Bayesian analysis of ESR dates, with application to Border Cave

Methods for Bayesian statistical analysis of stratigraphically related radiocarbon dates have been in use for over a decade. This paper extends these techniques to stratigraphically related ESR dates, allowing estimation of the dates of events not directly dated. A hierarchical model of the uncertainties in ESR dating is developed, to account for the correlation of error terms between samples. Using the new method, an analysis is made of the dating at Border Cave, Kwa Zulu Natal, South Africa. The results for individual dates and the dating of layer boundaries are more precise than previously obtained. The hominid fossils BC1 and BC2 are placed at either 71–91 ka (95% highest posterior density—hpd) or 152–171 ka (95% hpd) depending on the stratigraphic provenance assigned. BC3 is dated to 66–90 ka and BC5 to 61–72 ka (both 95% hpd). The estimated duration of the Howieson's Poort industry at Border Cave is demonstrated to have significant uncertainty, and the ESR dates, even with the increased precision of this analysis, are unable to decide between hypotheses that the industry lasted 10 ka and that it lasted 20 ka.