Current problems in dating Palaeolithic cave art: Candamo and Chauvet. (Method).

Pettitt, Paul ; Bahn, Paul

Introduction--style versus radiocarbon

It is now 12 years since Michel Lorblanchet first coined the term `Post-stylistic era' to denote the new period dawning in Palaeolithic art studies, in which direct dating was going to play the definitive role (Lorblanchet 1990: 20). Subsequent publications on this topic (e.g. Lorblanchet & Bahn 1991; 1993a) aroused controversy in some circles, but much of this was due to misunderstanding of the position adopted. In particular, some critics claimed that it was being argued that direct dating had entirely replaced, or would soon replace, the use of style in establishing chronologies. Nothing could be further from the truth. The use of the term `post-stylistic' merely denoted the arrival of a new phase, but did not reject the value of style: `It is self-evident that the impact of absolute dating methods on other areas of archaeological study, while enormous, has changed but by no means obliterated the role of typologies of stone tools or pottery. The term "post-stylistic" does not suggest the death of style, any more than the term "post-glacial" means that ice vanished from the face of the earth!' (Lorblanchet & Bahn 1993b: v).

It is obvious that style will continue to play a major role, because very few paintings are eligible for direct dating--few contain organic material, and in any case the procedure is extremely expensive, so the vast majority of Palaeolithic parietal figures will always have to be dated by other, indirect means. The present principle should be that direct dating and stylistic chronology will need to exist side by side, complementing each other or exposing inconsistencies. This paper discusses some examples where radiocarbon dating and stylistic dating are currently discordant. We suggest that while we must be ready to adapt traditional stylistic sequences to new absolute dates, the context and chemistry of the radiocarbon samples on the cave wall are critical, and can give rise to dates that may be anomalous.

Increasing numbers of radiocarbon determinations from samples of parietal art have been obtained over the past decade, with dates currently published from over a dozen caves in France and Spain (Bahn & Vertut 1997: 75). Most of the results conform well to stylistic dates which in turn are corroborated by comparisons with portable art--the Palaeolithic parietal art of Eurasia is blessed in that it stands alongside a corpus of thousands of contemporaneous portable images which are generally well dated. As predicted (Bahn 1993), radiocarbon has also proved useful in weeding out fakes and wrongly attributed motifs; for example, in the Pyrenean cave of Labastide, some `organised black dots' in three different places were thought to be Magdalenian signs, but have yielded radiocarbon dates of modern times, the 15th century AD, and (for `the most Magdalenian' examples) the 10th century AD (Simonnet 1999: 188).

Sources of error

At the same time, some radiocarbon dates seem to be in severe conflict with stylistic expectations, ostensibly demanding a startling new perspective on the artistic sequence. However, in these cases it may not always be the style that needs rethinking: sources of radiocarbon error are well known and can be determinant. For example, the date of charcoal used in drawing is the date of death of the tree-rings incorporated, not the date of the tree, or the date that the charcoal was made or used to draw with; so the date obtained can be much earlier than the drawing. The risks of contamination on an exposed rock surface are also high, both from material that is too old (fossil carbon) and material that is too young (microorganisms infesting the surface).

AMS dating, in particular, has taught us to be increasingly critical of what is measured, since very small samples adjacent to each other on a cave wall may in fact derive from different materials with different formation processes. The AMS dating of cave-art pigments is a relatively new application, and rock painting samples are some of the most difficult to date reliably and directly, for at least five reasons which we summarise here (Hedges et al. 1998):

* the necessarily small size of available samples (usually 10-50 mg) much of which is calcite or bedrock, and the very small amounts of datable carbon ultimately obtained from these (usually <1.5 mg, sometimes <0.5 mg),

* the exposure of such samples to the environment over long periods of time;

* the complex chemical history of charcoal, pigments, rock substrates and surface organic accumulations;

* the lack of association between such samples and other datable materials which may be used as dating crosschecks;

* the relative lack of experimentation in pre-treatment, understandable given the precious character of the original material

The principal sources of carbon in cave art are derived from charcoal, used for drawing or deposited as soot from lamps, and organic materials used in pigments, which may be extracted in solution as "humic acids". Forms of intrusive carbon are microfauna or micro-organisms infesting the cave walls. Charcoal should be stable and thus provide the most reliable samples, but even here, measurements can be inconsistent. Charcoal samples in apparent association selected for direct dating can have often widely different [sup.14]C ages, such as those fragments from the floor of Cosquer Cave which ranged from c. 15 500 to c. 27 900 BP in age, a difference of two half-lives (Clottes et al. 1992b). If one assumes that it is such charcoal fragments that formed the source material for execution of the art on the walls, it is easy to see how resulting AMS measurements will reflect the differing ages of the charcoal rather than the execution of the art.

Pigment and binding compounds are more subject to later changes by mineral, chemical or micro-biological invasion in ways that are not yet well understood. They may add more recent carbon by percolating in surface water, or by ingesting atmospheric carbon, or perhaps add earlier carbon by digesting fossil carbonates (Gillespie 1984). There is a possibility therefore that the dates obtained are composite results to which carbon from different sources and ages may have contributed.

One way in which the relative homogeneity of carbon in rock-art samples may be examined is by dating the soluble humic acid fraction, which can be isolated from the charcoal carbon (often referred to as the humin fraction) during routine pre-treatment for this material. `Humic acid' is the umbrella term for the products of the organic breakdown of plant materials. These acids are poorly understood, highly mobile in soils, sediments and through rocks, and often accumulate in the porous structures of charcoal, wood and bone (Gillespie 1984: 7). Humic acids, therefore, may contain carbon from a number of sources, each of which bears little relation to the age of the charcoal sample containing them in its chemical matrix. If, however, the resulting wood-charcoal and humic acid measurements are the same at 1 or 2 sigma then this may be a good indication of the relative integrity of the material selected. For example, two measurements of the wood-charcoal carbon on the depiction of a horse from Cos quer Cave (Gif A 92416, 18 840[+ or -] 240 BP; Gif A 92417, 18 820 [+ or -] 310 BP) are in statistical agreement with one on a humic fraction from the same depiction (Gif A 92422, 18 760 [+ or -] 220 BP) (Clottes et al. 1992a). But other paired measurements at Cosquer reveal that the situation is not entirely clear: charcoal carbon from a depiction of a bison was dated to 18 010 [+ or -] 190 BP (Gif A 92419) while the humic fraction was dated to 16 390 [+ or -] 260 BP (Gif A 92423) which is a statistically different age at 2 sigma and enough to raise doubts as to the homogeneity of the carbon. Humic fractions can also contain older carbon than the charcoal fractions, as in the case of both small and large left-facing bison at Altamira which, while closer in age than the above examples, still do not overlap at 2 sigma (Valladas et al. 1992).

A case study--Candamo

At the Palaeolithic decorated cave of Candamo, in Asturias (northern Spain), samples from black dots were initially dated to more than 32 000 years ago, which was surprising, especially as there were older red paintings beneath them (Fortea 1999: 6-27). An unusually full account of the dating procedure has now been provided, together with the results produced by another laboratory (Fortea 2000/1).

The original paint samples were taken in 1993 from intact black dots on a panel with several figures of bulls. This is a complex panel, with several phases of pictorial activity, some before and some after the black dots, and there is clear evidence for some retouching here and there (Fortea 2000/1:188-90). Sample CAN12 was composed of pigments from two black dots, one from the head of bull 15, another from just right of the tail of bull 16, which were combined by J. Fortea Perez and sent to the Laboratoire des Sciences du Climat et de l'Environnement (LSCE) at Gif-sur-Yvette (Fortea 2000/1:191). The result, produced in 1996 from a total mass of carbon of 1540 micrograms, was 32 310[+ or -]690 (Gif A 96138), i.e. 33 690-30 930 BP at 2 sigma. In 1999, LSCE provided a second determination from Candamo which broadly corroborated the first, produced from 460 micrograms which remained of CAN- 12:33 910[+ or -]840 BP (Gif A 98201). The laboratory suggested that either the prehistoric people might have used ancient charcoal for their drawings, or perhaps the drawings really were done at this remote period.

In 1997, the late Manuel Hoyos was asked by Javier Fortea to join him in sampling the same dots--from points immediately adjacent to those previously sampled. An equivalent amount of material was taken and first examined with a scanning electron microscope [SEM]. Hoyos' report (in Fortea 2000/1:191-96) showed that one sample (called CAD-1 and CAD-1B) was `normal', with the morphology, structure and composition of plant charcoal. The second sample, however, CAD-2, was burnt bone, which, since the carboniferous collagen is depleted during burning, is relatively deprived of carbon and can depress the date. The SEM also revealed the presence of micro-organisms in both samples, of a kind noted in others from Altamira and Tito Bustillo (Hoyos, in Fortea 2000/1:193). This extraneous organic material (bacteria, etc) is fundamentally composed of carbon, but may have drawn its carbon from pre-existing carbonates, not from the surrounding atmosphere (Gillespie 1984). These carbonates are inevitably more ancient than their host micro-organisms, so that the latter incorporate [sup.14]C, [sup.13]C and [sup.12]C from older samples. If the carbonate attacked by the microorganisms is of an infinite [sup.14]C age (effectively >50 000 BP), then its isotopic composition will only have [sup.13]C and [sup.12]C which, when added to the carbon sample being measured, will diminish the [sup.14]C ratio in the total, and thus make the result older than it should be. The sample may therefore have been rich in carbon but relatively reduced in [sup.14]C.

In February 2001 Fortea (2000/1:196-97) took new samples from exactly the same groups of dots in Candamo--"decidimos volver a muestrear ... los mismos lugares" (we decided to resample the same places) (Fortea ibid.: 196; and see Figure 5, p. 189 for the precise location from which the samples were taken) and sent them to Geochron Laboratory in the USA which produced results as follows:

CAN-3:15 160[+ or -]0 BP ([sup.13]C: -27.2. Weight of sample 87.1 mg. Weight of carbon for analysis: 280 micrograms. Carbon rendered: 0.32%) (GX-27841-AMS)

CAN-4:15 870[+ or -]90 BP ([sup.13]C: -27.0. Weight of sample: 64.9 mg. Weight of carbon for analysis: 289 micrograms. Carbon rendered: 0.43%). (GX-27842-AMS)

Geochron has affirmed that the small size of the samples was not critical, as the smallest sample dated by this lab was of only 30 micrograms of carbon.

We have an intriguing situation here, in that the two LSCE results are consistent, i.e. statistically the same age, as are the two Geochron results, but there is a discrepancy of approaching three half-lives between the two sets. Such a difference, if due to contamination alone, would require about 90% dead carbon contamination to change a `true' age of around 15 000 BP to around 33 000 BP, or alternatively about 13% modern carbon contamination to change a `true' age of around 33 000 BP to around 15 000 BP (M. Rowe pers. comm.).

In explanation of the discrepancy, Geochron suggested that either different parts of the wall were painted or retouched at different times, or that there were problems of contamination. Given the complex and small size of the samples there may have been absorption of organic contaminants of different ages through the effects of flowing water. The samples dated--as with all cave art pigments--contained a great quantity of calcareous wall (the carbon content of which has not been reported or used in the evaluation of the result) and less than 0.5% of `non-carbonate' carbon. Alternatively the more recently dated samples may have been permeated with more recent micro-organisms.

As the information currently stands, the Candamo paintings thus admit to one of three conclusions: that they were painted 33 000 years ago (Aurignacian), but some samples contained more recent carbon. Or that they were painted 15 000 years ago (Magdalenian), but some samples were contaminated by more ancient carbon. Or painting took place in both the Aurignacian and Magdalenian and that pigment survives from both these periods (Fortea 2000/1: 197-98). On learning of the Geochron results, LSCE agreed that there may have been types of pigment of differing ages, or that prehistoric people retouched the paintings at different times (Fortea ibid.: 197). It is important that lessons should be learned from the Candamo experience and applied to other caves, most notably Chauvet.

A second case study--Chauvet

The cave at Chauvet (Ardeche, France) provides perhaps the most interesting and controversial of all recent dating exercises in European cave art. Here, in some opinions, the radiocarbon dates can be said to be at odds with each other as well as with considerations of style. An example of conflicting dates obtained by radiocarbon dating is provided by a sample taken from a black horse figure. This gave a result of 20 790[+ or -]340 BP (Gif A 98157) from charcoal carbon but 29 670[+ or -]950 BP (Gif A 98160) from the humic fraction (Valladas et al. 2001: 33). The scholars working in this cave have chosen to adopt the earlier date, derived from the humic fraction, and asked for the style to be reconsidered in that light: "the age of the humic fraction resulting from the basic treatment ... appears to us to be more reliable" (Valladas et al. 2001: 216). But given the problems of formation processes associated with humic fractions, as mentioned above, it may be that the earlier date should be treated with caution.

In this case we also feel that stylistic arguments for a later date are persuasive. Zuchner (1996; 1999) has highlighted the many features of Chauvet's art which seem to him to imply a later period: the depiction of Megaloceros and reindeer, hand prints and hand stencils; the Magdalenian style of some bison, rhinos and big cats; the treatment of horse movement and anatomy; the rendering of volume, shading and perspective; the representation of herds and lines of animals; the full-face heads. He believes strongly that the cave's art belongs to the Gravettian (red figures) and to the late Solutrean/early-mid Magdalenian (black figures). The recent book on the cave (Clottes 2001 b) has reported yet more figures which seem to support Zuchner's scepticism: there is what looks like a red claviform, a sign which is closely linked with the middle Magdalenian. There is also an engraved tectiform (Clottes & Le Guillou 2001: 148), another classic Magdalenian phenomenon; and an indented circle (Aujoulat & Gely 2001: 91), yet another post-Aurignacian (probably Gravettian) feature. There are a whole series of scenes--the famous confronting rhinos, two scenes featuring lion couples, a complex hunting scene--and scenes, always extremely rare in Ice Age art, are not really known elsewhere till somewhat later than the Aurignacian, and primarily the Magdalenian. The extensive scraping of the wall surfaces prior to the figures being drawn is known elsewhere, but primarily at the Magdalenian caves of Altxerri and Covaciella (Aujoulat et al. 2001: 152). Clottes (1996b: 282; 2001a: 213) sees the cave's `vulvas' as being linked to Aurignacian specimens from south-west France, but they are in fact far more similar to examples from Magdalenian sites, and bear little resemblance to the range of Aurignacian motifs which some see as being vulvas (see Bahn 1986). In addition, it has to be borne in mind that virtually all of the other decorated caves in the Ardeche probably date to the Solutrean and Magdalenian.

Clottes (2001a: 219; see also 2001b: 63, 68) accepts that Magdalenians may have entered the cave, though he sees no evidence that they made any figures other than the claviform. Clottes is certainly correct to warn that stylistic criteria are by no means an infallible diagnostic tool (1996a: 27; 2001a). We are still ignorant of many aspects of Ice Age art such as the origin and duration of numerous features, and we know of remarkably sophisticated figurines from the European Aurignacian. Nevertheless, the rock and cave art which is definitely known to be Aurignacian looks pretty crude and simple, a long way from Chauvet--which of course is why the Chauvet dates caused such a shock. One certainly cannot deny that Chauvet may indeed date to the Aurignacian, but in simple terms of the laws of probability, what are the chances that a single Aurignacian cave would contain so many different features, themes, styles and techniques which, over a hundred years of study, have become so strongly and indubitably associated with much later periods?

Conclusion

These examples show how crucial is the dating of cave art and how difficult it remains, and where radiocarbon and stylistic studies disagree, it is clear that we need a critical approach to both., In radiocarbon determinations few things are more important than the composition and context of the sample: AMS dating is revealing that many previously dated samples were mixtures, so small samples, even taken from the same paint mark, may have very different origins and histories. The preparation of pigment may initially involve gathering the ashes from a hearth; but if this results in one sample being charcoal and another burnt bone, the dates obtained will vary. Similarly, the degree of contamination may vary greatly within a short distance, and the contaminants may raise or lower the date depending on whether the intrusive carbon is ancient or recent.

* One necessary prescription is to use a SEM to determine the nature and composition of samples as far as possible before measurement and isolate the material likely to have the greatest integrity.

* Another is to take control samples of unpainted bedrock close to the dated depictions and to recover humic fractions from them as a guide to the extent of contamination from the local chemical environment. Some contaminants (such as oxalates) may not always be successfully removed by standard acid pre-treatments.

* The obvious scientific strategy which must be followed is to broaden the range of AMS laboratories undertaking such cutting-edge measurements of complex samples. A call for independent verification of results by separate laboratories was first made by Lorblanchet & Bahn (1999:119) and it may be that given that large differences in date may be due to differences in pre-treatment it would be sensible for replication of measurement to become routine. Of more than 60 direct AMS radiocarbon dates published for French and Spanish Upper Palaeolithic cave art, all but two were undertaken by LSCE at Gif-sur Yvette.

For Candamo, attention is naturally attracted by the anomaly that one laboratory should only have samples from one period, and the other should only have samples from a different period apparently up to three half lives younger, when the samples were taken from essentially the same group of apparently identical dots. There is no reason to doubt that both laboratories routinely meet the highest standards in filtering and processing samples, and controlling for pollution by modern carbon. But it would be interesting to discover if current differences in pre-treatment, analysis and measurement methods were sufficient to account for such different results. For such a control the determinations would of course need to be made on material divided from exactly the same sample. In any case, given the potentially serious problems with plateaux and age inversions of measurements over c. 5 half-lives (i.e. 30,000 BP) we should in any case treat ages beyond this as provisional at best (e.g. Richards & Beck 2001; Pettitt & Pike 2001).

The implications of the Aurignacian dates at Candamo and Chauvet for our knowledge of the early development of Ice Age art are immense. It is therefore imperative that the dating programme be enhanced and the results corroborated as far as possible, by investigating the micro-composition of samples, investigating formation processes, testing thoroughly for possible contaminants, and splitting samples for use by several co-operating laboratories. In the recent sequencing of MtDNA from Neanderthal remains, independent replication of results by a number of other institutions was a crucial and integral part of the research design from the outset. Independent verification of results should be undertaken routinely with all cutting-edge science in archaeology. Our purpose here, therefore, is not to question the performance of the laboratory at Gif-sur-Yvette, which has led the pioneering programme for the AMS dating of cave art in Europe, but to call for replication of the results and the research and for full publication of the context and treatment of samples, a call with which all scientists will surely agree.

Acknowledgements

The authors would like to express their appreciation to Javier Fortea for making his Candamo data available, and to Marvin Rowe and an anonymous reviewer for invaluable comments, but all responsibility for the conclusions of this article is ours alone.

References

AUJOULAT, N., D. BAFFIER, V. FERUGLIO, C. FRITZ & G. TOSELLO. 2001. Les techniques de l'art parietal, in J. Clottes (ed.), La Grotte Chauvet. L'Art des Origines: 152-8. Paris: Le Seuil.

AUJOULAT, N. & B. GELY. 2001. La salle Hillaire ..., in Clottes (ed.) 88-95.

BAHN, P.G. 1986. No sex, please, we're Aurignacians,' Rock Art Research 3: 99-120. 1993. The `dead wood stage' of prehistoric art studies: style is not enough, in Lorblanchet & Bahn (ed.): 51-9.

BAHN, P.G. & J. VERTUT. 1997. Journey through the Ice Age. London: Weidenfeld & Nicolson.

CLOTTES, J. 1993. Post-stylistic?, in Lorblanchet & Bahn (ed.): 19-25.

--1996a. The Chauvet Cave dates implausible?,--INORA 13: 27-9.

--1996b. Thematic changes in Upper Palaeolithic art: a view from the Grotte Chauvet, Antiquity 70: 276-88.

--2001a. Conclusion, in Clottes (ed.): 210-14, 219.

--(Ed.). 2001b. La Grotte Chauvet. L'Art des Origines. Paris: Le Seuil.

CLOTTES, J., J. COURTIN & H. VALLADAS. 1992b. A well-dated Palaeolithic cave: the Cosquer Cave at Marseille, Rock Art Research 9: 122-9.

CLOTTES, J., J. COURTIN, H. VALLADAS, M. CACHIER, N. MERCIER & M. ARNOLD. 1992a. La Grotte Cosquer datee, Bulletin de la Societe Prehistorique Francaise 89 (8): 230-34.

CLOTTES, J. & Y. LE GUILLOU. 2001. La Salle du Fond, in Clottes (ed.): 128-48.

FORTEA, J. 1999. El arte paleolitico, El Campo de las Ciencias y las Artes 136(6): 1-37.

--2000/1. Los comienzos del Arte Paleolitico en Asturias: aportaciones desde una arqueologia contextual no postestilistica, Zephyrus 53/4: 177-216.

GILLESPIE, R. 1984. Radiocarbon user's handbook. Oxford: Oxford University Committee for Archaeology. Monograph 3.

HEDGES, R.E.M., C. BRONK RAMSEY, G.J. VAN KLINKEN, P.B. PETTITT, C. NIELSEN-MARSH, A. ETCHEGOYEN, J.O. FERNANDEZ NIELLO, M.T. BOSCHIN & A.M. LLAMAZARES. 1998. Methodological issues in the radiocarbon dating of rock paintings, Radiocarbon 40(1): 35-44.

LORBLANCHET, M. 1990. The archaeological significance of the results of pigment analyses in Quercy caves, Rock Art Research 7(1): 19-20.

LORBLANCHET, M. & P. BAHN. 1991. Rock art studies: the post-stylistic era. Where do we go from here? Rock Art Research 8 (1): 65.

--(Ed.). 1993a. Rock art studies: The post-stylistic era or Where do we go from here?: 51-9. Oxford: Oxbow. Monograph 35.

--1993b. Introduction, in Lorblanchet & Bahn (ed.): v-viii.

--1999. Diez anos despues de la `era postestilistica': Donde estamos ahora? Edades, Revista de Historia 6: 115-21

PETTITT, P.B. & A.W.G. PIKE. 2001. Blind in a cloud of data: problems with the chronology of Neanderthal extinction and early modern human expansion, Antiquity 75: 415-20.

RICHARDS, D.A. & J.W. BECK. 2001. Dramatic shifts in atmospheric radiocarbon during the last glacial period, Antiquity 75: 482-5.

SIMONNET, R. 1999. Les Magdaleniens dans les Pyrenees. La Grotte de Labastide (Htes- Pyrenees), Espace reel et espace imaginaire, Archeologie des Pyrenees Occidentales et des Landes 18: 183-209.

VALLADAS, H., H. CACHIER, P. MAURICE, F. BERNALDO DE QUIROS, J. CLOTTES, V. CABRERA VALDEZ, P. UZQUIANO & M. ARNOLD.--1992. Direct radiocarbon dates for prehistoric paintings at the Altamira, El Castillo and Niaux caves, Nature 357: 68-70.

VALLADAS, H., N. TISNERAT, M. ARNOLD, J. EVIN & C. OBERLIN. 2001. Les dates des frequentations, in Clottes (ed.): 32-4, 216.

ZUCHNER, C. 1996. The Chauvet Cave: radiocarbon versus Archaeology, INORA 13: 25-7.

--1999. La cueva Chauvet datada arqueologicamente, Edades, Revista de Historia 6: 167-85.

Paul Pettitt & Paul Bahn*

* Pettitt, Keble College, Oxford OX1 3PG, England. (paul.pettitt@keble.ox.ac.uk) * Bahn, 428 Anlaby Road, Hull HU3 6QP, England.

Received 29 May 2002; Revised 2 September 2002; Accepted 3 January 2003.