ESTIMATING THE DURATION OF ARCHAEOLOGICAL ACTIVITY USING 14C DETERMINATIONS

Summary. One of the questions that archaeologists have always hoped to be able to answer using radiocarbon dating is: ‘How long did this activity last?'. It has for some time been accepted that this question cannot adequately be addressed by simply calibrating single radiocarbon determinations. Rather, it is necessary to find means for coherently relating such determinations to one another and to the archaeology from which they came. In addition, most archaeologists are aware that estimation of duration is likely to be greatly affected by the presence of any aberrant determinations (‘outliers') and that they should, therefore, be allowed for in the estimation procedure. In this paper we look at the issues which relate to solving problems of this type and highlight some difficulties associated with one currently available method for approaching them. By adopting a Bayesian approach to estimating duration, it is possible to provide an explicit and coherent framework within which such investigations can take place and whereby each specific problem can be considered in its own right. In archaeological terms, this paper has, we hope, clarified the problems involved in estimating the duration of a period of time represented by a series of unordered radiocarbon determinations, and in providing estimates of the calendar dates for its beginning and end. These estimates take in any archaeological evidence available, for example, termini ante/post quos (here the arrival of the Spanish in Peru forms a terminus ante quern), and may be repeated when new evidence becomes available. For the first time, we provide an objective method for identifying outliers in the data and documenting the effect of their removal. We have given an example for the Peruvian Chancay culture to demonstrate the potential of this approach.