A Bayesian alternative for aoristic analyses in archaeology

Abstract

Aoristic analysis is often used to handle chronological uncertainties of datasets where scientific dates (e.g., ¹⁴C and OSL) are unavailable, and observations are described by association to archaeological periods or phases. Although several advances have been made over the last 2 decades, the basic principle of this approach remains fundamentally the same. Temporal windows of analyses are first divided into regularly sized time blocks, and probability weight is assigned to each of these for every observation. Weights are then aggregated by time block, and the resulting vector of summed probabilities is interpreted as a curve representing changes in the intensity over time of a particular phenomenon. This paper reviews the basic principles and assumptions of aoristic analyses in archaeology, highlighting several issues with its application and interpretation, advocating for a Bayesian alternative implemented via baorista, a new package written in R statistical computing language. The robustness of the proposed solution is evaluated through a series of experiments based on simulated datasets, which showcase key advantages over aoristic analysis. Two specific solutions are considered: a parametric approach where data are fitted to specific growth models and a nonparametric approach that allows for the visualisation of the changing frequencies of events, accounting for sampling error and the peculiarities of archaeological periodisation.