A note in the 1990 ANTIQUITY volume dealt with four issues crucial to the successful use of radiocarbon in archaeology (Bowman & Balaam 1990): selection and characterization of material and context; determination of the radiocarbon result and error term; interpretation and publication; and strategic resourcing. Since then much has been published, particularly on quality control of radiocarbon measurements ('determination'), and on the calibration of radiocarbon results ('interpretation'). Here is an update.
Determination -- accuracy and precision
Discussion of laboratory accuracy and precision in radiocarbon dating, dull for reader and writer alike, is seemingly well rehearsed. Nevertheless, the difference between these two types of error is central to any understanding of quality control in radiocarbon dating, and to any evaluation by the user of advice on how to use error terms. Accuracy determines how close the experimental result is to the true value, and precision determines the closeness of replicate measurements to each other: accurate results can be imprecise and vice versa (see, for example, Bowman 1990: figure 15). The error quoted by a laboratory is, in theory, the precision (the uncertainty due to random variability); but most radiocarbon laboratories estimate their precision rather than determine it by replication because of the resources involved in repeating measurements. In estimating there is the danger that not all potential sources of random error are included.
A proposed quality assurance protocol for radiocarbon (Long & Kalin 1990) covers all aspects of the dating procedure from documentation of samples received to issue of results, but it may not be obvious to the user whether a radiocarbon laboratory is providing quality control as far as accuracy and precision are concerned. Any laboratory should be able to provide information on its accuracy, either as evaluated in an intercomparison study with other laboratories or against known age samples. Much the most relevant samples of known age for archaeological applications of radiocarbon are dendrochronologically dated tree-rings from the same sequence as used for one of the high-precision calibration curves: laboratory accuracy demonstrated with such known age samples means users can be confident that their radiocarbon results can be calibrated against accepted curves. Laboratory participation in an intercomparison, on the other hand, is 'blind', i.e. the laboratory does not know the age of the samples before it submits its results for scrutiny. Furthermore, with the publication of the International Atomic Energy Agency (IAEA) intercomparison (Rozanski 1991) and Third International Radiocarbon Intercomparison ('TIRI', Scott et al. forthcoming) the results are no longer anonymous: users will be able to see for themselves how well participating laboratories fared relative to the consensus results from these studies. Anonymity, not unnaturally, suggested that laboratories had something to hide, but it also, perhaps surprisingly, provided a potential safeguard of benefit to the user. A laboratory on finding it has a systematic error (i.e. that it is producing inaccurate results) can, and should, investigate and rectify the problem. Its position on a 'league table' at any given time is therefore not immutable, and a correction made to the results of that laboratory produced at another time may well introduce, rather than remove, an error. Only the laboratory itself can tell the user whether or not it has, or has had, a systematic bias, and when.
This is not to say that intercomparison studies are not useful -- they are -- but largely to provide laboratories with independent checks of in-house procedures and, in the future, may well be required procedures whereby laboratories can be 'quality certified' to enable them to tender at least for government funded contracts. They also give the user a clear statement that the laboratory is committed to quality control: participation in an intercomparison study is no small use of limited resources. …