Relying heavily upon historical records to build archaeological chronologies, many Korean historical archaeologists question the utility of radiocarbon dating. To challenge this tendency, we test a model based on historical records by analyzing radiocarbon dates. We dated two bulk samples from Hongryeonbong Fortress II, Seoul, South Korea, which is known historically to have been occupied between A.D. 500 and 551. By dividing bulk samples into smaller aliquots, we dated 40 subsamples and statistically estimated consensus dates. The consensus dates do not support expectations of the historical model, as they do not lie within the temporal range provided by historical records. We propose an alternative model that explains both the radiocarbon dates and the historical records. It is suggested that radiocarbon dating can be successfully applied to the reconstruction of historical periods and is a way of mitigating possible biases of models based solely on historical records.

Abstract In interpreting radiocarbon dating results, it is important that archaeologists distinguish uncertainties derived from random errors and those from systematic errors, because the two must be dealt with in different ways. One of the problems that archaeologists face in practice, however, is that when receiving dating results from laboratories, they are rarely able to critically assess whether differences between multiple 14C dates of materials are caused by random or systematic errors. In this study, blind tests were carried out to check four possible sources of errors in dating results: repeatability of results generated under identical field and laboratory conditions, differences in results generated from the same sample given to the same laboratory submitted at different times, interlaboratory differences of results generated from the same sample, and differences in the results generated between inner and outer rings of wood. Five charred wood samples, collected from the Namgye settlement and Hongreyonbong fortress, South Korea, were divided into 80 subsamples and submitted to five internationally recognized 14C laboratories on a blind basis twice within a 2-month interval. The results are generally in good statistical accordance and present acceptable errors at an archaeological scale. However, one laboratory showed a statistically significant variance in ages between batches for all samples and sites. Calculation of the Bayesian partial posterior predictive p value and chi-squared tests rejected the null hypothesis that the errors randomly occurred, although the source of the error is not specifically known. Our experiment suggests that it is necessary for users of 14C dating to establish an organized strategy for dating sites before submitting samples to laboratories in order to avoid possible systematic errors.

Old wood effect refers to an error derived from difference between archaeological target date and radiocarbon dates of wood samples. Schiffer (1986) suggested that this effect would lead to radiocarbon dates older than target dates, and that to solve this problem, it is necessary to understand the r...