INTRODUCTION
The Cosquer Cave near Marseilles was discovered in 1991 by a scuba diver (Clottes et al. Reference Clottes, Beltran, Courtin and Cosquer1992a). Its entrance is now 40 m below sea level, but at the height of the last glaciation, the level of the Mediterranean Sea was 135 m lower than today and the entrance of the cavity was more than 6 km from the sea. In this area, the coast was lined by islands that surrounded a steppe plateau closed on three sides, which was favorable to large herbivores and their hunting. The food resources of the three habitats (sea, plains, and mountains) made this place particularly attractive for prehistoric people. Access to the cave entrance was closed by the postglacial marine transgression about 10,000 yr ago and only a fifth of the cavity surfaces accessible to Upper Paleolithic people have survived; the rest is underwater (Figure 1). The preserved part of the cave is richly decorated with rock paintings, drawings, and carvings scattered in all areas. These representations attributed to the Paleolithic period consist of 194 animal figures, 69 hand stencils, 240 geometric signs, and 2 human shapes. Cosquer is one of the most ornate caves from the Franco-Cantabrian group (Clottes et al. Reference Clottes, Courtin, Valladas, Cachier, Mercier and Arnold1992b, Reference Clottes, Courtin, Collina-Girard, Arnold and Valladas1997). The drawings were made with red or black pigments, the latter color being obtained from wood charcoal (Pinus sylvestris) that can be dated by the radiocarbon method. Engraved figures are also abundant on the wall surface, which is rather soft and smooth because it is mostly composed of moonmilk formation resulting from the alteration of the limestone. Observation of the wall reveals the presence of many hollows and deep scrapings, suggesting that this creamy sediment was collected by prehistoric people (Clottes et al. Reference Clottes, Courtin and Vanrell2005a).
Figure 1 Map of Cosquer Cave with the location of the dated samples
PREVIOUS RESULTS
From 1992 to 1998, J Courtin (and J Clottes) sampled four ground charcoal specimens and charcoal splinters on nine animal representations, three hand stencils, and two undetermined marks for accelerator mass spectrometry (AMS) 14C dating (Arnold et al. Reference Arnold, Bard, Maurice and Duplessy1987; Clottes and Courtin Reference Clottes and Courtin1994). The parietal samples consisted of three horses (CHV001, 005, and 007), two bison (BIS001 and 002), a feline (FEL001), a stag (CER001), a megaceros (MEG001), and a jellyfish shape, three hand stencils (MNR007, MNN12, and 19), and two geometric signs (SIG100 and 121). There is no picture of the sampling being conducted, but the exact location of the samples is known for most of them (Table 1, column 4). These first 14C results (Figure 2) obtained on the Tandétron (LSCE, Gif-sur-Yvette) suggested that the cave was decorated during at least two main phases about 10,000 yr apart (Clottes et al. Reference Clottes, Courtin, Valladas, Cachier, Mercier and Arnold1992b, Reference Clottes, Courtin, Collina-Girard, Arnold and Valladas1997; Valladas et al. Reference Valladas, Tisnerat-Laborde, Cachier, Arnold, Bernaldo de Quiros, Cabrera-Valdes, Clottes, Courtin, Fortea-Perez, Gonzales-Sainz and Moure-Romanillo2001): two hand stencils (MNR007 and MNN19), a bison (BIS002), and the oval mark S100 were placed during the first phase between 33,000 and 31,000 cal BP (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013), while six animals (CHV001, 007, BIS002, CER001, MEG001, FEL001) and the star-like sign (S121) were dated to the second phase, between 25,000 and 21,000 cal BP. It is noteworthy that these bison, BIS001 and BIS002, which were dated respectively to the second and first phases of decoration, are laid one beside the other and display the same drawing conventions. Such a situation could be explained in two ways: either the conventions persisted through ~10,000 yr or BIS002 is contemporary with BIS001 and was drawn with charcoal left by people who came into the cave during the first phase (Clottes et al. Reference Clottes, Courtin, Collina-Girard, Arnold and Valladas1997). Finally, one hand (MNN12) and a horse (CHV005) were dated to 29,000–28,000 cal BP. These latter results suggest the eventual existence of an intermediate period of decoration of the cave between the two main phases, although we could not exclude the possibility that these two ages had been artificially underestimated due to contamination of the sample by modern carbon. The reliability of the result obtained on the jellyfish sign dated at ~16,000 cal BP remains questionable considering the difficulty of dating that small amount (0.2 mg) of carbon 20 yr ago. The charcoal fragments collected on the ground surface fell within the two main occupation periods (Clottes et al. Reference Clottes, Beltran, Courtin and Cosquer1992a, Reference Clottes, Courtin, Valladas, Cachier, Mercier and Arnold1992b; Clottes et Courtin 1994).
Figure 2 Calibrated 14C dates obtained from 1992 to 1998 using OxCal v 4.2 and the IntCal13 calibration data (Bronk Ramsey Reference Bronk Ramsey2009; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013; at 2-standard deviation confidence level).
Table 1 Information on the charcoal samples and 14C analysis data. The acid humic fractions HAF are given in italics. The symbols *, °°, and ** respectively designate the following references: Clottes et al. (1992, 1997) and Vanrell and Olive (2012). The calibrated ages are given in the rightmost column.
The New Sampling
In order to obtain more information on the human frequentation of the cave, the Ministry of Culture and Communication (DRAC Provence-Alpes-Côte d’Azur SRA) decided in 2011 to collect 23 new samples to be dated by 14C on the Artemis AMS (LMC14, CEA, Saclay). To the extent possible, the pigment was taken in a restricted area of the representation, which was carefully described, while the full process was photographed, including the exact location of each of the samples (Vanrell and Olive Reference Vanrell and Olive2012). The samples were collected on 12 animal drawings: four horses (CHV001, 005, 017 and 057), two jellyfish shapes (1 and 2), four bison (BIS001, 002, 004, and 005), one penguin (PIN003), and an animal shape (AIN015). Four of these samples were collected on representations (CHV001 and 005, BIS001 and 002) studied during the first dating program. Two new hand stencils (MNN001 and 009) and four other parietal marks were also sampled, as well as five charcoal specimens collected in fireplaces present on the ground surface or at the foot of the ornate wall. At the same time, two fragments of calcite (Prv1105-27 and Prv1105-30) of the flowstone deposited on the cave soil were dated by the U/Th method. These dates aimed to establish if this carbonate layer was deposited between the Paleolithic human occupations as previously suggested (Clottes et al. Reference Clottes, Courtin and Vanrell2005b; Collina-Girard and Arfib Reference Collina-Girard and Arfib2010) or much later during the Holocene interglacial period. One of the calcite samples (Prv1105-30) was situated just above the dated charcoal (Prv1105-19).
MATERIAL AND METHODS
The parietal samples (or the ground charcoal) to be dated consist of splinters of charcoal mixed with calcite (or moonmilk) grains from the limestone wall. After removing with pliers the calcareous minerals as much as possible, the chemical pretreatment of charcoal varied in intensity according to the sample size (Valladas et al. Reference Valladas, Tisnerat, Cachier and Arnold1999, Reference Valladas, Tisnerat-Laborde, Cachier, Arnold, Bernaldo de Quiros, Cabrera-Valdes, Clottes, Courtin, Fortea-Perez, Gonzales-Sainz and Moure-Romanillo2001). The pretreatment involves a succession of 0.5N hydrochloric acid-base-acid (ABA) treatments, which first dissolve the remaining carbonate grains, the fulvic acids arising from the transformation of organic matter, and bacteria or other living microorganisms. The basic treatment (sodium hydroxide), gentle at first, is increased in intensity according to the fragility of the sample. As a rule, the treatment stops when the solution becomes highly colored. The coloration suggests that not only have the outer grain layers been stripped, but that a good fraction of the original charcoal has passed into the solution. If the treatment is not interrupted in time, no charcoal may be left for dating. The remaining charcoal grains are washed again with aqueous HCl (0.5N). A Fourier transform infrared (FTIR) analysis was not performed on the charcoal after this treatment. Taking into account the efficiency and the strength of the chemical attack, which usually eliminates more than 95% of the original sample, we are assuming that no residual carbonate remains in the dated charcoals. Before the combustion, the small pieces of charcoal are carefully examined under a microscope to check their integrity (absence of any visible contaminant).
Whatever remains is oxidized to carbon dioxide, then reduced to graphite and compressed into pellets for the Artemis accelerator (3MV Pelletron accelerator; Cottereau et al. Reference Cottereau, Arnold, Moreau, Baqué, Bavay, Caffy, Comby, Dumoulin, Hain, Perron, Salomon and Setti2007). These pellets usually contain somewhere between <0.5 mg to approximately 1 mg of carbon. During the processing of the Cosquer samples, blank values, which take into account the sample’s chemical pretreatment, the conversion into CO2, the graphitization, and the machine background contaminants, were estimated by measuring charcoal specimens from a Middle Stone Age layer (Border Cave, South Africa). The δ13C values of all samples were measured during the AMS analysis for the purpose of fractionation correction.
Two samples (Prv1105-3 and 19, collected respectively on CHV005 and on the ancient patina fireplace) were big enough to be divided into two parts, which were successively pretreated and 14C dated to test the reproducibility of the results. The humic acid fraction (HAF, in italics in Table 1) of six samples resulting from the basic treatment was also dated to evaluate the contamination level of the samples. This fraction contains a large part of the dissolved charcoal and the possible contamination by extraneous carbon. Therefore, the comparison of ages obtained on the humic acid fractions (HAF) to those of the purified charcoal specimens provides information on the sample’s quality and contamination (Batten et al. Reference Batten, Gillespie, Gowlett and Hedges1986; Valladas et al. Reference Valladas, Tisnerat-Laborde, Cachier, Arnold, Bernaldo de Quiros, Cabrera-Valdes, Clottes, Courtin, Fortea-Perez, Gonzales-Sainz and Moure-Romanillo2001). The two flowstone samples (~250 mg per sample) were prepared and analyzed at the LSCE on the Neptune Plus Plasma multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS) following the procedure described by Pons-Branchu et al. (Reference Pons-Branchu, Douville, Roy-Barman, Dumont, Branchu, Thil, Frank, Bordier and Borst2014) (see supplementary material 1).
RESULTS AND DISCUSSION
Information on the samples and their analysis is given in Table 1. Most of the 14C ages range from 27,000 to 16,000 yr BP (~32,000 to 19,000 cal BP) and fall within the same time period as the previous results. Surprisingly, the charcoal specimen found on the Crystal Gallery ground gives a much later age, ~4000 yr BP. At that time, Cosquer Cave was only accessible by diving and the only possible explanations for this recent age are that the charcoal was severely contaminated by modern carbon or was brought into the cave by sea currents during the Holocene period.
The duplicate results obtained on the 2011 samples collected in the ancient patina fireplace (Prv1105-19: 27,290±300 and 27,560±310 BP) and on the horse CHV005 (Prv1105-03: 22,440±130 and 22,920±160 BP) are in agreement. However, for this latter sample, the HAF gives an age about 1500 yr older (24,340±170 BP) than the charcoal duplicate ages. In general, we have found older dates for a given drawing sample to be more reliable after noting how much more frequent was contamination by recent carbon and consequent age reduction. Exposed pigments can be contaminated by organic materials, some of which can resist the chemical treatment meant to eliminate them. Some samples are so fragile that if the solid component is not to dissolve completely the purification has to be less rigorous. In such cases, the HAF, which consists of parts of original charcoal that were dissolved in the alkaline solution and reprecipitated, will give a more correct age. Therefore, for the horse drawing (Prv1105-03: CHV005), the oldest date obtained on the HAF (24,340±170 BP) should be the most relevant, and is also in agreement with the date (24,730±300 yr BP) obtained on the sample collected on the same representation in 1996 (Clottes et al. Reference Clottes, Courtin, Collina-Girard, Arnold and Valladas1997). A similar situation was encountered with the sample collected in the recent patina fireplace (Prv1105-14: charcoal and humic acid fraction dated respectively to 15,730±80 and 17,950±380 BP) and to a less extent with the hand stencil MNN001 (Prv1105-24: charcoal and HAF dated respectively to 26,900±290 and 28,060±550 BP). In our opinion, the oldest dates obtained on the HAF are the most reliable.
For the ancient patina fireplace (Prv1105-14) and the fixed lamp (Prv1105-18), the HAF ages are compatible with those obtained on the associated charcoals: for the Prv1105-14, the charcoal is dated to 27,290±300 BP and 27,560±310 BP and the HAF at 27,090±300 BP; for Prv1105-18, the charcoal and HAF are dated respectively to 27,020±290 and 27,990±520 BP. While this good agreement between the two sets of dates generally increases one’s confidence in the reliability of the dates, we can never exclude a remote possibility that both fractions may have been contaminated somehow. In the case of the horse CHV001 (Prv1105-01), the HAF (22,860±330 BP) is ~1500 yr BP younger than the charcoal (25,450±190 BP), suggesting the presence of modern carbon contamination in the pigment specimen. In this case, the most trustworthy age is the one of the purified charcoal.
Finally, we compare the sets of ages obtained in 1992–1996 and in 2012 on the four parietal representations sampled twice (CHV001 and 005, BIS001 and 002). The ages of the bison BIS002 are in good agreement. Those obtained on the horse CHV005 and the bison BIS001 sampled in 2011 are 1000 to 1500 yr younger than the previous results, and we suspect the presence of a remaining slight contamination in the second set of samples. The case of the horse CHV001 is more surprising: the 2010 date (25,450±190 BP) is much older than those obtained in 1992 on the same representation (18,840±250 and 18,820±310 BP). There is no straightforward explanation for this age difference, but several assumptions could be proposed as the two dated samples were not taken in the same part of the drawing: reutilization of old charcoal left on the soil surface (Bednarik Reference Bednarik1994) or a later repainting.
All the calibrated 14C dates obtained using the IntCal13 curve (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013) and the OxCal v 4.2 software (Bronk Ramsey Reference Bronk Ramsey2009) are reported in Figure 3. The 2010 results fall in the same time interval (33,000 and 20,000 cal BP) as the previous ones, but they show that prehistoric people went into the cave and made drawings not only during the two phases (33,000–30,000 and 25,000–21,000 cal BP, respectively) initially highlighted. The new dates confirm the existence of an intermediate period of decoration (~25,000 BP; 30,000–27,000 cal BP) between these two main phases. The horse CHV005, the animal shape (AIN015), the parietal marks (SIG125, 119), and the hand stencil MNN012 can be attributed to this intermediate period.
Figure 3 Calibrated 14C dates obtained from 1992 to 2012 using OxCal v 4.2 and the IntCal13 calibration data (Bronk Ramsey Reference Bronk Ramsey2009; Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht2013; at 2-standard deviation confidence level).
The ages obtained on the bison and the horse drawings are scattered from 32,000 to 22,000 cal BP; thus, the horses CHV017 and 007 and the bison BIS001 and 004 were dated between 25,000 and 22,000 cal BP, while the horse CHV005 and the bison BIS002 were placed between 32,000 and 28,000 cal BP. These results suggest that the stylistic conventions have persisted through several millennia, even if we cannot definitively exclude the reuse of old charcoal collected on the ground to realize the drawings by prehistoric people.
The ages obtained on the two calcite samples, 8300±137 on Prv1105-30 (LSCE5695) and 4288±260 yr BP on Prv1105-27 (LCE5694), show that the flowstone grew during the Holocene period, more than 10,000 yr after the second Paleolithic human frequentation. This flowstone is present on a large part of the cave soil surface and contributed to the good preservation of the underlying archaeological layers (supplemenaty material 1).
CONCLUSION
Two dating programs involving 41 samples (18 from 1992 to 1998 and 23 in 2012) have been devoted to the decorated Cosquer Cave since its discovery. The charcoal samples consist of 21 animal representations (9 between 1992 and 1998 and 12 in 2012), five hand stencils (3 and 2), seven signs (2 and 5), and eight specimens found on the soil surface (4 and 4). A total of 57 14C analyses including duplicate measurements were realized. They produce a coherent set of data, which shows that the Cosquer Cave was visited by prehistoric people from 33,000 to 20,000 cal BP. Throughout this time period, animal representations as well as hand stencils and several marks were drawn on the wall of the cavity. The oldest decoration period of Cosquer Cave falls in the same time range as the Chauvet Cave’s latest occupation dated between 31,000 and 29,000 cal BP (Clottes Reference Clottes2001; Geneste Reference Geneste2005; Quiles et al. Reference Quiles, Valladas, Bocherens, Delqué-Količ, Kaltnecker, van der Plicht, Delannoy, Feruglio, Fritz, Monney, Philippe, Tosello, Clottes and Geneste2016).
ACKNOWLEDGMENTS
The dating program of Cosquer Cave was made in the context of a PNRCC project “Dating Modeling of 14C and U/Th method for Upper Palaeolithic Painted Caves,” funded by the French Ministry of Culture. We are grateful to Cyrielle Messager for her support. We would like to thank the reviewers for their careful reading of the article and their helpful comments. This is LSCE contribution n°5803.
