1- Introduction

1Gatzarria Cave is an important site for many reasons regarding the Middle to Upper Palaeolithic transition. Among them, it is one of the few sites to contain a Vasconian Mousterian industry, a facies often thought to represent a late stage in the Mousterian (Deschamps 2008) and thereby potentially associated with the late Neanderthals. It is also among the few sites to contain a stratified sequence of Proto-Aurignacian and Classic Aurignacian levels. Thus, it can help us to appreciate the chronological relationship between the Proto-Aurignacian and other Early Upper Palaeolithic or late Mousterian industries and contribute to evaluating the taxonomic link between them.

2The principal objective of this article is to present the first ever radiocarbon dates done on the Vasconian Mousterian, Proto-Aurignacian and Classic Aurignacian layers of this site. We undertook dating of animal bones from three levels, sampling from the Musée National de Préhistoire (MNP) collections.

3Dating is the focus, but this could only be achieved reliably at this site by integrally linking sampling for dating with a consideration of the larger contextual information of the site; this, in order to avoid obtaining incorrect dates. Thus, the historical, lithic, excavation methodology and stratigraphic aspects that are crucial for understanding what, why, where and how we sampled, are also presented.

4We begin by presenting the site and giving an overview of the assemblages relevant to this study and the excavation/recovery methods which were used by Laplace (section 2). We then present the work we have done to establish the chronology of this site. This includes details of the way in which we evaluated the excavation stratigraphic integrity and determined the most appropriate zone for sampling (section 3). After presenting the methodology of the dating used (section 4), we end the paper by placing our new AMS radiocarbon dating results (section 5) into the larger technological, chronological and human behavioural issues surrounding the late Mousterian and early Aurignacian periods (section 6). A detailed techno-economic re-evaluation of the industries of Gatzarria, a full re-evaluation of previous work and detailed comparisons with other industries, will constitute the subject of future articles focusing on those aspects.

2 - Site characteristics

2.1 - History and methodology of excavation

5Gatzarria cave is located in the Basque area of France, in the Ossas-Suhare region, which is part of the Arbailles Massif (fig. 1). The cave, which faces north-east, is located within an Urgonian limestone cliff at the foot of Mount Hargagne. The main part of this small cavity is a 13m long, 4m wide, corridor and its maximum height was 2m from the top of the deposits (fig. 2). The cave floor is sub-horizontal at the entrance, but rapidly angles downward towards the back of the cave (see implications for this below). It was discovered in 1945 by a group of spelunkers. A few test pits were dug by P. Boucher and G. Laplace until 1958, revealing Mousteroid and Aurignacian industries. More extensive excavations, amounting to a total of 37 m2 were then conducted between 1961 and 1976 by Laplace. The thickness of deposits excavated was variable, but never reached more than 2m. At no time (neither during test pitting, nor during excavation) was the bedrock reached.

6The excavation and recording methods used at Gatzarria were those developed by Méroc and subsequently developed by him and Laplace (Laplace 1971 ; Laplace and Méroc 1954). This method involved the three-dimensional plotting of artifacts according to an orthogonal coordinate system. Excavation was done in horizontal bands (‘décapages’) following the sedimentary units. Pieces were labelled according to their three-dimensional coordinates. Some of them (especially the smaller fraction and/or unmodified lithic artifacts, as well as most of the fauna) were not labelled to precise depth, but only to archaeological layer and square, sometimes to sub-square and décapage. In the case of Gatzarria, a décapage corresponded to a 10 cm-thick cut (known as ‘Taille’ = T in Laplace nomenclature) or 5 cm-thick half-cut (T sup ; T inf) and each square metre was sub-divided into 9 sub-squares.

2.2 - Stratigraphy of the deposits and description of archaeological assemblages

7As mentioned above, the thickness of the excavated deposits never exceeded 2 m. The stratigraphy is shown in figure 3 and table 1 (see Saenz de Buruaga 1991 for more details).

9In this article, we focus particularly on the three layers which we dated, Cjr (Vasconian Mousterian), Cjn2 (Proto-Aurignacian), and Cbf (Classic Aurignacian). The studied set from layer Cjr contains 1552 lithic pieces. Quartzite was the raw material most chosen (89%), followed by flint (5%) and lydite (3%). All other raw materials were rarely used. The chaîne opératoire most commonly seen for the production of flakes is the unifacial Discoid method, followed by the bifacial method (fig. 4). The main goal was to make flakes whose entire periphery could be used as cutting edges. Naturally backed elements are relatively well represented due to the fact that the striking surfaces of quartzite cores are, for the most part, neocortical. We also find flakes whose back was formed by previous removals. It is interesting to note that there are very few pseudo-Levallois points. The assemblage contains six ophite cleavers (fig. 5). This raw material was used nearly exclusively to make this artifact-type. Retouched flake tools consist primarily of side-scrapers, followed by notches and denticulates (fig. 6); all other tool types are infrequent.

10Analyses are still underway regarding the Proto-Aurignacian layers, but preliminary results can already be presented. Archaeological layer Cjn2 (tabl. 2 and 3) fits completely within the definition and characterization of the Proto-Aurignacian. It is the richer of two superposed Proto-Aurignacian layers, containing 1737 flint pieces in the sampled set (see below for what is meant by ‘sampled set’). This industry is geared towards the exclusive production of small blades, as well as elongated, slender, rectilinear-profile bladelets produced as part of the same operational sequence. Blade-bladelet blanks thus dominate among the tools, representing 72.5 % of the transformed blanks (74 of 102, not including core tools such as carinated end-scrapers and burins). The rest of the tools were made on non-standardized blanks obtained from the shaping and maintenance phases of blade-bladelet production. Among the retouched blades and bladelets, Dufour bladelets (sub-type Dufour) are the most prevalent (37 out of 74 retouched blades/bladelets) (fig. 7 and 8). These are also the most standardized elements in the assemblage in terms of size and morphology, with thickness ranging from 1-2.5 mm and width from 3.5-9 mm. Most of them are broken, with the longest being 32.5 mm. As is typical for this artifact type, the bladelets are retouched on one edge of the ventral surface (inverse retouch), with some of them also being dorsally retouched on the opposite edge (alternate retouch). In this assemblage, the inverse retouch was nearly exclusively done on the right edge (34 of 37 cases), as is usually the case for this artifact type (Demars et Laurent 1989 - p. 102 ; Rigaud 1982 - p. 366). It is marginal (often very marginal) and semi-abrupt, thus barely transforming the blank. Blanks were made using local raw materials or generally those from the neighbouring environment ; flysch dominates, with 93.8 % of the pieces of identified raw material type belonging to this group (we could not identify the raw material of 20% percent of the assemblage).

11Although the assemblages from archaeological layers Cjn2 and Cjn1 both display Proto-Aurignacian characteristics, the Cjn1 assemblage is, in certain respects, more aligned with the Classic Aurignacian. It is for this reason that Laplace had described and viewed it as a Proto-Aurignacian transitioning towards the Classic Aurignacian. This is seen in the tool types represented. There are only 5 Dufour bladelets out of 50 tools on blades/bladelets and out of 92 tools in total (tabl. 4 and 5). Splintered pieces and carinated end-scrapers are well represented (28 % of the tools compared with 2.6 % in Cjn2) and retouch is much more frequently scaled and invasive, thus resembling Aurignacian retouch which is very rare or absent in Proto-Aurignacian assemblages. The raw material aspects are also noteworthy. An initial examination of the Classic Aurignacian assemblage of archaeological layer Cbf revealed that the types of flints used were overall the same as those used in Proto-Aurignacian assemblage Cjn2, but in reverse proportion. As such, Iholdy flysch flint, which accounts for 60 % of the raw material in Cjn2, is only marginally represented in Cbf. This rarity is also the case in Cjn1, where Iholdy flint represents only 8.5 % of the assemblage. It should be noted, though, that the assemblage from Cjn1 is smaller than Cjn2 and Cbci-Cbf (N = 1100) and that this ‘thin and sporadic layer, located in the upper part of the unit, is only occasionally seen as a thin line of hearths or, in other instances, as a horizon of minute traces of soot’ (Laplace 1966b - p. 125, our translation). Furthermore, it is frequently mentioned in the excavation notebooks (kept in the MNP archives) that it is very difficult or impossible to see the contact surface between Cbf and Cjn1 visually in the stratigraphy. Also, Cjn1 and Cjn2 are found in the same sedimentary unit (see below and Table 1) and are in contact with each other. Consequently, at this stage of our own analysis, we cannot exclude the hypothesis that the Cjn1 assemblage could be, at least in part, a mix of Proto-Aurignacian assemblage Cjn2 and Classic Aurignacian assemblage Cbf. For these reasons, as a precaution, we chose not to date Cjn1.

12The Classic Aurignacian assemblages of archaeological layers Cbci-cbf have not been re-analysed recently. Published findings (Laplace 1966a,b ; Saenz de Buruaga 1991), however, clearly show them to be attributable to the Classic Aurignacian. Readers are also invited to consult the recent synthesis of typological information carried out by Normand (2005) based on these publications and their drawings.

3 - Using the Laplace stratigraphy and nomenclature for new analyses

13In light of the risk of post-depositional disturbance, modern methods of study place much emphasis on ensuring the validity of archaeological layers. The likelihood of layers mixing is fairly high in karstic contexts and the extent of this can be difficult to estimate in excavations carried out a long time ago.

14This clearly poses a risk of error for any analysis of the material, although such an error might be noticeable in some cases if obviously out-of-place artifacts were present. It poses a serious and very likely risk of grave error, however, in the case of radiocarbon dating of the material in the collection because such stratigraphic problems would go unnoticed without refitting or similar post excavation analyses. Stratigraphically incoherent dates might be a sign of a stratigraphic problem, but a) this is late to discover that one has unnecessarily spent financial resources and b) dates can be incoherent for other reasons, such as contamination problems, having nothing to do with the stratigraphy.

15Below, we describe the analyses we did to assess the stratigraphy and to identify the most problematic zones at Gatzarria. Such work helped us identify the zones least affected by taphonomic excavation problems and thus the ones best suited for radiocarbon sampling. We then go on to describe the samples chosen for dating, the laboratory methodology used on them and the results.

3.1 - Assessment of Gatzarria Laplace excavation stratigraphic integrity

16Examining the stratigraphic profile of the infilling (fig. 2) can lead one to suspect that the deposits followed the topography of the site, and thus that they would be sub-horizontal at the front, dipping progressively towards the back of the cave. As a result, one would expect the layers at the back to be far more mixed. Our observations on the assemblages from Cjn2 and Cjr confirmed this hypothesis. Our analyses revealed that the material from the posterior portion of the cave was far less homogeneous than that of the anterior zone. We found this to be the case not only in terms of types of artifacts, but also in the representativeness of raw materials and in the surface condition of the pieces. For instance, we found a type of flint in the back portion of Cjn2 which is not present in the sub-horizontal front area of that layer but is, however, found higher up in the stratigraphy. These observations led us to consider only the material from the anterior zones of the cave for techno-typological analyses as well as for radiocarbon sampling. We refer to this material as the ‘sampled set’.

17In his 1962 site report, Laplace stated that “There was no sterile horizon between the various archaeological layers [of sedimentary unit CJ], although all are clearly individualized, with no indication or likely possibility of contamination between them” (Laplace 1962). Although he took care to identify sedimentary changes during excavation, as mentioned above he excavated in horizontal décapages. In addition, the four layers of CJ – the uppermost Mousterian layer (Cj), the Châtelperronian layer (Cjn3) and the two Proto-Aurignacian layers (Cjn2 and Cjn1) –― are all contained in the same sedimentary unit and were superposed without any sterile layers separating them (Laplace 1966b). As a result, “[CJ was divided] into horizons/archaeological levels using only Cartesian coordinates” (Laplace 1963). Given the sloping topography and the horizontal way in which the site was excavated, it is therefore to be expected that mixing will have occurred between these layers, at least in part.

18Thus, we used the method of vertically projecting artifacts to “correct or confirm, within the sequence of strata, the boundaries of cultural levels as established during the excavation” (Villa 1982 - p. 277). We found that the layers were indeed in contact with each other and that they overlapped partly. The analyses we conducted on Cjn2 and Cjr indicate that in the sub-horizontal zones of the cave, these layers are internally coherent in terms of techno-typology. As well, the surface condition is identical across artifacts, bolstering the argument of overall homogeneity of the anterior zone parts of these levels. According to us, the techno-typo-taphonomic homogeneity of the sampled assemblages is a strong argument in favour of their stratigraphic integrity, thus making sampling for dating feasible within this zone.

19In her faunal analysis of Mousterian archaeological layer Cj, Ready (2010) came to a similar conclusion regarding overlapping and mixing of layers, as well as problems in the layer attribution of artifacts, finding that this was especially problematic in certain parts of the cave. Although this was not a layer studied by us, her findings bolster our argument that care must be taken in choosing which artifacts to include in any analysis from Gatzarria, be it for subsistence-related analyses, as in her case, or for choosing samples for dating, as in ours.

20The stratigraphy at Gatzarria also includes a layer which was attributed to the Châtelperronian by Laplace (Cjn3). We did not have the chance to re-evaluate this assemblage given that others were currently working on it. As such, we did not take samples from it. Based on our existing research at the site, however, a number of factors lead us to suspect that this layer may be mixed with material from layers above and below it. For one thing, layer Cjn3 is very thin. As described above, our analyses have shown that post-depositional processes were acting on this site, so a thin layer is problematic. In addition, Cjn3 is part of the same sedimentary unit as layers above and below it and is thus indistinguishable from these. This hypothesis of mixing needs to be tested more extensively on the assemblage itself, but existing information already made this layer questionable and problematic for dating.

3.2 - Selecting the sampling zone

21The zone from which to sample was thus chosen in light of the various issues outlined above. The projection of pieces also shows that the slumping towards the back of the cave is strong from row 5, but is already happening to some extent in row 4 (see fig. 2). Thus, sampling from rows 4 and above was not considered reliable by us. The material from the front of the excavation (rows 22 and 21) were less precisely extracted and labelled ; row 1, also near the front of the cave, is poor in material. As a result, sampling was avoided from these front rows. This left us with rows 2 and 3. Given that the latter has more material and has a more dilated stratigraphy (easier to avoid contact surface with other layers) our preference was for row 3. In the North-South direction, rows E and F are the richest in material. Taking all this into account, samples were examined for selection from preferred squares 3E and 3F, but also 2E and 2F.

22It is also worth noting that the study of the Proto-Aurignacian layers included some non systematic refitting and conjoining work (by LE). The number of pieces is very small (N=23) and thus not sufficient on its own, but combined with other evidence described, it confirms our conclusions. Work of this type, done for the purpose of stratigraphic assessment, has been shown to be particularly useful in karstic contexts. It consists of making projections of such artifacts within and between layers to assess the degree of dispersion within the deposits and thus test the integrity and amount of post-depositional disturbance of the stratigraphy (Bordes 2002 ; Cahen and Moeyersons 1977 ; Cziesla et al. 1990 ; Hofman 1986 ; Tixier 1978 ; Villa 1982 ; Villa and Courtin 1983). The results of the initial refitting and conjoining test carried out at Gatzarria (fig. 9) demonstrate no significant vertical displacements in the zone under consideration. The only inter-level refit found (refit # 5 in Figure 9), which unites two pieces marked as coming from Cjn1 and Cjn2, is consistent with this finding. The stratigraphically higher one was found at the contact zone between the two layers and could therefore have been part of Cjn2; the one found lower was aligned with the North-South slope of the cave. In future we hope to expand this study. For now, this test bolsters the techno-typological findings of the lithics from this zone, demonstrating minimal risk of contamination between layers in the sampled zone.

4 - Dating methodology

4.1 – Sampling

23Our radiocarbon dating at Gatzarria relates to a larger Wenner-Gren Foundation-funded chronometric project, directed by one of us (CCBS), to gain a more precise and accurate handle on the late Mousterian and early Upper Palaeolithic chronology, particularly in the larger southern France region (Mediterranean, Pyrenean and Southwestern regions) (Szmidt 2009 ; Szmidt et al. 2010a,b,c). The dating sampling protocol of this larger chronometric project has been described already (Szmidt et al. 2010c). Readers are invited to consult these for details. We have applied this protocol as much as possible here, while operating within the well-known constraints imposed by selecting from a previously excavated site, whose labelling and recording protocol and standards differ from the ideals of today. The fairly detailed 3D positioning and labelling used by Laplace made this far more feasible than would be possible for other sites. The willingness and openness of the curators of the museum to grant us access to all the material for the purposes of looking for potential samples, also helped enormously. Curatorial issues, including preservation, often make it difficult for museums to allow extensive access and/or sampling of artifacts.

24As per the sampling protocol of this overall regional dating project, for Gatzarria we wanted samples which were as precisely provenienced as possible (preferably z,x,y coordinates, but at least x,y and décapage) ; that could be identified taxonomically (preferably to species, but at least enough to rule out bones of carnivores) ; preferably thick long bones in good condition ; and preferably bones with direct evidence of human involvement (cutmarks or impact scars).

25Two samples from each of the three layers (Cbf, Cjn2, Cjr) were chosen (see tabl. 6 for provenience and taxonomic details). Layer Cjn2 is located between 60 cm and 90 cm depth in row 3. In order to minimize the potential problem of mixing between it and adjoining layers Cjn3 (below) and Cjn1 (above) at their contact surfaces, we sampled from the middle of the layer. Thus, when choosing samples for dating from Cjn2, we only considered bones bearing precise depth coordinates to make this feasible (as opposed to only décapage information). As shown in Table 6, the two dated samples were found at 69 cm and 78 cm depths (shown as squares in all the graphs of Figure 9).

26Regarding the Vasconian Mousterian layer (Cjr), the analysis by Deschamps (2008) found that the back of the cave was stratigraphically mixed, as was the case for the Aurignacian levels. As a result, we sampled for dating from the middle section, in particular the square which had been most studied in this recent technological evaluation (square 3F). This square showed a concentration of cleavers, the typologically diagnostic artifact of this industry, at a depth of 140 cm. T15 sup (T15 upper) is the Laplace half-cut décapage corresponding to a depth of 140-144 cm, so we selected bone samples for radiocarbon dating from within that half-cut décapage. In particular, we were especially keen to choose cutmarked identified bones so as to have as clear a link as possible to human activity. The use of caves by carnivores and bears in the Mousterian period is well known, so it was particularly important to avoid these species for samples from this period.

4.2 - Radiocarbon AMS analysis

27The bones were sent to the Oxford Radiocarbon Accelerator Unit (ORAU) for all stages of the radiocarbon analysis by accelerator mass spectrometry (AMS). Pre-treatment, combustion, graphitization and measurement were done according to the current standard procedures for bones used in this laboratory. This included an ultrafiltration step which the ORAU deems to be an improvement to the thoroughness of eliminating contaminants, based on much research and experiments in recent years. For ample details regarding their techniques and procedures, the reader is invited to consult the following publications (Brock et al. 2007, 2010 ; Bronk-Ramsey et al. 2002, 2004a,b ; Brown et al. 1988 ; Higham et al. 2006 ; Jacobi et al. 2006 ; Wood et al. 2010). Eliminating contaminants is especially important in samples of Aurignacian/Mousterian age, where even a trace amount could significantly rejuvenate a date (Aitken 1990). None of the six samples has ever been coated in preservatives or consolidants and casts had not been made. They were photographed before submission for dating.

28To summarize our approach to the dating of the Gatzarria collection, we did numerous studies before sampling, as outlined more amply above. Firstly, we carried out a complete techno-typological, raw material and surface condition study of the lithics from Cjn2 and Cjr. This showed homogeneity at the front of the cave, but not at the back, so we focused on the front. Secondly, for the front zone, we projected the lithic pieces bearing three-dimensional coordinates. This confirmed that layers touched each other and overlapped slightly within sedimentary unit CJ. This also gave us general depth limits of layers, enabling us to select away from the contact surfaces between layers. Thirdly, we did refitting and conjoining work on the sampling zone being considered to check that there was no major vertical displacement there. Then, within these zone and depth constraints, we focused on selecting taxonomically identified bones in good condition, preferably with cutmarks. Furthermore, in terms of the laboratory aspects, we made sure that good pretreatment methods were applied to such Palaeolithic samples, including ultrafiltration, and we verified that chemical and elemental parameters were normal.

5 - Results

29Results are quoted in radiocarbon years before present (BP) using the Libby 14C half-life and have been corrected for natural isotopic fractionation. The collagen yield percentage, as well as the elemental and isotopic values (C:N, 13C, 15N) of all the samples are within normal parameters (tabl. 7) (Burleigh et al. 1984 ; DeNiro 1985 ; Hedges and van Klinken 1992 ; van Klinken 1999). The only exception is that the collagen yield percentage of one of the Mousterian samples (cjr 3F (9) T15 sup #1) is low.

30As can be seen, the dates are stratigraphically coherent, with some overlapping of dates. The two layer Cbf dates (Classic Aurignacian) are statistically identical to each other (overlap at one standard deviation). Within layer Cjn2 (Proto-Aurignacian), where we had precise depth measurements for the samples as opposed to simply square and décapage, the stratigraphically lower sample (cjn2 3E 78.25.70) is older than the stratigraphically higher sample (cjn2 3E 69.99.95). The Mousterian samples are also consistent, both being older than is possible to measure, being over 47 400 14C BP in age.

6 – Discussion

31In this section, we will place our new results in the larger context of issues of this period. We begin by reviewing the more salient points of debate of this period (section 6.1), then turn to the particular results obtained at Gatzarria to see the way in which they fit into these larger discussions (section 6.2).

6.1 - Background Middle to Upper Palaeolithic transition issues

6.1.1 - The early phases of the Aurignacian

32There has been much debate in recent years regarding the definition and characterization of the earliest Aurignacian industries, Proto-Aurignacian (Aurignacien archaïque) versus Classic Aurignacian (Early Aurignacian/Aurignacian I/Typical Aurignacian/Aurignacien ancien), and the techno-typological and chronological relationship between them. Until recently the Aurignacian (aside from its initial phases) was perceived by most scholars as a pan-European, homogeneous culture (for an overview see Bon 2002 ; Le Brun-Ricalens and Bordes 2007). In contrast, the Proto-Aurignacian seemed more impoverished in terms of material culture and was mostly confined to the southern European zones.

33The Proto-Aurignacian and Classic Aurignacian have been amply described elsewhere and the reader is invited to consult these works for detailed information (Arrizabalaga et al. 2009 ; Bazile 2005 ; Bazile and Sicard 1999 ; Bon 2002, 2006a and b ; Bon et al. eds. 2006 ; Bordes 2005, 2006 ; Bordes et al. 2008 ; Broglio 1996 ; Chazan 2001 ; Conard and Bolus 2006 ; Kuhn and Stiner 1998 ; Laplace 1966b ; Le Brun-Ricalens 1993, 2005 ; Le Brun-Ricalens et al. 2009 ; Mellars 2006 ; Normand 2002 ; Normand and Turq 2005; Onoratini 1986, 2004 ; Teyssandier 2008).

34In France, the Proto-Aurignacian is found primarily in the Mediterranean and Pyrenean regions and the number of chronometrically-dated ones is small. Even more rare are sites which contain both industries. In wanting to compare the Proto-Aurignacian and Classic Aurignacian in terms of technology and typology, in order to assess their taxonomic relationship, and in wanting to assess their chronological relationship, such multi-stratified sites are of particular interest.

35Based on evidence available until recently, it was usually thought that the Proto-Aurignacian and Classic Aurignacian were separated geographically within Europe (Mediterranean Europe versus Central Europe/Atlantic region, respectively ; cf. note 3). Given the evidence available in these different regions, the question of their taxonomic and chronological relationship thus emerged.

37Some more recent findings and studies, however, are making us re-evaluate these views. Regarding the chronological, geographic and/or taxonomic separation and independence of these two industries, a few cases have emerged of a techno-typologically Proto-Aurignacian lithic industry associated with split-based points (Brou 1997 ; Liolios 2006 ; Schmider ed. 2002 ; Schmider and Perpère 1997 ; Soler Subils et al. 2008 ; Szmidt et al. 2010a). Until now, the latter were instead exclusively associated with the Classic Aurignacian. As well, Proto-Aurignacian assemblages are being found outside their ‘expected’ geographic range. Recent technological analyses of previously excavated assemblages and new excavations have revealed Proto-Aurignacian assemblages in southwestern and northern parts of France (Le Piage : Bordes 2002; Bordes et al. 2008 ; Les Cottés : Talamo et al. 2012 ; Grotte du Renne, level VII : Bon and Bodu 2002 ; Trou de la Mère Clochette : Brou 1997 ; Szmidt et al. 2010a), as well as in central Europe (Kozarnika level VII : Guadelli et al. 2005 ; Sirakov et al. 2007 ; Tsanova 2008 ; Krems and Tincova : Teyssandier 2006, 2008 ; Zilhão 2006). These demonstrate a geographical range which overlaps and even goes beyond that of the Classic Aurignacian.

38Some of these Proto-Aurignacian assemblages have dates extending well into the range of the Classic Aurignacian (Szmidt et al. 2010a). In addition, there is evidence at some sites for in situ development of a Classic Aurignacian industry from Proto-Aurignacian roots, such as at Isturitz cave and possibly at other neighbouring Pyrenean sites (Arrizabalaga et al. 2009 ; Normand 2002 ; Normand et al. 2007 ; Szmidt et al. 2010c). This relationship and possible in situ transformation had also been proposed for the site of Gatzarria by Laplace (1966a) and is currently being re-evaluated by one of us (LE).

6.1.2 - The Vasconian Mousterian

39The Vasconian was initially defined by Bordes (1953) as a regional Mousterian facies characterized by the presence of cleaver flakes, usually made from non-flint stone. Although this artifact is frequently found in the Acheulian, it disappears at the beginning of the Middle Palaeolithic. After that, it is only found in any significant number in a very restricted geographical region : the Basque area (Spain and France) and Cantabria.

40In the 1970s and 1980s, as part of the larger culture-function debate challenging the Mousterian groupings of Bordes and the ethnic/cultural implications he attributed to them (see Szmidt 2003 for a detailed discussion of the Mousterian debate), a number of researchers (among them Cabrera Valdés 1983 ; Freeman, 1966, 1969-70) questioned the existence of the Vasconian as a separate facies. Based largely on typological evaluations and comparisons of Mousterian assemblages, they argued instead that it was simply a functional variant or even a normal part of the very heterogeneous nature of Mousterian assemblages in the Vasco-Cantabrian zone. Cabrera Valdés (1983) made the point that the Vasconian assemblages were very heterogeneous in terms of Bordes’ typological and technical indices, which would argue against their unity as a separate entity.

41It is difficult to defend the idea, however, that cleavers had a special function that could not have been served equally well by one of the other Mousterian tools. Use-wear data are only available on Acheulian quartzite and flint cleavers (and only on a very small sample), but they show that they were used on wood and meat (Utrilla and Mazo 1996), like many other Mousterian tools. A large-scale, interdisciplinary project is currently underway (Thiébaut ed. 2009) which includes use-wear analysis and experimentation aimed at discerning the possible function(s) of Mousterian cleavers.

42In addition, a recent technological study of a number of Vasconian assemblages has found that they are united not simply by cleaver production and a high frequency of cleavers, but also by technological characteristics, including the use of the Discoid method for most lithic reduction and similar retouch type (Deschamps 2008, 2009). Further study is pending on a larger sample of assemblages (M. Deschamps, Ph.D. in progress), but it seems that their unity and their existence as a separate entity deserves to be reconsidered. Chronological unity would bolster such an argument. Dates so far on such sites indicate a late Mousterian attribution, with most being < 45 000 and many < 40 000 BP (Allard 1993 ; Cabrera et al. 1996 ; Deschamps 2008 ; Maillo Fernandez et al. 2004 ; Mourre et al. 2005, 2006 ; Muñoz et al. 1989 ; Saenz de Buruaga and Ébrard 2004 ; Utrilla 2000). The Vasconian thus plays a key role in debates about late Neanderthal adaptations. If it is indeed a late industry, it could very well have overlapped in space and time with Aurignacian industries in the region.

6.2 - Discussion of Gatzarria results in these larger debates

43As outlined in this paper, Gatzarria is an important site for three reasons regarding the Middle to Upper Palaeolithic transition. (1) It is a good case study into how to approach collections of excavations carried out before the benefit of taphonomic analyses for the purposes of conducting new analyses. It shows the great care that must be exercised, especially regarding chronometric sampling, if one is to minimize the risk of error that could easily otherwise go unnoticed if the initial excavation data were taken at face value ; (2) It is one of the few sites to contain a Vasconian Mousterian industry ; (3) It is among the few sites to contain a stratified sequence of Proto-Aurignacian and Classic Aurignacian levels. In this section, we will take each of these aspects in turn.

6.2.1 - Use of older excavations/collections for radiocarbon sampling

44There are different ways of approaching a museum collection for radiocarbon dating of material excavated a while ago, depending on the type of question being asked and the required precision. If a site contains a single component or the components are clearly distinguishable and physically far apart, it may be less critical to do a detailed stratigraphic assessment for the purposes of selecting samples to date. Dating of organic artifacts characteristic of a particular period can also make such an assessment less necessary.

45It is very clear at Gatzarria that selecting organic samples for dating based on the layer ascribed to them by Laplace would, depending on the square sampled, invariably lead to erroneous dates. The number could be very precise and could be correct in terms of measurement, but it would not relate to the event (human occupation layer) one thought one was dating (i.e. it would not be accurate). This is one of the major methodological points we hope to get across in this article. Due to the age of excavation, we wanted to verify the integrity of the stratigraphy and layer attributions before doing any sampling for dating. Recent controversies regarding the dating of other older collections revolve in part around this issue. We urge archaeologists sampling for dating in museum collections of older excavations to take this issue into account a priori. We are also in favour of using strict protocols to choose, clean and evaluate samples, as we have done here and on other sites (Langlais et al. 2010 ; Szmidt et al. 2009 a,b , 2010 a,b,c).

6.2.2 - Vasconian Mousterian

46As mentioned, one of the key issues regarding the Vasconian is establishing its age. Dated sites until now have produced nearly exclusively recent dates (< 45 000, many < 40 000 BP). The new AMS 14C dates at Gatzarria, on cutmarked bones, help address this issue. According to these new dates (> 47 400 and > 50 300 14C BP), this industry at Gatzarria is clearly not a terminal Mousterian overlapping with the Aurignacian. This result is consistent with the fact that this layer is separated from the Aurignacian ones by a Châtelperronian layer and another Mousterian layer. In other words, there is no evidence that the Vasconian Mousterian Neanderthals (presumed makers of the Vasconian ?) lived contemporaneously with ‘Aurignacians’ based on these results. They are clearly separated by at least 7-10 millenia. Whether it dates to MIS 3 (the period in which these infinite dates fall) or an older period, however, cannot be determined by these dates. It is important to realize, though, that these results do not preclude other Vasconian Mousterian sites from having been inhabited later (i.e. for the Vasconian to have overlapped the Aurignacian). Overall, the chronological period to which the Vasconian Mousterian is attributed differs depending on the data examined ; this is what may be causing controversy and uncertainty in its date. Radiocarbon evidence is pointing to a date around 40 000 BP, taking into account many sites. The faunal spectrum identified at a number of Vasconian sites, however, is characteristic of a temperate climate, more likely attributable to MIS 5 according to some authors (Jaubert 2007). So either the Vasconian existed earlier than was thought (not only the end of the Mousterian), or the dates at 40 ka BP from other sites need to be re-evaluated. New zooarchaeological work on a number of sites (Olha, Gatzarria) would provide new data on this debate.

47Answering these issues will require more dating of samples from multiple sites. It will also require the use of a variety of dating methods on samples bearing clear association to human occupation (an important aspect in Mousterian contexts given that those caves were also often used by carnivores or denning bears) and which come from zones that have clearly been established as largely undisturbed.

6.2.3 - Proto-Aurignacian-Classic Aurignacian taxonomic and chronological relationship

48Regarding the Aurignacian, as outlined in section 6.1.1, one unresolved issue is whether the Proto-Aurignacian and the Classic Aurignacian are contemporary or not. Also, if they are contemporary, do they overlap substantially in time or only during a short time ? Purely stratigraphically, including at Gatzarria, the Proto-Aurignacian is below the Classic Aurignacian. However, 1) the issue as to whether these industries may have been contemporaneous across adjacent regions (Atlantic/Mediterranean) remains unresolved and (2) we do not yet have a firm enough grasp of the period in question in the Atlantic zone, i.e. the very region in which one sometimes finds both industries present within a sequence. The small number of Proto-Aurignacian dates and the large standard deviation of many Aurignacian (sensu lato) dates make it difficult to evaluate their chronological relationship. The new chronometric dates at Gatzarria give us the opportunity to discuss these questions, including some important methodological aspects.

50Looking further afield, this Gatzarria date is older than, or is statistically equivalent to, other Proto-Aurignacian dates. These include the new AMS radiocarbon dates on charcoal at Fumane level A2 (Higham et al. 2009), the dates of the deeper part of Mochi level G (Douka et al. 2012 ; Kuhn and Stiner 1998), Paglicci 24Bi (Palma di Cesnola 1999 ; Riel-Salvatore and Negrino 2009), the recently done dates of Grotte du Renne level VII (Higham et al. 2010) and Les Cottés Unit 04 lower (Talamo et al. 2012), as well as the conventional radiocarbon dates at Esquicho-Grapaou SLC1 (Bazile and Sicard 1999 ; Delibrias and Evin 1980). It is important to note that although many of these others appear to be younger than Gatzarria when only comparing the date itself, the situation is not so clear when we take into account, as one should, the standard deviations (overlap at one or two standard deviations). On the other hand, it is also true that other assemblages ascribed to the Proto-Aurignacian in Italy (Cala, Castelcivita, Serino) appear much younger than Gatzarria according to their dates (Riel-Salvatore and Negrino 2009), even if we take into account the standard deviations (no overlap even at two standard deviations). The recently obtained AMS dates on the Proto-Aurignacian at Serino, however, are much older than its conventional radiocarbon date (especially the dates done using the ABOx-SC protocol) (Wood et al. 2012).

51New dating on some of these assemblages/sites may need to be done in light of new archaeological and laboratory methodological advances in dating. Also, importantly, establishing whether all of these assemblages truly belong to the same industrial complex as Cjn2 requires re-assessment using identical criteria and methods of definition. The extent of variability within what has been termed ‘Proto-Aurignacian’ is not yet known, many of these assemblages having been examined prior to recent technological definitions of this industry. Thus it is important to remember that these assemblages may not be strictly comparable, nor accurate in their dates ; conclusions may change in the near future in light of new analyses.

52At the other end, the dates of layer Cbf at Gatzarria (Classic Aurignacian) (34 250 ± 550 and 34 400 ± 550 14C BP) are statistically equivalent and are in accordance with the older Classic Aurignacian dates of the neighbouring and recently-dated site of Brassempouy (layers 2F and 2DE) (Fontugne in Henry-Gambier D. and Bon F. eds., in preparation). They are also on the older end of many Classic Aurignacian sites in the Dordogne region and other neighbouring regions (Arrizabalaga et al. 2009 ; Jöris and Street 2008 ; Mellars 2000 ; Talamo et al. 2012 ; Zilhão et al. 2008). These dates are as old, or older, than some of those obtained for the Proto-Aurignacian in southeastern France and Italy.

54Until recently, young dates in the Aurignacian (both the Proto-Aurignacian and Classic Aurignacian) may have been more likely caused by a small amount of more recent contamination not having been fully removed during pre-treatment. Based on recent research done, ultrafiltration seems to reduce this risk by eliminating lower weight molecules likely to be potential contaminants or degraded collagen (Brock et al. 2007 ; Bronk Ramsey et al. 2004a ; Higham et al. 2006 ; Jacobi et al. 2006). This hypothesis of a later-lasting Proto-Aurignacian still needs to be confirmed from a larger number of recently-dated sites to bolster the argument of the pattern seen so far. Those of Les Cottés Unit 04 lower, also done using ultrafiltration, show similar results (Talamo et al. 2012).

55At Isturitz Cave, also located in the Pyrenean region (fig. 1, #9), and also bearing both industries in its stratigraphic sequence, there is evidence that an in situ transformation from Proto-Aurignacian to Classic Aurignacian might have occurred in layer C 4c4, dated to 37 180 ± 420 ¹⁴C BP (thus 38 020 - 36 340 BP, at two SD) (Szmidt et al. 2010c). At Gatzarria, the occurrence of the changeover is seen a bit later. Whether we are witnessing a replacement or instead an in situ transformation at Gatzarria has yet to be determined through detailed re-examination of the Cbf material in the way in which Cjn1 and Cjn2 have been re-analyzed. In any case, at Gatzarria, based on the dates produced by these samples, this would have occurred sometime between 34 900 and 33 300 BP (fig. 10).

56To sum up, the results obtained at Gatzarria indicate the presence of the Proto-Aurignacian in the Atlantic zone at least as early as in regions East of it, judging by the date on the deeper Cjn2 sample (36 300 ± 700 14C BP). As mentioned above, this date fits perfectly with other Proto-Aurignacian dates of the region. The Gatzarria results also show that the Classic Aurignacian may have developed earlier than usually suspected, possibly prior to 34 000 BP judging by the two statistically equivalent dates obtained for Cbf (34 250 ± 550 and 34 400 ± 550 14C BP). Do these observations allow us to conclude not only that the Proto-Aurignacian began at least as early in the Atlantic zone as in the Mediterranean zone, but also that it ended earlier in the former, giving way to the Classic Aurignacian earlier than in the latter region ? As discussed above, these results do not help resolve the issue of contemporaneity ― we do not yet know the precise timing of the change from one facies to the other, given that the dates of the two facies overlap. Redating Proto-Aurignacian sites using more modern cleaning, measurement and archaeological selection techniques (as is being done in this larger dating project), would help to resolve this issue, given that it often leads to improved precision (smaller standard deviations) and accuracy.

We are very grateful to the Executor of the G. Laplace estate, the late F. Lévêque, as well as to J.-J. Cleyet-Merle, Curator of the Musée National de Préhistoire (Les Eyzies-de-Tayac, France) for having granted us permission to study and sample the collection from Gatzarria. All of the staff at the Musée National de Préhistoire helped to make our stay welcoming and our access and work on the samples efficient and productive. In particular, we wish to express our thanks to Stéphane Madelaine for having carried out the taxomonic identification of the dated bones and to Philippe Jugie for having taken photographs of the bones prior to their dating. We also wish to thank the staff of the ORAU for their work. François Bon and three reviewers read an earlier draft of this paper and provided some very good comments, for which we are grateful.