little Scardon River (Figure 2). The site yielded fossil bones of large mammals and flint “axes,” but it quickly became a subject of controversy with the exhumation in 1863 of an “ante-diluvian” human mandible, found in a stratigraphic context analogous to the aforementioned ‘axes’ (Abbeville, 1866; Boucher de Perthes, 1847; Quatrefages, 1863a; Quatrefages, 1863b; Boucher de Perthes, 1864a; Boucher de Perthes, 1864b; Milne, 1864). (see Supplementary Material for details).

The gradual disappearance of quarries and urbanization in the second half of the nineteenth century in Abbeville made it even more difficult to correlate Moulin Quignon with other sequences from the Carpentier and Léon sites and arbitrary comparisons were sometimes advanced (D'Ault du Mesnil, 1889; D'Ault du Mesnil, 1896; Commont, 1909). (Breuil, 1939; Breuil et al., 1939; Hurel et al., 2016a; Hurel et al., 2016b).

The collections of Abbeville Moulin Quignon were thus forgotten for almost 150 years. Research resumed in 2012 as part of a programme dedicated to the earliest settlements and to the Acheulean, in particular in the Somme valley (Hurel, 2014; Antoine et al., 2016; Moncel et al., 2016).

The geological context of the discoveries was re-examined through the testimonies left by Boucher de Perthes, then by the geologists and archaeologists who subsequently worked on the fossil alluvium from the Somme to Abbeville (Bahain et al., 2016). This critical approach was supplemented by new studies and in particular by the results of work carried out at Carrière Carpentier, as well as by a process of mapping alluvial formations in the Abbeville sector (Antoine et al., 2015; Antoine et al., 2016) (Figures 2A,B). The main results confirmed the fluvial nature of the archaeological layers of Moulin Quignon, which are embedded in sands and gravels of the high terrace of the Somme (Antoine et al., 2019). The terrace lies on an incision step in the chalk bedrock located about 40 m above the maximum incision of the Somme valley in Abbeville. The dating carried out at Moulin Quignon using ESR-quartz from fluvial quartz grains provides an age ranging between 709 ± 55 ka and 650 ± 37 ka to the glacial fluvial deposits. The study of the malacofauna collected at the Carrière Carpentier, as well as the mammal fauna collected since the end of the nineteenth century in the Marne Blanche, place these alluvial deposits in an interglacial during the second half of the Cromerian (Cr. III), or to marine isotopic stage (MIS) 15 (Voinchet et al., 2015; Antoine et al., 2016; Bahain et al., 2016) ESR-quartz ages and ESR-U/Th on mammal teeth attributes an age of 584 ± 48 ka to these interglacial fluvial deposits of the Marne Blanche. This is perfectly in line with the position of the Carpentier quarry fluvial formation in the Somme terrace system (Antoine et al., 2007; Antoine et al., 2016) (Figure 2).

In addition to geological data, archaeological material (bone remains, lithic tools) from the old quarries of Moulin Quignon, deposited by Boucher de Perthes at the Museum National d’Histoire Naturelle, has been re-examined, in particular the carefully annotated bifaces. Their stratigraphic location is sometimes indicated by labels written by the hand of Boucher de Perthes. However, we cannot distinguish the pieces found by Boucher de Perthes from those brought by workers and given to Boucher de Perthes. Some could come from other sites. The revision of this material provided the first proof of the authenticity of the site, even though it is still impossible to specify the stratigraphic position of these pieces (Moncel et al., 2016). The few malacofaunal remains collected by Boucher de Perthes at Moulin Quignon, and curated in the collections of the MNHN, include Cromerian species confirming such antiquity (Bahain et al., 2016).

During glacial periods, the palaeo-Somme valley was one of the main tributaries of the Channel River, flowing into the Atlantic, and was a likely major route for human migrations between continental Europe and southern England (Ehlers et al., 2011) (Figure 1). The Somme River exhibits a stepped Quaternary fluvial terrace system incised into Upper Cretaceous chalk and composed of 10 stepped alluvial formations spread between +5/6 m and +55 m above the maximal incision of the present-day valley in the Cretaceous chalk bedrock (Antoine et al., 2007) (Figure 2A). The relative height of the contact between the successive alluvial sequences and their respective bedrock steps can be used to securely associate archaeological layers such as Moulin Quignon with the Somme River system. In addition, in the Somme valley, each alluvial formation corresponds to the morpho-sedimentary glacial-interglacial cycle.

The terrace series can be summarized by the following phases ( Figures 2A,B–4 ):

(1) Early glacial: heterogeneous coarse sediments made up of slightly reworked large chalk blocks and flint nodules embedded in a calcareous sandy matrix, including some thin lenses of fluvial calcareous silts (slope deposits slightly reworked by fluvial processes). These deposits (between 3 and 4 m thick) are only preserved in the outcrops located very close to the former slope of the valley, as in the Cagny-La Garenne or Carpentier quarry sequences. Overall, they are attributed to continental temperate conditions.

New fieldwork was recently conducted at the Carpentier, Léon and Moulin Quignon quarries, all located on the same +40 m alluvial formation of the Somme system (Antoine et al., 2016; Antoine et al., 2016). The stratigraphic sequence of the Carpentier quarry (average altimetry of the top of the sequence +31 m a.s.l.) is used as a reference to describe the whole sequence of this alluvial formation. It consists of a succession of fluvial gravels covered by whitish fluvial calcareous sandy silts with calcareous nodules (“Marnes Blanches” = White Marl) (Antoine et al., 2016). The fluvial succession is then covered by a slope sequence composed of hillwashed sands and thick clayey soliflucted gravel beds locally trapped in deep sinkholes. During the new fieldwork performed from 2010 to 2014 at this site, no unquestionable artefacts were discovered in the White Marl or in the underlying gravel layer, in keeping with Victor Commont’s assertions, while several faunal remains were recovered in the White Marl. Moreover, no clear evidence of human intervention can be shown on the bones although the long bones systematically show helicoidal fractures generally thought to be typical of human intervention (marrow extraction) (Antoine et al., 2016). The rare potentially flaked flints were considered to be geofacts due to post-depositional effects (shocks and pressure flaking between the largest flint blocks during deposition). Two Acheulean bifaces were discovered at the base of the slope deposits directly overlying the fluvial sequence, i.e., at the same stratigraphic location as the pieces recovered by Commont, (1909). Similar types of oval-shaped bifaces were even described and related to the “Acheulean” pieces found in the gravels of the Fréville Formation in Amiens (Commont, 1909) and at the Moulin Quignon site in Abbeville (Aufrère, 1937, Aufrère, 1950; Breuil, 1939). Based on our observations and analyses, the artefacts of Boucher de Perthes are most likely coeval with the end of MIS 15 or with an early stage of MIS 14, between 550 and 500 ka (Antoine et al., 2016).

In this context, new investigations were undertaken in 2016–2017 at Moulin Quignon (50°06′18″N/1°50′89″E, Figures 3, 4), as a result of the redevelopment of the suburb of Abbeville, leading to the rediscovery of this emblematic Palaeolithic site 170 years later (Antoine et al., 2019). In 2016, seventeen 4-to-5-m-deep test pits were excavated, resulting in the discovery of undisturbed fluvial gravels and sands, well preserved below thick layers of reworked sandy gravels (from the former quarry) and modern dump deposits. The average bedrock altitude measured at Moulin Quignon (26.5 m a.s.l.) in test pits and in the archaeological excavation is very close to the altitudes obtained for the Léon 1 and Carpentier quarries, located 200 and 400 m northwards respectively (relative height: 39–40 m). This approach demonstrates that the three alluvial remnants are located on the same bedrock step and correspond to the same alluvial formation.

The relative elevation of this Moulin Quignon alluvial formation above the altitude of the maximum incision of the chalk below the present-day valley (+40 m) correlates it with Alluvial Formation VII of the Somme system (Antoine et al., 2007).

In addition, we compiled all the available stratigraphic information for this alluvial formation from the area located between Carpentier Quarry and Moulin Quignon, taking into account our recent observations at Carpentier, Léon 1 and Moulin Quignon quarries between 2012 and 2017, as well as the former data published by D’Ault-du-Mesnil, Commont and Breuil (D’Ault du Mesnil, 1889; D’Ault du Mesnil, 1896; Breuil, 1939).

This approach was completed by a new test pit campaign during spring 2019 over an abandoned area further to the north (140–180 m) of the Moulin Quignon 2017 excavation, closer to Carpentier Quarry (Figures 3, 4). During this extensive survey (17 test pits on 2,600 m²), in situ remnants of fluvial gravels and sands were discovered (thickness about 1 m), overlying the same chalk bedrock step as in Carpentier Quarry and Moulin Quignon 2017.

These results strongly support the direct stratigraphic correlation between the fluvial sequences from Moulin Quignon and Carpentier Quarry, despite their distance.

The facies and stratigraphic succession observed in the new Moulin Quignon excavation are very similar to descriptions published in the nineteenth century by Boucher de Perthes and Prestwich, in terms of overall thickness of the sequences and sedimentary facies. Boucher de Perthes described a sequence composed of gravel beds and sand layers, with a thickness of about 4.5 m. This is perfectly compatible with the difference between the average altitude of the bedrock step measured at the base of the sands and gravels during the new excavations and test pits (26.5 m) and that of the surface of the topsoil (30.5–31 m). Moreover, at the very base of the fluvial sequence, we also found only at Moulin Quignon a 0.4–0.5 m-thick layer of gravels strongly coloured by iron and manganese oxides, which was formerly described by Boucher de Perthes (“Couche noire”) and Prestwich (the Black band) in the same stratigraphic position, and in which we also discovered Palaeolithic artefacts.

The stratigraphic synthesis shows that the Carpentier interglacial deposit, the “White marl,” extends to the south towards the area corresponding to the Chemin de Fer and Léon 2 quarries located at less than 100 m from the area explored at Moulin Quignon in 2016–2017 and that they overlie the coarse gravels and sands from the Moulin Quignon Formation.

The interglacial calcareous facies of the “White marl” is thus only preserved in sequences located in the external part of the alluvial formation. This configuration results from the occurrence of a lateral channel close to the right bank of the Palaeo-Somme valley, whereas interglacial deposits are not preserved towards the central part of the former river valley, showing a markedly thicker lower gravel unit (3–3.5 m), like in the Moulin Quignon area. Thus, combining the new observations and former data, we can demonstrate:

Given the stratigraphic information exposed above, the geochronological data from the deposits of the Carpentier quarry, and especially the ESR and ESR/U-series dates from the interglacial deposits of the White marl, can be used to reliably infer the age of the Moulin Quignon findings. Besides, three samples of sandy sediments were extracted during the 2017 archaeological excavation from the main stratigraphic units and in situ gamma-ray measurements were performed in each sampling hole. The results obtained for the three samples from Moulin Quignon, using the aluminium (Al) and lithium titanium (Ti-Li) centres of quartz, are consistent with each other (overlapping of uncertainty domains at 2 σ). An average age of 672 ± 54 ka was obtained for the whole fluvial sands and gravels formation of Moulin Quignon, confirming the antiquity of the site.

Although the average age of the Moulin Quignon fluvial deposits corresponds to the early phases of MIS 16 (Glacial b of the Cromerian Complex), and thus to clearly cold and markedly different conditions compared to the “Marne Blanche” deposits, no accurate palaeoenvironmental reconstruction can be attempted due to the absence of bio-indicators in the Moulin Quignon gravel formation (fully decalcified deposits).

The interpretation that the Moulin Quignon gravels and sands were deposited in a periglacial environment is based on the occurrence of some very large rounded Tertiary sandstone blocks (up to 0.5–0.8 m in length), at the base and in the fluvial gravels, reworked during episodes of periglacial mudflows from the slopes of the valley towards the alluvial plain. Indeed, in the alluvial formations of the Somme River, as in numerous rivers of Northwest Europe, these large ice-rafted blocks are described within or at the very base of Quaternary alluvial formations. The reworking of these sometime huge rounded blocks (>1 ton) from the slopes to the alluvial plains, and their subsequent transport by fluvial systems, imply that major episodes of solifluction occurred on the slopes and that extremely strong river dynamics transported large ice-rafted blocks and heavy loads of coarse and poorly sorted material (flint nodules and chalk blocks) during spring break-ups and floods.

These processes characterise full glacial conditions and are typical of braided river systems in Quaternary periglacial and present-day northern tundra environments. The deposition of the Moulin Quignon sequence during periglacial conditions is also shown (Figures 3, 4) by 1) the occurrence of well-sorted flint gravel layers including a main stratified sand lens deposited in a shallow channel structure (at least 5–6 m wide) typical of braided river systems (Gr-n, Grs-j, Grs-br and Fs), 2) a very low proportion of sandy matrix as well as iron oxide and manganese coatings on the gravels (Sa-br), that are both typical of the basal units of the fluvial sequences of the other Somme terraces and, 3) the occurrence of numerous large and poorly rolled flint blocks (20–40 cm in length at the base of Gr-n), as generally found at the base of the gravel bars and islands in braided river systems (lag deposits).

Finer facies (grey silty sands, yellow laminated sands) observed in Moulin Quignon sections that could have reflected different climatic conditions, but no bio-indicators, are preserved due to post-depositional weathering processes. The thin layers of clayey-silty sands occurring at the base of the main sandy lens have been tested for pollen without success. These fine-grained sediments were probably deposited in the same braided river environment in an abandoned shallow channel between gravel bars at the end of a flood episode.

The homogeneous character of the physical characteristics (abrasion and patina) of the archaeological pieces from Moulin Quignon excludes reworking from an older terrace formation and no traces of occupation have ever been described in the older alluvial sequences of the Somme system despite more than 150 years of intensive research.

The archaeological excavation showed that the artefacts are not regularly distributed in the fluvial deposit but are rather localized in spatially discrete concentrations separated by archaeologically sterile zones.

The Acheulean artefacts discovered at Moulin Quignon thus probably represent human occupations during a cold period, located in the former alluvial plain of the Somme River on a gravel bar close to the right bank of the valley, at the margins of the valley. In this position the site was regularly flooded during spring floods and the flint artefacts were probably displaced over short distances (a few metres), before being buried in the sandy-gravelly material.

However, it is highly likely that human occupations from this period only correspond to short incursions of human groups during the summer season. Taking into account the wide dating error bars, it is also possible that the Moulin Quignon human occupation took place during a short interstadial period (≤1–2 ka) characterised by a milder climate (especially during summer season), as is the case for example for Upper Palaeolithic hunters during the Last glacial period in the area (Renancourt: Paris et al., 2016; Moine et al., 2021), or for the Acheulean occupation at Cagny-la-Garenne during the Early MIS 12 glacial period (Tuffreau and Lamotte, 2010; Lamotte and Tuffreau, 2016).

At that time, Acheulean “pioneers” probably followed the tracks made by the Somme valley and river banks along a straightforward trajectory. In this context, the occurrence of a confluence between the Somme and the Scardon River valleys, as is the case in Moulin Quignon, was a particularly good marker in the landscape. This configuration is also auspicious for large mammals, and was particularly attractive for the first Acheulean “pioneers” in the Somme Valley.

The lithic material attributed to Moulin Quignon is abundant in museum collections; at the Musée de l’Homme and the Institut de Paléontologie Humaine (Paris), or in other French museums, despite partial destruction during the two world wars. However, for most of the pieces, no precise geographic or stratigraphic location is indicated. These series are undoubtedly from surface collections and not from archaeological excavations or quarries.

In 1860, Jacques Boucher de Perthes gave the Museum National d’Histoire Naturelle Paris a sample of the archaeological material that he collected in Abbeville and its region, consisting of several hundred pieces (Hurel, 2014). The collection of these pieces began in 1840 and the last entry in the inventory is in 1859. There is no doubt about the authenticity of the pieces that make up this corpus (with some very rare exceptions). Twenty-two of them bear various indications and are recorded in the inventory. There are, in particular, bifaces from Moulin Quignon with indications written by the hand of Boucher de Perthes between 1845 and 1863.

Our goal was to identify the pieces with clear stratigraphic attributions handwritten on labels stuck directly on the piece or on the box. These annotations have been systematically checked in the inventory.

To date, only fourteen pieces of often very patinated flint can be attributed to a layer or a “depth” in the sequence of Moulin Quignon (3, 4 and 5 m deep). According to the inscriptions, which indicate a clear stratigraphic attribution, these lithic pieces were discovered by workers in the presence of Quatrefages or collected by Boucher de Perthes himself. The two Quatrefages bifaces, extracted in April 1863 from the same layer as the “mandible”, are now curated with the collection donated by Boucher de Perthes in 1860, in the reserves of the Musée de l’Homme (Nos. 2977 and 2978).

In addition, there are two bifaces (Nos. 2157-31 and 2837/27) without a precise stratigraphic position, but inventoried as coming from the sequence. They are described separately.

Finally, two other pieces (Nos. 2155-29 and 1931-27), a small ball in flint and a gelifract (discovered in 1866), with no stratigraphic position but inventoried, were considered as natural and eliminated from the study.

Our study focused above all on the technological characteristics of the lithic pieces.

A total of 254 flakes, 15 cores, 5 bifaces and 4 shaped tools were discovered in situ at the base of the deposits in several stratigraphic groups (Gr-n, Grs-j and Grs-j) during the recent test pits (2016–2019) and excavation (2017) (Antoine et al., 2019) (Table 1). The material is made on flint collected from the local Middle Coniacian C4b (flint with green cortex), from secondary flint reworked in the Tertiary formations or from the Pleistocene gravels. Only eight pieces show traces of crushing and 19 are clearly flake-tools. The vast majority are fresh. The geological features of the sequence and the taphonomy of the artefacts indicate that they were moved little by natural processes (≤1 m). As there are no significant differences between the sub-layers, the corpus is studied as a whole.

The series is composed of 15 cores made on flint nodules, which is rather a low number compared to the number of flakes (n = 254) (Tables 2–4). The core/flake ratio shows a core deficit.

The organization of removals on cores indicates short debitage sequences with unipolar or centripetal removals, on one, two (orthogonal or bifacial secant) or multiple surfaces. Most of the cores, however, are largely cortical with few removals.

The nodules used were of various sizes, often with irregular shapes, and the flint is of medium quality (sometimes presence of inclusions). No particular selection of nodule morphology seems to have been applied.

The scars, as well as the diversified nodule sizes, indicate the production of flakes of disparate shapes and sizes. No core preparation is visible and debitage exploits the natural morphology of the nodule. The relatively low frequency of hinged scars on the core surfaces seems to indicate well-controlled management of knapping angles despite the low proportion of prepared striking platforms.

Most of the cores are unifacial with one (or two) series of unipolar and centripetal removals. The striking platform is cortical or with preparation limited to reducing the angle (around 60°), when necessary. Debitage produces thick flakes with a large and thick butt, and frequently uses the natural cortical convexities of core edges. Sometimes, the lateral convexities of the flaking surface are prepared by hinged removals. The core technology of these cores indicates efficiency with very limited predetermination.

Rare cores show more intensive flaking. They are unifacial, bifacial secant or multifacial, with unipolar, centripetal or crossed removals produced in one or several series. The striking platform is partially prepared with previous removals (Figure 6). Bifacial cores indicate flaking on two faces with two or three series of removals. Some small final removals on two cores can be considered as retouch or final debitage attempts. Two cores indicate that flaking can become independent of the natural geometry of the nodule. They show peripheral unipolar removals obtained from a prepared striking surface and could be described as “semi-tournant”-type cores. One other core surface shows additional debitage by invasive and hinged unipolar removals.

The largest core (163 mm × 150 mm × 92 mm) is a piece with three orthogonal surfaces. Two opposite surfaces and the edges of the core were flaked. Invasive centripetal removals on one face are obtained from cortical striking platforms or previous removals. Finally, peripheral and orthogonal removals are produced on part of the core periphery. Then there is alternating flaking of two orthogonal surfaces with some additional removals on the first flaked surface. The lower surface of the core shows some scars extracted from other cortical striking platforms. Traces of strong impacts on the edges, with angles close to 90°, suggest that the core was recycled as a hammerstone.

Most of the products come from on-site debitage. A large proportion of flakes are cortical, from the first phases of flaking (Figure 7). They measure between 40 and 80 mm long (Table 2). Most of them are “robust” and thick flakes with edge angles of around 70°. Cortical flakes indicate the flaking of oval or irregular nodules. Convergent unipolar or unipolar removals eliminate the cortex by a series of thick flakes. The butts indicate that striking platforms were prepared or cortical. A majority of flakes have a cortical back and some have a cortical butt-back, from the edges of the core or nodule.

Non-cortical flakes show that a longer debitage sequence can continue on the cores by unipolar, centripetal or crossed flaking. The frequency of backs for non-cortical flakes also indicates that flaking continued on core edges. Some more elongated flakes mainly result from the use of the scars and core edges. Flakes are often thick, and some are truncations of a convex or pyramidal flaking surface. Flake sections show abrupt facets and deep scars. Butts are flat, but also punctiform, dihedral and “faceted” (with more than two scars). Few flakes are hinged, confirming control of the striking platform, despite some evidence of hinged scars on cores. One of the main production characteristics is the presence of wide and thick butts with an acute angle (50–60°). Impact points are positioned far from core edges, as is the case for “Clactonian” type debitage (Ashton et al., 1994). We can mention two Kombewa flakes, resulting from the breakage of a nodule and the knapping of the broken surface.

Less than 31 flakes show retouch (one or more edges with marginal or irregular retouch), but only seven are unquestionably retouched: three convergent scrapers made by irregular retouch and two end scrapers on the largest elongated flake (128 mm) and a cortical flake, with retouch on both lateral edges (Figure 7).

Above the c. 45th N., the site of Pont-de-Lavaud in the centre of France yielded pollen grains associated with human occupations related to core-and-flake type assemblages and dated to 1.055 ± 0.055 Ma. Pollen grains indicate a forested environment in a very humid warm temperate climate. Hominids were therefore present during an interglacial temperate climate at this mid-latitude (Marquer et al., 2011; Despriée et al., 2018).

Plant and faunal remains from British sites such as Happisburgh 3 (MIS 21, 866–814 ka, or MIS 25, 970–936 ka) and Pakefield (MIS 19 or MIS 17, about 750–680 ka) attest however that the climate was rather continental and cool for occupations dated between MIS 25 and 17 at this latitude (Parfitt et al., 2005; Parfitt et al., 2010; Farjon et al., 2020). These two sites yielded series of cores and flakes, with no evidence of biface production. At Happisburgh 3, artefacts are associated with Pinus and Picea pollens and conifer wood and pine cones, indicating regional conifer-dominated forest. Temperature estimations point to mean summer temperatures between 16 and 18°C and mean winter temperatures between 0 and –3°, with similar conifer-dominated woodland to that of present-day southern Scandinavia. Local grassland is indicated by the range of grazers (Equus suessenbornensis, Bovidae and Microtus spp.) and indirectly, by preserved pollen from hyaena coprolites. Mammals include Mammuthus cf. meridionalis, Bovidae, Equus suessenbornensis and at least two species of Mimomys. An assemblage of 78 flint cores, flakes and flake-tools has been excavated according to studies from fluvial gravels and laminated estuarine sands and silts or glaciomarginal fans desposited in a proglacial lake complex (Cromer Forest-bed Formation CF-bF) (Gibbard et al., 2009; Gibbard et al., 2011; Gibbard et al., 2012; Gibbard et al., 2018; Hijma et al. 2012; Parfitt et al., 2010; West, 2019; West et al., 2019; Turner et al., 2020; Gibbard et al., 2021; West and Gibbard, 2021).

At Pakefield, macro remains and fauna from the Cromerian Forest-bed indicate warmer summers and mild winters (between 18 and 23°C for summer) with broad-leaf woodland. Macro-remains (Trapa natans, Salvinia natans and Corema album), beetles (Cybister lateralimarginalis, Oxytelus opacus and Valgus hemipterus) and the presence of Hippopotamus indicate warmer summers and mild winters (between 18 and 23° for summer). Mammuthus trogontherii, Stephanorhinus hundsheimensis, Megaloceros savini, M. dawkinsi, Bison cf. schoetensacki, and carnivores (Homotherium sp., Panthera leo, Canis lupus and Crocuta crocuta) make up the faunal assemblage. Pollen analysis indicates an interglacial with broad-leaf woodland including Carpinus. Mimomys savini and M. aff. Pusillus are both present.

The British sites of Brandon Fields, Maidscross Hill and Warren Hill yielded evidence of bifacial technology on flint. The age of the associated sediments was recently dated by ESR-quartz to 550–600 ka suggesting an age of MIS 15-14 for the earliest bifaces in Great Britain (Voinchet et al., 2015; Lewis et al., 2021). However, microfaunal data are not in keeping with the ESR dates, and place these sites in MIS 13 under a landscape of boreal species and open grassland (Candy et al., 2015).

At la Noira, in the centre of France, hominids were present at the end of MIS 17 and the beginning of MIS 16 and disappeared from the area when it became too cold (Despriée et al., 2011; Moncel et al., 2020c). The site is located along the Cher River, a tributary of the Loire River. The substratum of Tertiary lacustrine limestone (stratum a) is overlain by five strata. A coarse slope deposit (Stratum b) is covered by two sequences of fluvial sediments (Strata c and d). The artefacts are associated with the basal coarse sand of stratum b, the lower part of which consists of an accumulation of local “millstone” in Oligocene lacustrine limestone. The earliest archaeological level is located at the bottom of the sequence. The position of the artefacts between the deposition of slope materials on the limestone bedrock and later phases of gelifluction and cryoturbation, would suggest that hominins were present after the period of river incision at the beginning of a cold climatic stage. ESR dates (mean age of 655 ± 55 ka) place this stage at the beginning of MIS 16 (Moncel et al., 2013; Moncel et al., 2015, Moncel et al., 2020c; Moncel et al., 2021a).

At the Caune de l’Arago (France), levels P-Q are dated back to 532 ± 106 ka by ESR/U-series (Falguères et al., 2015). Sedimentological analysis shows that the Unit I deposit is essentially composed of sands blown into the cave by strong north-eastern winds during a dry and cold phase of MIS 14. The levels yielded assemblages with bifaces (Barsky and Lumley, 2010; Barsky, 2013). The faunal spectrum includes Ursus deningeri, Cuon priscus, Vulpes vulpes, Lynx spelaeus, Panthera cf. pardus, Equus mosbachensis, Stephanorhinus hemitoechus, Bison sp., Ovis ammon antiqua, Hemitragus bonali, Rangifer tarandus and Cervus elaphus (Moigne et al., 2006). Horse (28%), reindeer (20%) and bear (28%) are the most abundant taxa indicating an open and steppe landscape. The cave was alternately a bear den or a den for other carnivores and a habitat for short-term human occupations.

In Spain, at Atapuerca, Sima del Elefante and Gran Dolina yielded several archaeological levels from this period. At Sima del Elefante, lower units TE7 to TE16 show reversed polarity magnetization attributed to the Matuyama Chron, confirmed by cosmogenic nucleids with 1.2 Ma to TD9 (Parès et al., 2006; Carbonell et al., 2008; de Lombera-Hermida et al., 2015). The occupations were characterised by a humid and warm climate before a period of cooling recorded at MIS 22 (ca. 900 ka). Pollen shows the presence of Pinus sylvestris. Unit TE16 is dated to 804 ± 47 ka and 864 ± 88 ka. The luminiscence ages obtained for TE17 are 724 ± 43 ka and 781 ± 63 ka.

Layer TD6 at Gran Dolina, composed of 2–2.5 m of blocks and gravels in a poor clayey matrix, has been divided into three lithostratigraphical units and yielded hominin remains (Carbonell et al., 2005; Carbonell et al., 2010). TD6 has a pre-Matuyama negative polarity (>0.78 Ma, MIS 21) (Blain et al., 2013; Parès et al., 2013) confirmed by biostratigraphy and radiometric dating by electron spin resonance of optically-bleached quartz and U-series methods (Rodriguez et al., 2011; Duval et al., 2018). Sub-unit TD6-2 is characterised by warm and humid climatic conditions (continental Mediterranean climate) (Rodriguez et al., 2011). The landscape indicates open woodland and steppe. This can be correlated with the transitional phase of forest development during a cold to warm climatic transition (possibly correlated with the MIS 22/21 transition). A moderate increase in the presence of open dry taxa occurs in several samples from sub-unit TD6-3 to TD6-1 and indicates steppe habitats in a mosaic environment.

The site of Barranc de la Boella in Spain is located in a fluvio-deltaic area with an aquatic environment. Six lithostratigraphic units (Units I–VI, from bottom to top) are described (Mosquera et al., 2013; Vallverdú et al., 2014; Mosquera et al., 2015; Mosquera et al., 2016), including one with crude bifacial tools. Fieldwork has been carried out in three different localities, at the same stratigraphic position: Pit 1 (P1), la Mina (LM), and el Forn (EF). Fossils recovered up until now are almost exclusively from Unit II (Vallverdú et al., 2014). The P1 faunal sample is almost entirely composed of mammoth remains, associated with a lithic sample of more than 100 remains, including a pick and a cleaver cited as proof of the early arrival of Acheulean technology in Europe. A solid geochronological age for the lithic assemblages found in BB Unit II could indicate a position in the late Early Pleistocene or late Matuyama Chron (0.96–0.78 Ma). The presence of Mammuthus meridionalis and the morphology of Mimomys savini molars from the top of Unit II at the three localities support the age of the archaeological deposits. The faunal assemblage of level 2 (Unit II) of the Pit 1 locality is dominated by Mammuthus meridionalis, associated with a few Dama cf. vallonetensis and Equus sp. (Pineda et al., 2015; Pineda et al., 2017).

Notarchirico in Italy is a 10-m-thick fluvial-derived sedimentary sequence rich in volcanic materials from the Monte Vulture stratovolcano (Piperno, 1999; Lefèvre et al., 2010). Hominins circulated regularly along these water channels or lakeshores. Several radio-isotopic methods (ESR and ⁴⁰Ar/³⁹Ar) were applied to quartz grains and volcanic minerals (mainly sanidines and clinopyroxenes) in the sediments. The results constrain the period of occupation of this site to between 695 ± 6 ka and 614 ± 12, spanning the entire cold stage MIS 16 and the end of MIS 17 (Pereira et al., 2015; Moncel et al., 2020b). The site thus records one of the earliest known occurrences of bifacial technology in Southern Europe (levels B, D, F, G) (Piperno, 1999; Moncel et al., 2020a; Santagata, et al., 2020). Raw material procurement and core technology are similar throughout the sequence and focus on the production of small flint end-products. The oldest handaxes found at Notarchirico (levels F and G) are now securely dated to more than 670 ± 4 ka and prove that hominid populations lived there as early as at the beginning of MIS 16 and at the end of the MIS 17. The faunal remains found in the lower levels (E/E1, F, and G) are Elephas antiquus, Dama clactoniana, Bos primigenius, and Bison schoetensacki. The faunal assemblage of Notarchirico belongs to the “Ponte Galeria” faunal unit. From the microfaunal point of view, micro-mammals indicate a dry climate typical of the Middle Pleistocene glacial period in the Italian peninsula. Results from the top of the sequence show that vegetation was characteristic of open and cold environmental conditions with Poaceae meadows (Piperno, 1999).

The stratigraphy of Isernia-la-Pineta (Italy), near Notarchirico, is composed of five units (Gallotti and Peretto, 2015). Unit 3 contains the archaeological deposits and is sub-divided into three sub-units. Unit 3a is the richest level, composed of a high concentration of flint and limestone artefacts (without bifaces). Recent ³⁹Ar/⁴⁰Ar dates of 583–561 ka corresponding to the MIS 15/MIS 14 transition were proposed by Peretto et al. (2015) for this site, based on more than 90 single crystal sanidine dates. The faunal list is composed of Elephas (Palaeoloxodon) antiquus, Hippopotamus cf. antiquus, Stephanorhinus hundsheimensis, Bison schoetensacki, Praemegaceros solilhacus, Cervus elaphus cf. acoronatus, Dama cf. roberti, Capreolus sp., Sus scrofa, Hemitragus cf. bonali, Panthera leo fossilis, Panthera pardus and Homo. The most abundant carnivore remains are those of Ursus deningeri (Thun Hohenstein et al., 2009; Peretto et al., 2015). The large number of herbivores indicates that the Isernia region was an area of open vegetation of woodland steppe, rich in pastures, with bison herds and numerous pachyderms during the MIS 15/MIS 14 period (Thun Hohenstein et al., 2009; Pineda et al., 2020). This kind of environment indicates a climate with a long and arid season coupled with a shorter one with concentrated annual rainfall (Lebreton, 2002; Orain et al., 2013).

The available sites are mainly open-air habitats beside rivers, lakes or swamps, except some sites in caves. This may be due to the low number of sites for this long period of time and conservation conditions. However, it may also indicate common patterns of territory and resource management. Regardless, such locations allow for easy access to quantities of local raw materials, as the sites are located on shores with abundant stones, and herbivore carcasses. Faunal remains indicate, when preserved, accumulations of large herbivores, including megaherbivores, along rivers and possibly scavenging and butchery activities in competition or not with carnivores (i.e., Barranc de la Boella, Notarchirico, Isernia-la-Pineta) (Piperno, 1999; Vallverdu et al., 2014; Gallotti and Peretto, 2015; Mosquera et al., 2015; Pineda et al., 2017; Moncel et al., 2020b). When visible, cut marks and fragmented bones demonstrate meat processing, butchery and scavenging on herbivore carcasses of different sizes, including megaherbivores (Pineda et al., 2015; Pineda et al., 2017; Pineda et al., 2020). Sites seem to be either multi-activity sites (la Noira; Hardy et al., 2018) or specialized sites (possibly scavenging sites at Isernia-La-Pineta (Longo et al., 1997; Thun Hohenstein et al., 2009; Pineda et al., 2020). They point either to mobile groups using local raw materials with little evidence of semi-local stone procurement, or less mobile groups with strong ties to a specific place and environment (Gallotti and Peretto, 2015; Mosquera et al., 2015; Ollé et al., 2016). Isotopes of strontium on the human tooth of Isernia-La-Pineta indicate the relatively limited mobility of the corresponding individual (Lugli et al., 2017).

No correlations have been established at la Noira between types of tools and function, and crudely-made bifacial tools appear to have been used in the same way as bifaces (Hardy et al., 2018). This suggests that the lithic assemblage could represent above all a cultural package, and not merely tool kits devoted to specific functions. Sites described as butchery sites for large herbivores of the same age indicate that small flakes were widely used and were possibly as valuable as heavy-duty tools for these activities (for instance Isernia-La-Pineta in Italy, Thun Hohenstein et al., 2009; Pineda et al., 2020).

Raw material procurement is mainly local and related to the type of site (along water areas and rivers). Hominins used various available stones, such as cubic blocks, slabs, nodules or cobbles/pebbles and selected stone quality or thickness depending on the type of activity (Moncel et al., 2020c; Santagata et al., 2020). Medium or large-size nodules or slabs of local flint are used at Moulin Quignon and millstone slabs at la Noira while small chertlocal nodules were recovered at Notarchirico, mainly for debitage or direct retouch. Limestone pebbles are collected à Notarchirico for pebble tools and LCTs. At Moulin Quignon, all the parts of irregular flint nodules were used, even some rounded extremities. Nodule quality is mainly good. At la Noira, raw material quality seems to have been a criterion for selection, and slabs with no evidence of frost cracking were chosen (Despriée et al., 2016; Moncel et al., 2020a). Three main lithotypes were identified at Notarchirico for the small nodules used for the debitage or direct retouch: silicified calcarenites (flysch chert), nodular chert (carbonate platform) and radiolarite (basin) (Synthem of Palazzo San Gervasio (Eramo et al., in Moncel et al., 2020b). No clear selection of a specific type of chert or radiolarite is observed, whatever the occupation phase. Selection seems to have focused on nodule size and shape, for flaking or direct retouch (thick and flat small nodules for abrupt and denticulate retouch). At Pont-de-Lavaud, quartz was used, perhaps as a result of its abundance near the site (Despriée et al., 2018). The only exception is the Caune de l’Arago with some long-distance stones collected in a perimeter of 15–30 km, especially for debitage and LCTs (Barsky, 2013).

Flaking comprises long and short sequences of removals, showing a diversity of hominin skills and cognition (unifacial and multifacial, bipolar or discoid-type). However, flaking sequences cannot always be considered as expedient. We observe adaptation to the shape of the blank with common rules, even on quartz. These rules indicate flexibility and a selection of stone shape for the best results. Platforms are prepared if necessary. Some reduction sequences are more structured and more independent of blank shape and fully managed on good-quality stones, as clearly observed at la Noira, Moulin Quignon and the slightly younger site of Isernia-la-Pineta.

Product size is highest at la Noira and Happisburgh (flakes >145 mm), as a result of flaking nodules or slabs, and smallest at Moulin Quignon on medium-sized nodules. At Notarchirico, the use of the small available nodules results in very small flakes (10–20 mm) and very small and thin cubic nodules are sometimes directly retouched. However, there is not always a strict correlation between the natural blank and product size. Small flakes (40 mm long) are also produced at la Noira on millstone cores and large flakes (80–100 mm long) are obtained from large cobbles at Notarchirico (Supplementary Figures S1–S3).

Platform angles are acute for plain and thick butts at Moulin Quignon, Happisburgh 3 and Pakefield, and more obtuse (80–90°) on slabs at la Noira and small cubic nodules at Notarchirico. The impact of stone shape alone cannot explain production aims. We observe thicker and more robust flakes in the northwest with acute retouched and unretouched cutting edges. Relationships with activities must be examined.

The ratio of flake-tools differs (mainly unilateral or bilateral retouch and some notches) depending on the site and TD6 and Happisburgh 3 seem to be an exception for Mode 1 series. In most sites, flake-tools bear one or several retouched cutting edges. Angles are mainly dependent on the type of raw material, and are lower for flint with acute unretouched angles. The diversity of the number of retouched edges, creating simple, multiple or an association of simple tools on the same blank, can be correlated to needs. At the Caune de l’Arago, discoidal working was applied to quartz or siliceous stone cobbles from 15–30 km away. Retouched flakes are abundant and mainly pointed.

The number of bifaces with overall volume management is low for each site and we cannot assess whether or not these tools were more mobile than other tools. Raw materials are local. Despite the small numbers, the series associate various intensities of shaping, with crudely-made tools and very reduced bifaces. The artefacts from la Noira illustrate the diversity of the heavy-duty component, with diverse heavy-duty tools made on irregular fragments of slabs (Iovita et al., 2017; Moncel et al., 2020c).

Series are characterised by the diversity and lack of standardization of bifacial tools and the heavy-duty component, although some are more homogeneous, due to the raw materials. Cutting edge angles vary depending on the sites and stones, as does biface size, which is smaller at Notarchirico, longer at la Noira. However, we have to keep in mind that some bifaces, like at Brandon Field, Moulin Quignon and la Noira, may have undergone resharpening, resulting in decreased tool size. Raw material size could also partly explain the diversity of biface size (Supplementary Figure S4). At Caune de l’Arago, bifacial tools on a variety of raw materials include well-worked bifaces of various sizes with overall volume management. There are also some cleavers on flakes.

The limited number of sites and the diversity of contexts hinder comparisons of the specific features of each site and the investigation of adaptation to colder climates and environmental contexts. Stone types do not seem to be associated with specific flaking methods, except for quartz. Debitage modes on siliceous rocks are similar, with adaptation to stone shape (slabs, nodules…). Flaking products differ according to raw material size, ranging from large flint flakes in the northwest of Europe to mainly small flakes in the South of Italy. Observations are similar for LCT management and production. We also observe a wide diversity of cutting edge angles among the light and heavy-duty component, possibly pointing to diverse functional tool kits and cultural packages. The diversity of skills, routines and strategies among populations in Western Europe between 1 Ma and 600 ka seems to have been favourable to adaptation to varied environmental contexts.

In Moncel et al. (2018a), we discussed current data from the archaeological record for which the climatic context cannot be accurately assessed. The link between global climate variations and regional environments is difficult to establish for this period of time. During the main Middle Pleistocene glacial periods such as MIS 16 and MIS 12, the ice sheet probably had an impact on hominin occupations in Eurasia, erasing northern evidence of human sites from previous interglacials. In addition, many sites are not dated accurately enough to be confidently correlated with regional climatic and palaeoenvironmental records. The position of a human occupation cannot always be strictly linked to a glacial or interglacial solely on the basis of age and stratigraphic position. On the other hand, faunal remains and thus biostratigraphy are markers of specific environmental conditions. Furthermore, occupation duration (i.e., seasons, months, years?) in a specific site cannot be directly appraised. The archaeological characteristics of sites are related to hominin occupation and faunal assemblages can be biased by the nature of the site (e.g., butchery site, seasonal settlement) or by the origin of the fauna (selective hunting, opportunistic butchery…).

At northern sites, records indicate that hominin occupations took place either in dry steppic environments or woodland settings. Some sites yield more reliable environmental data based on multiple proxies such as the British sites (Happisburg 3 and Pakefield), while others are less relevant due to the lack of well-preserved remains (Bytham River sites, Somme Valley sites) (Antoine et al., 2007; Gibbard et al., 2009; Candy et al., 2015). For la Noira (Centre France), hominins disappeared when very cold and dry conditions emerged at the beginning of MIS 16, considered as one of the harshest glacial phases since 1 Ma. This would account for the chronological gap in hominin occupations in the Centre of France and in the Somme Valley between 700 and 500 ka (“bottleneck”) up until now (Despriée et al., 2011; Moncel et al., 2020a).

In Southern Europe, if we enlarge the chronological span of the study, Spanish sites suggests recurrent humid and warm environments during occupations but steppic environments are also recorded, for instance at Orce dated to around 1.2 Ma (Rodríguez et al., 2011; Huguet et al., 2013; Rodríguez-Gómez et al., 2014; Rodríguez-Gómez et al., 2016; Blain et al., 2021). At Atapuerca, the early sequences of Sima del Elefante, lower units TE7 to TE16 (Parès et al., 2006; de Lombera-Hermida et al., 2015) were characterised by a humid and warm climate before the cooling recorded from MIS 22 onwards (ca. 900 ka) (Cuenca-Bescós et al., 2011; Blain et al., 2013; Parès et al., 2013). Then the landscape indicates open woodland and steppe correlated with the transitional phase of forest development (possibly correlated with the MIS 22/21 transition) (Rodriguez et al., 2011). The site of La Boella (0.96–0.78 Ma) indicates a mixture of temperate environments (Vallverdú et al., 2014; Mosquera et al., 2015; Mosquera et al., 2016). In Italy and the South of France, dry MIS 16 or 14 environments are associated with occupations, such as Notarchirico, in both MIS 17 and 16, without clear differences in fauna and occupation along lakeshores or water channels (Pereira et al., 2015; Moncel et al., 2020b). This suggests less intense glacial phases in the south compared to the north.

The possibility, that Northwest Europe, including Northern France and Southern England, was at the limit of the “Oekoumen” of Homo heidelbergenis of hominins leaving in Europe at that time, especially during the severe MIS 16 glacial event (around 650–620 ka), has be discussed by the scarcity of ocupations. Evidence of occupations during interglacial events or at the transition are more common (see for instance Boxgrove, MIS 13, Souther England; Robert and Parfitt, 1999; Pope, 2002; Leroyer, 2016; Pope et al., 2017). However, some traces of occupations show that hominins could tolerate the environmental conditions and especially cold conditions during glacial events as “permanent residents” or “punctual visitors” (cf. Cohen et al., 2012; MacDonald et al., 2012) and outperform their capabilities and not completely depopulate these territories. Southern climatic data suggest that humidity and temperatures were the two main prerequisites for human occupations (Jouzel et al., 2007; Joannin et al., 2011; Combourieu-Nebout et al., 2015). Pakefield climatic data suggest cold winters but some faunal and plant evidence suggests that low temperatures were possibly rare (Parfitt et al., 2005). To settle these north-western territories, with cold and arid climates, with similar technologies, hominins may possibly have developed specific strategies that cannot be seen in the stone tools. Due to the low quantities of human fossils, no discussions or comparisons of the anatomy of hominins living in Northwest Europe can be undertaken. Indeed, the Mauer mandibule found in Germany merely gives us an idea of the robustness of the hominin living at this latitude (Wagner et al., 2010; Wagner et al., 2011; Fiedler et al., 2019), but we have no idea of the anatomical plasticity of hominins such as Homo heidelbergensis during long periods of slow climatic changes (Hosfield and Cole, 2018). de León et al. (2021) show that the brain size of Homo enabled him to adapt to various scenarios and environments. So far, there is no evidence of the use of fire and taphonomic processes could have destroyed traces of fire if it existed before MIS 11 (i.e., Ashton et al., 2016; Gowlett, 2016). Faunal remains are not preserved at Moulin Quignon or at la Noira for instance. However, at Notarchirico and Isernia-la-Pineta, the bones do not show any evidence of the use of fire and the scant evidence recorded at the Caune de l’Arago could be due to wind or water transport of micro-charcoals (Deldicque et al., 2021). Seasonal mobility and visits of human groups following migrations of herds and big game could explain the sporadic presence of hominins in these northern regions (i.e., Dennell et al., 2011; Fielder et al., 2019). However, we also have to consider the possibility that hominins could permanently occupy these areas and find strategies to survive in more or less harsh or mild episodes.

Despite footprints suggesting bare feet in some situations and family groups moving in swamps to find food (wet sediments at Happisburgh; Ashton et al., 2014), various solutions can be envisaged to survive winter conditions with clothes (animal skins) and increased energy intakes, ingesting the necessary quantity of food and high nutrition (fat) by extracting animal marrow or collecting plants for instance. Ashton and Lewis, 2012, Ashton, 2015; McDonald et al., 2018, Hosfield et al., 2016 and Hosfield and Cole, 2018 have developed detailed accounts of what was required to survive cold conditions in these latitudes. We also have to bear in mind that hominins could have occupied smaller and specific areas with microclimates, such as sunny parts of valleys or slopes protected from cold winds. Summers (as indicated by the mean temperatures estimated at Happisburgh 3 and Pakefield) may have been warmer than winters with sub-zero temperatures, few available plants and short daylight hours. Food storage could have been practised during winter, facilitated by a diversity of tools, combining light-duty and heavy-duty tools such as bifaces with long functional cutting edges. Moreover, the diversity of the landscape and rich environmental resources (i.e., around Happisburgh and Pakefield) could have been favourable to collecting plants (berries, nuts,…) for winter, as well as the large river estuary near the coast (marine and terrestrial resources) (Ashton et al., 2014). Preece and Parfitt, (2012) indicate cut marks on bison bones from Happisburgh (series collected in 1870s). Similar situations may have occurred at Moulin Quignon, located along the Somme Valley close to the confluence with the Scardon River (hot spot for large mammals?), with herds of game, or at la Noira along a tributary of the Loire River. Animal carcasses and plant resources could have been collected to prepare winter. The sites of Notarchirico and Isernia-la-Pineta in Italy confirm that hominins sought out banks of water zones for recovering meat on large herbivore carcasses (Elephants, bison, horses and cervids) or for hunting (Ben-Dor et al., 2011). At Carpentier Quarry (Antoine et al., 2016), new excavations have revealed that abundant fauna lived there during the deposit of the White Marl (MIS 15) with bovids, cervids, proboscideans, rhinoceros, horse, wild boar, indicating grassland and wooded areas. In Northern France, more than 20 taxa have been recorded during such interglacial phases (Auguste, 2009). Hominins at Moulin Quignon at the glacial/interglacial transition, or possibly during a short cooler event (unrecorded in the sequence), could have discovered this diversity of animals before the onset of cold conditions (Antoine et al., 2019). At Cagny-la-Garenne (Somme Valley, beginning of MIS 12, top of gravels), the fauna was composed of elephants, horses and cervids in a steppe-type landscape (Auguste, 1995; Auguste, 2009; Auguste, 2012).

The hypothesis of a strong relationship during the Middle Pleistocene between hominins and megafauna such as elephants is attractive for explaining the ability to occupy diversified territories even in the northwest (i.e., Reshef and Barkai, 2015; Solodenko et al., 2015; Barkai, 2016). Hominins would have taken advantage of the plentiful carcasses available along water areas with easy primary access and residues and micro-wear traces indicate domestic activities (Hardy et al., 2018; Moncel et al., 2020c). The sequence of the Caune de l’Arago illustrates the possible functional link between bifaces (LCTs) and megafaunal remains (i.e., elephant), during cold events (for instance MIS 14), suggesting that these tools could have been efficient and useful for cutting up large herbivore carcasses (Barsky and Lumley, 2010; Barsky, 2013).