BCal help system: The Tutorial

Step D11: Viewing a plot of the estimate of the time elapsed between two parameters

We can now see a plot for an estimate of the length of time elapsed between the end of Layer III and the start of Layer II. Immediately, we note that this range includes zero (which is equivalent to modelling the layers as abutting), but that this is no more likely than many other time periods represented on the plot. In other words, on the basis of the radiocarbon evidence alone, we cannot be sure whether or not there is a hiatus between Layers III and II. This means that we must carefully reconsider the archaeological information that caused us to model the two layers as abutting. Are we really sure that this evidence is strong enough and reliable enough to be allowed to influence the results upon which our final results will be based?

At the fishpond in question, there is in fact little doubt that the two layers are abutting. We are using an abutting relationship to represent a short-lived construction period before aquaculture commenced and so we would be happy to exclude the possibility of a hiatus on the basis of the archaeological evidence even though the radiocarbon evidence suggests that it is possible. Such decisions are often not this clear cut, however, and experts often do not agree exactly how archaeological evidence should be interpreted. This is not problematic when using BCal, it simply means that several models will need to be investigated and reported upon so that all possibilities are given an airing when the site reaches publication stage. Be open minded, consider all the archaeological possibilities and be sure to report on your findings so that others can share them too.

Since BCal uses a simulation approach (MCMC) to obtain the results it gives you there is one other important issue that you must consider (in addition to sensitivity analysis) before you assume that your results are reliable enough to report to others. This issue is convergence checking and we will focus on this now by considering the reproducibility of the results obtained for our fishpond example.