Stone tools – their design details and the way in which they are made – make up the core data through which we understand the deep past. True enough, the term ‘Stone Age’ may be misleading given what we now know about how ancient hominins used wood in sophisticated ways. Yet, stone tools preserve well in many different habitats; they simply make up the bulk of the material we have at our disposal to study past processes of cultural transmission, migration, and adaptation.
That said, there are countless ways to approach stone tools and in Europe alone a number of different schools have developed more or less compatible ways of doing so. Typological classification and the recording of technological traits are among the most important here but have their roots in largely pre-computational times when the efficient handling of large amounts of data was difficult. Stone tool analysis also rarely – all too rarely, really – considers issues of intra- and inter-observer variability. A lot of previous research has highlighted how political and linguistic borders, for instance, have had an impact on how we demarcate and hence understand past cultural variability. All of these factors and biases work against large-scale synthesis.
In a pair of papers recently published in Scientific Data and in PloS One respectively, we take stock of a major recent effort to truly synthesise stone tool data across Europe with a specific focus on the very end of the last Ice Age, the period from 15,000 to 11,000 before present. This effort is part of the ERC-funded project CLIOARCH that seeks to understand human niche change and adaptation in Europe at the very end of the Pleistocene. The data we use was generated through expert-elicitation with leading researchers from different regions across Europe contributing tabulated data, images and references for their respective areas – it’s a ton of data. We then use a suite of computational methods to see how the incoming data cluster in relation to traditional cultural taxonomic labels, and whether space or time have major impacts on the similarities and dissimilarities among these objects and regions.
The results are, in many ways, quite challenging. None of the data we use – stone tool shape, typology or technology – really offer any neat spatial or temporal patterns, something we already saw in an earlier study of organic tools as well. Using clustering algorithms, we investigate the hierarchies of relatedness among the objects and cultural taxonomic units suggested by our regional experts, but the resulting clusters often don’t place cultures that have the same names (e.g. the Brommean) but come from different regions (and experts) together, nor do units labelled as, for example, Early Azilian and Late Azilian line up tidily in the implied chronological sense. Evidently and in contrast to many traditional textbook scenarios, the data do not support a simple view of neatly ordered hunter-gatherer cultures who change and adapt their tool kit to changing climates or other pressures at this time. This goes contrary to previous visions for this period and does raise the question of how we bring archaeological data together with emerging palaeogenomic data for this period. Since we should not really be talking about clearly bounded cultural units at this time, we should also not interpret the ancient genetic data to reflect the movement of clearly bounded demographic units. The story is likely more complex and we might need to develop fundamentally different analytical approaches to synthesise not only stone tool data themselves but to articulate these data to genetic data.