Defining similarity: An arithmetic method for archaeological source provenance targeting using geochemical data

• New, petrologically-grounded arithmetic method for assessing lithic artefact-outcrop provenance linkages. • Method provides a quantified ‘Similarity Score’ based on geochemical analysis. • Demonstrated applicability across rock and data types using published examples. • Able to identify prospective outcrops, rock suites and areas for further targeted archaeological study. Geochemical analysis has long been a key tool in studies attempting to determine the source provenance of lithic artefacts. Cogenetic linkages between outcrop sources and artefacts have been variously argued through visual, statistical and arithmetic interrogations of geochemical data. However, such approaches can overlook the petrological significance of particular elements and suggest artefact-outcrop ‘matches’ that may not pass geological scrutiny. Recent investigations of the same geochemical dataset for the sarsens of Stonehenge, for example, have yielded very different, and irreconcilable, interpretations regarding their provenance. This disagreement has provided the opportunity to develop and test a new arithmetic method for determining the geochemical similarity between lithic artefacts and potential source outcrops. The method, involving the simultaneous evaluation of multiple trace element/Zr (E i /Zr) ratios, is mathematically simple, petrologically grounded and offers a standardised way to target subsequent outcrop sampling. We first test the method on the Stonehenge sarsen dataset and reaffirm previously suggested provenance linkages. Further tests on datasets from provenancing studies involving basalt, andesite, dolerite and obsidian artefacts demonstrate its utility across a range of lithologies and on datasets obtained using higher and lower precision analytical methods. In most cases our test results confirm the source interpretations of the original investigations, but in others provide petrologically grounded avenues for further investigation not previously considered. Throughout the paper, we caution that the trace elements included in any arithmetic comparison need to be considered carefully in terms of their petrogenetic behaviour.