First Geochemical ‘Fingerprinting’ of Balkan and Prut Flint from Palaeolithic Romania: Potentials, Limitations and Future Directions

Abstract

Long-distance raw material transfers across Romania prior to the Last Glacial Maximum have previously been inferred from either visual and/or petrographic observations of East Carpathian sites. We investigated the potential to 'fingerprint' flint from archaeological sites at Mitoc-Malu Galben and Bistricioara-Lutărie III in Eastern Romania, using in situ highprecision analyses of 28 major, minor and trace elements determined by laser ablationinductively coupled plasmamass spectrometry (LA-ICP-MS) in combination with multivariate statistical analysis. Our results suggest that geochemical analyses have the ability to distinguish between different geographical sources but are unable to positively associate flint artefacts from archaeological contexts to these geochemical groups. The mismatches of signatures between artefacts and geological materials, however, raise new questions and open unforeseen perspectives.

Evaluating machine learning techniques for archaeological lithic sourcing: a case study of flint in Britain

“…Despite widespread documentation on the correct usage of these techniques, several problems can be identified in the recent literature on lithic sourcing. These include basic prerequisites such as inadequate sampling from individual geological sources for machine learning techniques to effectively learn from 23 , to perhaps more importantly, a large number which use classification techniques with no ‘none of the above’ or ‘other’ class or method to discriminate from the geological sample sites used to compare artefacts with 20 – 22 , 24 , 25 . The failure to create such a class or method for the model to use can lead to false positive results (type I errors), allowing no option to rule out the geological sample sites used.…”

Geochemical fingerprinting of Pleistocene stone tools from the Tràng An Landscape Complex, Ninh Bình Province, Vietnam

“…Portable XRF has limited analytical resolution, and some lighter diagnostic elements that might further narrow down the geological source (e.g., sodium, potassium, silicon, magnesium) are not detected well by this method [ 90 ]. The application of more precise methods, including LA-ICP-MS [ 83 ] and thin sectioning [ 59 , 91 ], might help shed more light on the geographic source of the archaeological igneous material, though will carry the downside of being destructive.…”

“…Principal Component Analysis (PCA) is statistical clustering method used to simultaneously minimize information loss and increase interpretability of multivariate datasets by transforming large sets of variables into smaller components [ 81 ]. It has been previously and successfully applied as a method to analyze XRF data of stone tools [ 66 , 67 , 82 , 83 ]. Here, it is used to explore variability within raw material subgroups.…”

Geochemistry and petrology of striped chert as a provenance tool for artefacts from the KRZEMIONKI NEOLITHIC mining area (Poland)

“…Numerous geologic sites of chert occurrences and prehistoric mining and dressing are located primarily within the Iłża‐Ożarów zone of the northeastern Permo‐Mesozoic margin of the HCM (Figure 1a). This zone is characterized by the presence of the most decorative striped chert concretions unusual in other sites throughout Poland and the world (e.g., Högberg et al, 2012; Hughes et al, 2012; Kochman, Kozłowski, & Matyszkiewicz, 2020; Kochman, Matyszkiewicz, & Wasilewski, 2020; Mathur et al, 2020; Moreau et al, 2019). Although striped cherts also occur in the southwestern Permo‐Mesozoic margin of the HCM, they do not show such clear‐cut banded patterns, and moreover, they are usually cracked and less cohesive (Kutek, 1962; Migaszewski, 2005; Migaszewski & Olszewska, 2002).…”