Grotte des Fées de Châtelperron

6 minute read

Gravina et al. (2005) report on the radiocarbon stratigraphy of Grotte des Fées de Châtelperron, which is the original type site of the Châtelperronian industry. I think the paper is a very fine example of most archaeological papers I have read. The thing is, like many other papers, I'm not sure it provides enough information for me to understand the conclusion.

The most relevant finding is this:

Hitherto, direct archaeological evidence for [the temporal overlap of Chatelperronian and Aurignacian industries] has proved controversial, with the suggestion that supposedly direct 'interstratifications' of Chatelperronian and Aurignacian levels at three separate sites in western France and northern Spain (Roc de Combe and Le Piage in southwest France, and El Pendo in northwest Spain) might in fact reflect serious confusions or misinterpretations of the stratigraphy at these sites. The results reported here seem to provide clear evidence for a direct interstratification of distinctively Chatelperronian and Aurignacian occupations at the type-site of Châtelperron itself, closely dated by a sequence of 13 high-resolution radiocarbon accelerator mass spectrometry (AMS) measurements by the Oxford Radiocarbon Laboratory (Gravina et al. 2005:1).

The stratigraphy by itself would seem to be adequate for the conclusion: there are 10 typologically Aurignacian artifacts with provenience from the Chatelperronian levels, two perforated animal teeth consistent with Aurignacian typology, and nineteenth-century excavations recovered three unprovenienced Aurignacian blades and one split-based bone or antler point. The interpretation of "interleaving" comes from the observation that the majority of the provenienced Aurignacian artifacts came from one level (B4) of the Chatelperronian, both above and below other Chatelperronian levels (B1-B5). If these typologically Aurignacian artifacts were made by modern humans, then they demonstrate the temporal overlap of Neandertals and modern humans at the site.

Compared to this--a careful historical review--the dates are a bit of an anticlimax: 40,000 radiocarbon years for the bottom of the sequence, 34,500 to 36,000 for the top, and between 39,000 and 36,000 for level B4 with many of the Aurignacian-type artifacts.

The authors attempt a comparison between these radiocarbon dates and the Cariaco Basin deep sea core, which leads them to an interesting argument:

What is clear from the Cariaco basin data is that the overall time range of the human occupation is likely to be much shorter than the range of the radiocarbon dates, with the occupation of B5 centred on 42,000-43,000 yr BP, B4 on ~41,000-42,000 yr BP and that of B1-B3 on ~40,000-41,000 yr BP (ages estimated from the GISP2 ice core). Interestingly, the occupation in phases B5 and B1-B3 seems to be correlated with brief warmer spells visible in the delta 18O record. The two samples from B4 are somewhat different in age and correlate with the end of one warmer period and the start of the next, separated by an intervening colder phase. This coincides with the brief episode of Aurignacian occupation in level B4. A displacement of Aurignacian populations from central Europe to France in response to sharply colder conditions at this time (especially perhaps the onset of severe winters) would hardly be surprising in ecological and demographic terms, as winter temperatures in the more oceanic areas of western Europe are likely to have been significantly milder than those further to the east... (ibid., 5).

Interesting to me is the possibility that all this stuff actually unfolded over around 2000-3000 years of real (non-radiocarbon) time. This seems more intuitive than the persistence of distinct, static cultures across 6000 years or more, while intermittently in contact with different groups.

Reading between the lines of my description, you may sense a bit of hesitancy toward the conclusions. I'm not an archaeologist, and many of the concerns of trained archaeologists are lost on me. I wouldn't know an Aurignacian blade from a Chatelperron backed point.

But perhaps unlike many non-archaeologists, I have looked to see if I can tell the difference. I happen to have the edited volume Context of a Late Neandertal: Implications of Multidisciplinary Research for the Transition to Upper Paleolithic Adaptations at Saint-Césaire, Charente-Maritime, France on my shelf. Here are five of the Aurignacian blades from Chatelperron (Gravina et al. 2005:3):

Aurignacian edge-retouched blades from Chatelperron

Here are some of the Chatelperron backed points from Saint-Césaire, approximately to the same scale (Leveque 2003):

Chatelperron backed points from Saint-Césaire (Leveque 1993)

There is of course an obvious difference: the Aurignacian blades are retouched on both sides; the Chatelperron points are "backed", or retouched only on one side. But despite their names, the "points" are not necessarily pointier; the tools would appear to be close functional analogues.

Now here's my question: if you have a sample of 65 Chatelperron points from a site (unlike the upper Chatelperronian level of Saint-Césaire where there are only 8), how likely is it that you might find a few that look like Aurignacian blades?

The answer to that might be "very unlikely". Indeed, the fact that some of the Chatelperron "Aurignacian" artifacts are made on materials brought from far away might give an additional reason to think that they stand apart from the average Chatelperron point. And the split-based point is a likely marker of at least strong Aurignacian influence (although it is unprovenienced).

But I would be more convinced by a statistical answer to that question than a typological one. Out of those 65 backed points, how many of them have some retouch on the backed side? Any extensive? There are 750 artifacts in the Chatelperronian levels. Where did they come from? How many of them come from level H4? From the description here, it appears that the Aurignacian-type artifacts come directly from Chatelperronian layers -- that is to say, they are not "interleaved", they found with Chatelperronian-type artifacts. Is that accurate? If not, why not?

These are questions that I could ultimately answer myself through the magic of interlibrary loan, but that would take weeks. They ought to be in this paper. There is no possibility of understanding the importance of 10 atypical artifacts without some assessment of the range of variation of the "typical" ones. All this paper gives is an assertion that they are in fact Aurignacian. That they may be, but how do we know it?

Other related questions I wouldn't expect the paper to answer, but I would expect somebody else to look at soon. Are there artifacts at other Chatelperronian sites that fit these criteria? What would a finite mixture of two industries look like -- especially in the case where there is a lot of one industry and only a few artifacts of another? After all, the Aurignacian and Chatelperronian differ in a few fossiles directeurs, but for the most part they share a lot of tool types at different frequencies. Would that mix of frequencies be statistically identifiable, or would it be indistinguishable from different facies of the same industry?

Probably I'm thinking too much like a biologist. And to be honest, finite mixture analysis may be a bit much to expect from anybody.

So, the paper may be entirely correct for all I know. I'm just asking naive questions, beyond my expertise. But when two pictures look like the ones above, and they are supposed to be typologically identifiable products of "modern humans" on the one hand, and "Neandertals" on the other -- well, it seems to me there needs to be a bit more than an edge of retouch behind that conclusion.


Gravina, B., Mellars, P., & Ramsey, C. B. (2005). Radiocarbon dating of interstratified Neanderthal and early modern human occupations at the Chatelperronian type-site. Nature, 438(7064), 51-56. doi:10.1038/nature04006