Main

As Mellars points out1, ultrafiltration of bone samples can yield final age determinations that are significantly different from ages derived by traditional methods1,2. The latest results from ultrafiltration show that the ages originally reported from archaeological contexts should be considered as only minimum estimates. Although Mellars says that bones are the most widely available materials for dating1, most (75%) of the ages plotted in Fig. 3 of ref. 1 are based on charcoal (P. Mellars, personal communication). As with bone, important developments have taken place in the preparation of charcoal for radiocarbon dating. Advances in the pretreatment and graphitization of charcoal3 have extended its dating range to at least 55,000 yr before present (BP) and make it probable that many of the original age determinations for charcoal are minima. Unfortunately, Mellars reconstructs routes and rates of modern human dispersal across Eurasia from published ages of samples that had not been pretreated using the new methods.

Mellars also erroneously refers to NotCal04 (ref. 4) as a 'calibration'; however, NotCal04 is a summary of data sets that extend beyond the 26,000 calibrated years BP limit of the internationally accepted radiocarbon curve, IntCal04, and is not a calibration curve (hence the name4). The records that contribute to the NotCal04 comparison curve deviate by up to several millennia from one another5,6,7,8 (Fig. 1). Significant variability in atmospheric radiocarbon content is also indicated by a recently produced data set9. Some of this variability is due to uncertainties in the dating of individual records, but features common to all records will include, for example, short-lived excursions in the Earth's magnetic field8.

Figure 1: Selected radiocarbon comparison curves for the period 25,000–60,000 calibrated years before present (BP).
figure 1

The graph shows the NotCal04 best-fit curve4, and Cariaco Basin sediments5, Lake Suigetsu varves6, the Bahamas speleothem7, and tropical Atlantic and Pacific corals9 radiocarbon data sets. Age ranges at 1σ uncertainty (including those for the NotCal04 curve) are plotted only on the 'radiocarbon years BP' axis.

Full size image

Atmospheric 14C variability has not followed a simple, smooth pattern, as suggested by Mellars1. Instead, smoothing took place during the statistical analysis of these data sets to develop the NotCal04 mean best-fit line. By using the mid-point of the mean best-fit line, Mellars artificially improves the apparent precision of calibrated ages in his Fig. 3; even 'infinitely' old ages are reported with improved precision, whereas calibration almost invariably results in age ranges that are significantly larger than the radiocarbon measurement error10.

We appreciate that Mellars' review was restricted to radiocarbon dating, principally of bone, but it is recommended practice that multiple methods and materials should be investigated to avoid any possible pitfalls that might be associated with a single technique or sample type. Several numerical-age methods are now available for the dating of bone11, sediments and artefacts12 that complement one another and add much-needed rigour to radiocarbon chronologies. We recommend that the accepted nomenclature for ages should be strictly followed to avoid confusion13: radiocarbon ages should be reported as 'BP', time before present (where 'present' is AD 1950); calibrated ages as 'cal. yr BP'; and ages from alternative dating methods in calendar years.

Inevitably, researchers will need to compare finds older than 26,000 cal. yr BP to other records measured on calendar timescales. When ages are 'calibrated', it must be specified which curve has been used. Meanwhile, radiocarbon ages should be reported in their uncorrected form for recalculation when an accepted calibration curve is developed.

Only by taking these considerations into account can we hope to develop robust models for human dispersal.