The chronology of long Upper Pleistocene loess sequences in Eurasia is based on combined pedostratigraphy and radiocarbon dating of high-quality charcoal. The accuracy of such a chronology depends on the reproducibility and precision of the 14C dates. However, certain dates may show discrepancies with regard to their chronostratigraphic context based on series of coherent dates. In order to evaluate the consistency and variation in the 14C dates obtained from small charcoal pieces, this question was tested on a set of spruce wood remains with well-preserved tree rings found in the Middle Pleniglacial loess-loam sequence of Kurtak (central Siberia). Tree-ring analysis of five fairly large wood pieces from three successive layers, dated to about 30.0, 30.8, and 32.2–32.5 ka BP previously, was done by continuous sampling of 90–150 rings on each wood piece. This enabled direct comparison of the succession of tree rings with the 14C dates. A total of 133 dates was obtained for the five wood pieces. The results show fluctuations in the 14C dates within a time range between 1000 and 2000 yr. Four possible causes for such variation will be discussed herein: (1) internal variability of the AMS dating method; (2) outliers; (3) variations in the 14C background; and (4) external factors such as past atmospheric 14C variations.

The paper is Variability in Radiocarbon Dates in Middle Pleniglacial Wood from Kurtak (Central Siberia) by P Haesaerts, F Damblon, N Drozdov, V Checha, J van der Plicht appearing in Radiocarbon, Vol 56, No 3 (2014).

Tell me again about how reliable carbon dating always is.

Hat tip to Joe Marino