The increase in the chronological data provided by dating laboratories as well as by excavations and artefact analysis over the last few decades makes it important that they be treated in an efficient and reproducible manner, in order to built the most reliable chronologies possible. The aim is to fix events on the calendric time scale and to estimate the span of the phenomena studied, taking into account all the uncertainties. Since the 1990s, new software has made new tools for chronological analysis available to the archaeological community. These tools are based on Bayesian statistical modeling which makes it possible to formalize chronological reasoning in archaeology. We can now explore the probability space of chronological scenarios deduced from the complex relationships among the chronological data from archaeological sites. We detail the Bayesian paradigm, applying it to the calendric calibration of measurements obtained by dating laboratories (14C, TL/OSL ages, archaeomagnetic measurements, etc.). Then we present the chronological models implemented in software such as BCal or OxCal (calibration, combination, phase, sequence), and the new models or tools implemented in the software RenDateModel, including “fact” or “event” modeling, outlier treatment, and predictive distribution of phases. An application to the chronology of the “Morbihan” Neolithic (Brittany, France), is presented, which demonstrates the potential of the Bayesian approach.

Modélisation statistique bayésienne des données chronologiques

Bayesian statistical modelling of chronological data

Abstract