It is a pleasure for us to thank Jean Clottes and Jean-Pierre Giraud for entrusting us with the different analyses and reevaluations of Cuzoul de Vers. We also extend warm thanks to Pierre Chalard for his support and advice, as well as to the Service Régional de l’Archéologie de Midi-Pyrénées for funding some of the dates presented here (Artémis program).
Two of the eleven measurements were funded by a grant from the Fondation des Treilles obtained by one of us (JMP). A big thank you to Bernard Marty and Olivier Gaiffe who facilitated access to the antler industries curated in the regional depot in Toulouse, as well as to Jean-Christophe Castel for his help in selecting the Solutrean remains. We also wish to thank François Lacrampe-Cuyaubère, docteur ès Data DeskTM and Olivier Ferullo for access to certain comparative data. The acquisition of these data owes a lot to the logistic help of Laurent Brou (for the loan of the drill bits) and Didier Galop (for his hospitality in the GEODE laboratory and the loan of material), and we sincerely thank them. We also extend thanks to the dating teams from the Lyon (CDRC), Saclay (CEA) and Oxford (ORAU) laboratories. We are grateful to Carolyn Barshay-Szmidt, Emmanuel Discamps and the two anonymous reviewers for their critical comments which considerably improved the quality of this article.

Introduction

1Cuzoul de Vers Rock shelter (Lot) was discovered in 1982 and excavated until 1986 under the direction of Jean Clottes and Jean-Pierre Giraud. It is located at the foot of a south-facing limestone cliff, at the edge of the RN 653, which separates the shelter from the Lot River, situated about 30 metres from the back of the shelter. During the course of a 5 year rescue excavation program, this site yielded a sedimentary record spanning all or part of the first half of the Last Glacial Maximum (DMG; sensu Mix, Bard, Schneider 2001, MARGO project 2009). In archaeological and sedimentary terms, this reference stratigraphy is made up of three layers attributed to the Upper Solutrean (c.31 to 29), overlain by 27 layers with lithic and bone material ascribed to the Badegoulian, from the very beginning of this culture (Early Badegoulian: l. 27 to 22) until the “classical” phase, from a historiographical point of view (Recent Badegoulian with raclettes: l. 21 to 1; Clottes and Giraud 2012).

2At the end of 2012, the site monograph was published, representing multidisciplinary studies carried out over a period of nearly 15 years (Clottes et al. dir., 2012). Due to the abundance and diversity of the remains (lithic industry, osseous industry, personal ornaments, fauna and microfauna, portable art, pigments, human remains, etc.), specialized analyses were planned and conducted from the end of the 1990s onwards, following approaches classically applied (Clottes and Giraud 2012). These studies led to interdisciplinary approaches and laid the foundations for new reflections on the socio-economic trajectory of societies known during the first half of the LGM in present-day France (Ducasse and Renard 2012).

5The main objectives of this work were to (1) test the older and more closely grouped AMS dates, (2) to try to gauge the extent and the chronological position of each of the main documented occupation phases (Upper Solutrean /Early Badegoulian/Recent Badegoulian) and in doing so, (3) to more accurately identify the “transition” phases from one to the other and/or any possible gaps in site occupation. Our contribution is, as of yet, preliminary, but it extends the results obtained in the scope of the monograph by proposing a critical interpretation of the new 14C measurements. These measurements were taken from samples selected following a simple but systematic protocol that backs up the value of the obtained results and allows for an ulterior critical analysis of our interpretations (cf. infra: technologically identified antler waste chrono-culturally characteristic of the Badegoulian; anatomically and/or species-determined faunal remains selected according to strictly defined location criteria for the Solutrean; in all cases the exact stratigraphic position of the pieces is known and can be verified).

Archaeosequence versus 14C dates: problems and questions raised

6The nine excavation seasons brought to light a differentially preserved site stratigraphy over two meters thick (Clottes and Giraud 1989, 2012). The construction of the road running alongside the Lot River and the ensuing developments aiming to improve traffic visibility truncated the sequence at the top of the infilling (i.e., l.s 1 to 4: only 2 to 10 m² conserved). In addition, some of the lowest levels, uncovered over more extensive surfaces (i.e., layer 13 to 31: generally excavated over more than 30 m²), are partly sealed by the present-day RN 653. Although most of the identified layers thus provide different levels of information, the archaeological material allows for the identification of diachronic variations which clearly cut through any possible representativeness problems linked to these different constraints. From the human occupations attributed to the end of the Solutrean (op. cit.; Renard 2008 and 2012), this exceptionally broad sequence allows for the observation of the development and evolution of new typo- and techno-economic standards introduced during the Badegoulian north of the Pyrenees. The identification of an occupation phase contemporaneous with the earliest Badegoulian was one of the major attractions of this site, as the development of this phase at Cuzoul de Vers has no equivalent throughout Western Europe.

7Several dates were obtained very early on and as the excavation advanced, in order to provide accurate data concerning the chronological position of each level and to allow for comparisons with the rare previously dated stratigraphies from this period (Laugerie-Haute: Evin, Marien, Pachaudi 1976; Abri Fritsch: Evin, Marien, Pachaudi 1978).

The classical method

What the AMS dates changed

11Two samples were entrusted to the Gif-sur-Yvette laboratory, already in charge of the first series of dates (tab. 1: l. 27 and 31). The Lyon laboratory looked after the preparation of another 11 pieces to be dated (purification and decontamination treatments), and then sent the samples to Oxford (nine cases) and Poznan (two cases) so that the measurements could be conducted.

12The results obtained presented some divergences with the previously available dates (fig. 1B). The most obvious of these is the ageing of the radiometric context: once the dates are calibrated to two sigma, the end of the Badegoulian occupations could be established at 19 ka cal BP with the classical method (l. 5: 15980 ± 150 BP, or 19640-18910 cal. BP), whereas the AMS dates establish this same limit around 21.5 ka cal. BP (l. 6: 18180 ± 170 BP, or 22417-21614 cal. BP). In addition, these data considerably reduce the chronological span of the Cuzoul de Vers occupations, varying from 5,000 years (classical 14C: from 24 to 19 ka BP) to 3,500 years (14C by AMS: from 25 to 21.5 ka BP). Finally, this double ageing/narrowing effect is completed by the disappearance of the hiatuses induced by the first two dating series (fig. 1B, on the left).

13This “reorganization” is probably more in keeping with the real rhythm of the sequence of occupations, but nonetheless comprises clear overlaps between the different chronocultural phases (e.g., impossible to distinguish between the Upper Solutrean and the Early Badegoulian: fig. 1B, on the right; Oberlin and Valladas 2012, p.80), as well as several inversions which raise problems even after the calibration of the measurements (e.g., l. 11 and 16).

Classical dates versus 14C by AMS: For a disjointed use

What about comparability between laboratories?

What about date inversions?

18The recent commencement of reevaluation work by two of us (CR and SD), focusing on l. 31 to 29, 28 and 27, now enables us to show that inter-layer refits linking the Solutrean level to the first Badegoulian levels (fig. 2: 2 links between c27 and c29) are extremely rare and probably only denote the existence of limited and localized disturbances, or may even be linked to stratigraphic interpretation difficulties (Clottes and Giraud 2012, p. 27). In the same way, the status of l. 28, located at the Upper Solutrean and Early Badegoulian interface should be reassessed.

19This layer is present over practically the whole surface of the excavated zone and is described as “(…) almost sterile, as if the rock shelter had been abandoned for some time after the Solutrean occupations” (Clottes and Giraud 1989, p. 87). It contains remains which appear to belong to two levels during stratigraphic projections: in places they appear to be clEarly distinct from the first recorded Solutrean levels but in direct contact with the remains from l. 27 (i.e. Early Badegoulian), and in others, they seem to be interleaved between l. 27 and 29 (i.e. Upper Solutrean, cf. infra fig. 7B). Attempts at refits and associations between these three layers confirm the “mixed” character of l. 28, as physical links were clearly established (fig. 2) between l. 27 and 28 on one hand (four cases), and between l. 28 and 29 on the other (one case), in addition to the combined presence of several fragments of pieces with flat and invasive retouch and of several antler flakes identical to the Badegoulian antler waste presented in figure 3 (n° 2 to 6; Pétillon and Averbouh 2012). In each case, several matches (litho-technological relations) were also documented, strengthening the hypothesis of an artificial assemblage, resulting from the association of distinct chronocultural elements.

20The intensification and extension of systematic inter-layer refit and association research to the most recent levels concerned (l. 11 and 15/16) will be one of the main aims of future analyses.

A new series of AMS dates

21Thus, the coherent radiometric framework that was consistent with the data available at the time (classical 14C: Bosselin and Djindjian 1988), gives way today to an older, tighter (between 24.5 et and 22 ka cal. BP, once the Poznan dates are removed) and much less coherent framework, with no regional or inter-regional published equivalent.

22In 2011, as part of an extended study on the diversity and evolution of Badegoulian reindeer antler production methods (Pétillon and Averbouh 2012; Pétillon and Sacchi in press), two of us (JMP and SD) attempted to carry out several direct dates on antler débitage waste, in order to assess certain technological associations (débitage by percussion versus groove and splinter technique: Pétillon and Ducasse 2012). Alongside the reevaluation of ambiguous contexts (e.g., Lassac: Sacchi 2003; Reverdit: Bourdier et al. 2014; Les Harpons: Ducasse and Renard dir. 2013, p. 57-63), the construction of AMS reference dates from reliable archaeological contexts was thus initiated. In this way, the Badegoulian industries from Cuzoul de Vers, representing one of the best documented series of débitage by percussion (Pétillon and Averbouh 2012), contribute to the development of chronological markers for this method. Initially, two 14C measurements were entrusted to the Oxford laboratory in the aim of publishing the results quickly (Pétillon and Ducasse 2012: table 5, p 158), but the dating program was then extended to seven additional measurements when funding was granted by the ARTEMIS program in 2011. Lastly, as part of the PCR SaM and ARTEMIS 2012, four remains from l. 29 to 31 were also dated, in order the evaluate the diachrony of the Solutrean occupations but also in order to obtain comparative “quality” data (cf. sampling strategy), as for the Badegoulian (Ducasse and Renard 2014). This extension thus provided the opportunity to combine a thematic question linked to antler Badegoulian material with the radiometric reevaluation of this reference sequence for the LGM in Western Europe.

Sampling strategies

24As the excavation data had not yet been computerized at that time, it was not possible to systematically project each dated remain. However, in several cases, the planimetric and altimetric proximity with specific lithic objects was observed, such as the percussion flake from l. 21, located right beside a fragment of a Badegoulian shouldered point (fig. 3, n° 5; J4 c.21c n° 1504: the stratigraphic section chosen in fig. 6A does not illustrate this spatial relationship). Nonetheless, the projection of all the pieces became possible at a later stage (fig. 6 and 7) and will be the subject of future archaeostratigraphic studies (cf. supra). The computerization of the excavation notebooks resulted in the correction of the stratigraphic origin of one of the dated pieces: the cortical splinter selected from l. 23 (according to its marking ; fig. 3, n° 4), turned out to belong to l. 22 (according to excavation notebook data). Although in theory, this reattribution only has a slight impact on the dating strategy (both layers being attributed to the Early Badegoulian), note that l. 22 had been deliberately excluded from the selection process, given its intermediate position between the Early and Recent Badegoulian, but also on account of the presence of several raclettes which are clearly spatially associated with the concentration of raclettes recorded in l. 20 and 21 (fig. 2; Ducasse 2010, p. 230). We will see that these elements of discussion shed light on the possible interpretation of the 14C results obtained from this sample (cf. infra).

26Thus, given the extreme rarity of the Solutrean bone industry, which is, moreover, not very characteristic (Le Guillou 2012; Pétillon and Averbouh 2012), our choice focused in this case on the faunal remains from each of the three layers. This selection was made at the beginning of the year 2013, and thus, unlike the Badegoulian levels, was accompanied by the stratigraphic projection of each selected object, in order to test the hypothesis that these three layers identified during fieldwork may in fact make up a single and coherent layer (Renard 2012) with specific “activity zones” to the east and west of the site. This hypothesis is based on typological and techno-economic observations, on the interpretation of the different refits and associations (cf. fig. 2), but also on “complementary” spatial distribution (Fourment and Giraud 2012, p.409)but apparently contradicted by 14C data (cf. supra). Four coordinated bone remains were thus selected for dating with the collaboration of Jean-Christophe Castel (fig. 4A), taking account of their position and spatial distribution in relation to each layer (fig. 4B). A reindeer metatarsal (fig. 4A, n° 1) was selected in layer 29 – a layer extending mainly over the east of the rock shelter -, and two remains (n° 2 and 3) were taken from layer 30, which corresponds to two disjointed concentrations correlated to two hearths, one to the west of the rock shelter, and the other to the east. The two selected pieces come from each of these concentrations (fig. 4B; a reindeer tibia from l. 30-east and a medium-sized ungulate shaft fragment from l. 30-west). The fourth and last object concerned (n° 4: medium-sized ungulate shaft) comes from the base of l. 31 located to the east of the rock shelter and interpreted as the extension of the hearth present in this zone in l. 30 (cf. note 6).

Sampling methods

27The pieces selected from the three Solutrean layers do not present any major constraints in terms of conservation (i.e., they are not very “sensitive” faunal remains, samples taken directly by the Lyon laboratory), but this is not the case for the Badegoulian elements for which it was important to limit sample destruction. Thus, although the status of these pieces (i.e., reindeer antler débitage waste) differentiates them from finished objects sometimes selected for dating (e. g. Aujoulat et al. 1998; Barandiaran 1988; Chauvière et al. 2006; Pétillon et al. in press; Szmidt et al. 2009; Tisnerat-Laborde, Valladas, Ladier 1997), the sampling methods applied sought to preserve, as much as possible, the morphological, dimensional and technological condition of each remain (fig. 3). This sine qua non condition is respected in order to technologically reexamine the pieces in the future and to critically assess the results and interpretations presented here (even to apply additional radiometric analyses to the same pieces). A specific sampling protocol was thus elaborated before submitting the samples to the laboratory, directly inspired by the procedure developed by Laurent Brou (Brou 2006; Ducasse and Renard dir. 2013, p. 32-34 with several adaptations). After recording the morphometric and technological data and taking photos, each piece was subject to the same series of operations: (1) selection of the best preserved zone (i.e., good thickness of compact tissue, non-fissured zone in order to avoid the fragmentation of the object, absence of visible technical marks); (2) cleaning of the selected zone by abrasion of the outer layer of the object, potentially contaminated (tool used: wood drill with a 3 mm bit mounted on a mini drill); (3) removal of the material to date by successive micro-drilling (e.g., fig. 3, n° 6) in order to obtain several solid samples with a total mass varying between 450 and 1040 mg depending on the pieces (tools used: drill bit with a diameter of 7 mm, precision scale). These three stages were repeated for each piece, the drill and drill bits were cleaned by ultrasound immersion between each series of samples in order to avoid any possible inter-sample pollution. The seven samples obtained in this way were stored in sterile Eppendorf tubes and then submitted to the dating laboratories.

Results and discussion

29Both these measurements, and the new date obtained for l. 29 (19410 ± 100 BP), are more or less identical to the results obtained for the same level by the Oxford laboratory in 2001 (tab. 1, n° 8: 18730 ± 110 BP, n° 19: 19510 ± 110 BP).

30This first element provides a guarantee of the excellent level of comparability and rules out the “laboratory” factor from any discussion of this new series of dates. These dates can thus be integrated with the previously available AMS corpus.

Confirmation and refinement of the AMS framework: The oldest west European Badegoulian

A tighter chronology for the Solutrean levels

32The four dates obtained for the Upper Solutrean (tab. 2) provide important – if not decisive – elements of discussion for this level. They display marked consistency and tend to back up the hypothesis advanced following the techno-economic and taphonomic analysis of the lithic industries (cf. supra), of a short-chronology assemblage linked to a very limited number of occupations (a single and unique occupation?). Three of these dates, from all of the three levels, overlap perfectly (fig. 5A in red), and also echo the AMS date obtained earlier for l. 29 and considered to be the more coherent of the two, after inter-site comparison. Note also, that unlike for the Badegoulian where classical dates and AMS dates are generally discordant (cf. supra), these new dates are perfectly coherent with the classical 14C measurement obtained for l. 30 (tab. 1: 19400 ± 210 BP) although this measurement was made on a bulk sample. It is tempting in this case to see this as another sign of the chronological homogeneity of the 29-30-31 levels, which seems to slot between 23 and 23.5 ka cal. BP. The only discordant note is the measurement obtained for l. 30, which yielded a much younger age (19050 ± 100 BP; fig. 7C), and although this date fits well with other dates from regional and extra-regional sites (Hinguant and Biard 2013; Bodu, Dumarçay, Naton 2014; Ducasse and Renard 2014), it is not in keeping with the stratigraphic position of the dated object (fig. 7B; base of the hollow BC3-4: Fourment and Giraud 2012, p. 411). Considering (1) the nature of the dated material, distinct from the three others (tab. 2: burnt bone) and (2) the – partial –insertion of this date in the group formed by the first three once the measurement was calibrated (fig. 5A), the hypothesis of a more recent Solutrean occupation or of a contamination of this level by remains of Badegoulian age must be considered with caution in the absence of concrete evidence.

33Lastly, note that the date obtained for l. 31 (19320 ± 100 BP) for the sample taken from the base of the hearth in J4 (1) invalidates the AMS date for l. 31 (20110 ± 180 BP, GifA 101440), attributed to the Solutrean level, and (2) is perfectly in keeping with the date for the sample selected from the overlying l. 30 in I4 (19380 ± 100 BP). Although we cannot explain the early age of the date obtained for l. 31 (cf. supra) in 2001 by the Gif-sur-Yvette laboratory, note that this new series of measurements tends to further isolate this archaeologically incoherent result.

Diachronic variations of the material and 14C chronology: what is the degree of resolution?

34We have seen that this tighter chronology for the Cuzoul de Vers sequence involves more or less significant changes in the lithic and osseous material (Clottes, Giraud, Chalard dir. 2012; Ducasse and Renard 2012).

35These changes are both typological and techno-economic and are manifest at an intracultural (Early Badegoulian/Recent Badegoulien transition and diachronic variation of the material during the course of the Recent Badegoulian), as well as an intercultural scale (Solutrean/Badegoulian transition). As we have seen, the aim of the renewed radiometric framework is also to obtain 14C measurements to better document the duration of the different occupation phases as well as their respective limits.

36Given these objectives and the critical analysis of all the available AMS dates for Cuzoul de Vers (cf. supra), we decided to exclude the following elements from the ensuing discussions:

1. the dates obtained from bulk bone samples (l. 6 and 11: tab. 1, n° 9 and 11) which involve the same sampling methods as the classical dates and which thus present the same biases (possible “average” dates);

2. the measurements that were reliable from a physico-chemical viewpoint, but that gave incoherent, difficult to interpret results (l. 11, 16 and 31: tab. 1, n° 10,13 and 20; l. 15 from the 2011-2012 series: tab. 2);

3. the measurements with a standard deviation superior to ± 200 (± 310 for l. 25 and 26, ± 240 for l. 27: tab. 1, n° 16 to 18) which are thus not accurate enough (probabilities spread over about 1500 years cal BP to 2 sigma), and are of no use to the questions raised here.

37Fifteen measurements were thus retained out of the 24 AMS dates carried out between 2001 and 2013 for all the levels. Seven dates correspond to the Recent Badegoulian (five layers were dated between l.6 and l.21), three dates to the Early Badegoulian (betweeen l.22 and l.27) and five dates to the Upper Solutrean (l.29 to 31), so that in theory, the limits of each of these subdivisions have a radiometric marker. Once the two measurements obtained for the same remain from l. 6 are combined (fig. 8: 18640 ± 71 BP; tool “R_Combine” from the OxCal software [version 4.2.3: Bronk Ramsey 2013], used before calibration), the calibration of the 14 dates retained (tab. 3) gives an interval of 23 to 22.3 ka cal. BP for the Recent Badegoulian, whereas the Early Badegoulian extends between 23.5 and 22.5 ka cal. BP. As shown in figure 9, this marked overlap between the Early and Recent Badegoulian is mainly due to the measurement obtained for l. 22a, which is identical to the date for l. 21. Yet, given (1) the contextual elements presented above for l. 22 and (2) the stratigraphic projection of the dated element which shows its spatial proximity with the first level with raclettes (c.21 : fig. 6B), the hypothesis that this measurement dates the base of the Recent Badegoulian appears plausible. Thus, at Cuzoul de Vers, both of the 14C markers likely to correspond to the industries attributed to the Early Badegoulian place this episode between 23.5 and 23 ka cal. BP. Although the overlap margin between the two Badegoulian phases is considerably reduced by this hypothesis, the Solutrean-Badegoulian transition remains rather nebulous (Oberlin and Valladas 2012). This transition was spared by potential 14C plateaus since the publication of the IntCal13 curve (Reimer et al. 2013), but is nonetheless characterized by an incompressible radiometric overlap which includes the whole of the Early Badegoulian. If there was an occupation hiatus between the last Solutreans and the first Badegoulians in Aquitaine (e.g., Bosselin and Djindjian 1997; Djindjian 2000; a hypothesis largely based on a corpus of non-calibrated dates obtained by the classical method), then the 14C method would not appear to be the most effective tool for bringing to light such a phenomenon, if it existed. On the contrary, in spite of the imprecision of the results, these data tend to show not only the early age, but also the relative rapidity of the reconfiguration of techno-economic standards. Finally we note that the Early Badegoulian at Cuzoul de Vers does not seem to last after 23 ka cal. BP, a period when Recent Badegoulian industries with raclettes develop in France (a level that lasts until about 21 ka cal. BP: Langlais et al. 2010; Ducasse 2012; Pétillon and Ducasse 2012).

38This chronological outline thus confirms that it is essential to refine the meaning of the stratigraphic dilatation of this sequence and the succession of the different sedimentary levels (cf. supra). Thus the narrowing of the 14C measurements points to a much shorter occupation phase of the rock shelter than previously thought, whereas the different physical links established between some of the layers (fig. 2) allow for the individualization of coherent and significant assemblages from a typological and techno-economic point of view. Nonetheless, the most striking observation remains the extreme similarity between the measurements obtained for the whole of the Recent Badegoulian (from l. 6 to l. 21: fig. 9). Whereas the proximity of the ages obtained for l. 19 to 21 is in keeping with the typo-technological and taphonomic similarities for l. 16-21 (fig. 2: presence of raclettes, shouldered points with abrupt retouch and virtual absence of antler points; lithic refits and associations; Ducasse 2010, p. 30-33, 78 and Ducasse in preparation), their equivalence with the available measurements for the assemblages documented between l.s 6 and 15 was less predictable, given the differences in hunting equipment (virtual disappearance of lithic arrow heads and development of antler points: fig. 2; Ducasse and Renard 2012, figure 4; Le Guillou 2012). If this radiometric equivalence does not result from confusion and/or difficult to identify stratigraphic disturbances in the absence of systematic refits, the contribution of radiometric data is limited, apart from the fact that they point to the transience of certain occupation episodes and the rapidity of some of these transitions. At an inter-site scale, this predictable distortion of time between the relative chronology and the 14C chronology probably contributes to maintaining the image of the synchronic diversity of Badegoulian weapons and reminds us, if need be, that this synchrony is often just a “short diachrony” (Boëda 2005).

An approach to be extended

39It is important not to grant an excessive role to absolute dates, or to forget their limits and shortcomings, but nonetheless, we must acknowledge the importance of these dates in debates focusing on the evolutionary dynamics of Palaeolithic societies, in the long term (rhythm of evolution, palaeo-climatic and palaeo-environmental correlations) or throughout space (regional divergences of these evolutions, development of cultural “mosaics”, etc.). From this point of view, the Solutrean-Badgeloulian “transition” appears to be a particularly fertile field of study. This transition is considered as a period of profound reconfiguration of techno and socio-economic norms (Ducasse 2012; Ducasse and Renard 2012), but its rhythms are still poorly known, thereby limiting the potential understanding of the exact causes and hindering the assessment of its possible diversity throughout south-western Europe. The very varied state of the archaeological documentation considerably limits the scope of the different hypotheses and calls for a critical review of the data, zone by zone.