Radiocarbon dates as data: quantitative strategies for estimating colonization front speeds and event densities

Abstract

Archaeological analysis of large-scale prehistoric population history requires us to estimate rates of spatial spread during dispersals, and rates and magnitudes of temporal contraction during crashes. Using OxCal’s MCMC sampling routine, I introduce and demonstrate a simple and easily implemented method of estimating front speeds that takes due account of the uncertainty in the archaeological data (in both dates and distances), and argue that this method is more appropriate than those most often used in front speed estimation at present. I also propose a simple and easily implemented method of estimating event densities as a demographic proxy, as an alternative to summed calibrated probability distributions. I argue that this alternative is a significantly better technique, and show that its use also enables us to identify individual archaeological dates that are exerting particularly strong influence on the results, so that we can efficiently allocate our attention when assessing the possible effects of exogenous sampling uncertainty. To illustrate these methods I re-analyse two published datasets relating to the early Paleoindian colonization of North America. My results with the new technique indicate that even with a very noisy dataset, there was clear evidence in the framework of the INTCAL04 calibration curve for a drastic reduction in archaeological event densities following the Younger Dryas onset, followed by a prolonged period of reduced human activity, and a possible renewed phase of rapid growth after the Younger Dryas termination and onset of the Holocene. However, the revised estimate of the Younger Dryas marine reservoir offset in the INTCAL09 calibration curve for ∼12,550–12,900 cal BP changes the picture significantly, by flattening the peak in Clovis-age events and pushing it forward in time into the early Younger Dryas itself.