A computational linguistic methodology for assessing semiotic structure in prehistoric art and the meaning of southern Scandinavian Mesolithic ornamentation

Abstract

Non-figurative prehistoric art is comparatively common yet challenging to decode and hence to understand. At the heart of the difficulty of assessing the presence of semiotic structure in prehistoric art is a lack of appropriate, replicable, and case-transferable methodologies. We here propose a novel approach derived from computational linguistics, in which k-skip-n-gram (skipgram) models and associated pointwise mutual information (PMI) measures are customised to the analysis of prehistoric art. In applying this methodology to a large corpus of portable art from the South Scandinavian Mesolithic, we demonstrate how the mutual relationship between individual motifs can be established. In the case of Mesolithic portable art, our analysis strongly suggests that there is no evident semiotic structure – this was likely not a form of proto-writing – but salient changes of motif occurrence over time are detectable. These changes are sensitive to changes in population density, structure, and connectedness, and may relate to increased territoriality in the Late Mesolithic. The method presented here is readily case-transferable and renders possible further linguistic and semiotic analyses of prehistoric art.