On behalf of the organising committee, it is a great pleasure to welcome you all to this 14C workshop and indeed to Scotland. The organising committee has already indicated its ability by arranging sunshine for a place and time in which horizontal rain is more common. We plan to build on this initial success by having an outstanding week of good science and pleasant social activity. Scientifically, we have the opportunity firstly to look back and review previous research on the accuracy and precision of 14C dates. Then we will hear and discuss some important new results from the final stage of the present international intercomparison study. Finally, we will discuss and plan mechanisms and procedures by which, in future, we can improve our general level of performance. Paralleling this and of equal if not more importance, we have arranged a social programme which we hope will give us the opportunity to eat and drink well together, see some of the country, get to know each other better and discuss our science informally. So our hopes are high and our welcome sincere.

Many interlaboratory studies have been made in the 14C community at irregular intervals over the past ten years. At times, the results from these studies have been contentious, mostly because of the lack of consistency in their findings. The importance of regular exercises has become particularly acute due to the large number of operating laboratories and the diversity of their methodologies. Hence, we briefly review the studies that have been made in the 1980s, focusing on those in which our laboratories participated. These include the 14C Interlaboratory Comparison in the UK (Otlet et al 1980), the International Comparison (ISG 1982, 1983) and the first two parts of the current International Collaborative Program (Scott et al 1989a, b). The development of each study, its findings and shortcomings, are highlighted in order to assess the concordance of the conclusions.

The success of any intercomparison exercise depends largely on participation and cooperation of a sufficient number of laboratories and the selection of a suitable suite of samples. Unless the latter is satisfactorily devised, the former cannot be guaranteed. The hierarchical nature of this study has necessarily resulted in a far more comprehensive set of sample types than has previously been employed. The exercise was structured to satisfy the following criteria: 1) to enable the participating laboratories to assess the experimental precision and accuracy of the component stages of the dating process; 2) samples should be typical of those routinely dated by the laboratories. This takes on a particular significance in Stage 1 where they should resemble as closely as possible the counting medium; 3) an objective statistical analysis of the results at each component stage of the study.

This report on the third and final stage of the International Collaborative Program concentrates on the analysis of internal and external variability of 14C dates obtained from samples involved in the full 14C dating process. Thirty-eight laboratories took part in this stage with most producing 8 14C dates from 3 sets of duplicate material (eg, wood, shell and peat) and 2 single samples of wood of known ages 190 yr BP apart. From the 3 sets of duplicates for each laboratory, the internal precision of most laboratories was adequate; 6 labs grossly underestimated their internal reproducibility. From the 14C determinations from the 5 distinct samples for each laboratory, we discovered significant systematic biases, often greater than 100 years, in 15 laboratories and even accounting for bias, 12 laboratories had significantly greater external variability than explained by their quoted errors. In total, 23 out of the 38 laboratories in this stage of the study, FAILED to meet these 3 basic criteria for an adequate performance in the production of 14C dates.

A new high-pressure methane-filled counter system for 14C dating was installed in 1986 when the first stage of the International Collaborative Study (ICS) started. Random errors in the new measuring system and in the process of chemical pretreatment and preparation were checked during the three years of intercomparison. Results show that the most important source of error in our laboratory is gas contamination. This causes variation of the count rate to exceed the statistically expected variability. Other sources of error are also discussed and limits of their contributions are given.

Results of the International Collaborative Study show an unexpectedly large scatter of individual dates as well as systematic biases. Very high values of linear correlation coefficients are observed for all results of Stage 2 and for benzene samples of Stage 1. We observed moderate correlations for carbonate samples and the lowest for natural samples of wood and peat of Stage 3. The correlation is practically negligible among results obtained in different stages. The probable reasons for such effects are seen in medium-term changes in the calibration of the counting systems.

The Radiocarbon and Tritium Laboratory of the Rudjer Bošković Institute, Zagreb, participated in the International Collaborative Study (ICS) in all three stages. All measurements were made by proportional counting of methane. We present here a statistical analysis of our results. A comparison with the mean or median values of reported ICS values showed that our results are generally slightly younger.

We discuss here the variability, for our laboratory, in counting for radiocarbon dating of replicate measurements of background and secondary modern standard, duplicate measurements of samples provided by the International Collaborative Study, and replicate measurements of the dilution of the 14C-labeled benzene standard. The variability in the measurements of the International Collaborative Study samples suggest the existence of systematic bias.

The International Collaborative Study involved a wide range of sample materials and ages and, on completion, assessed each stage independently (Scott et al 1989; Aitchison et al 1990). We combine here the three stages of the study and provide an overview of the uncertainties in the dating procedure as a whole and in its component processes. Three key optimal performance indices, related to internal and external precision and to bias, have been defined to allow quantitative assessment of Internal Consistency and External Consistency (Aitchison et al 1990). We believe that these measures provide quantitative descriptions of a laboratory's reproducibility, accuracy and precision.

For the internal consistency, we have defined the Internal Error Multiplier of the quoted error and, for the external consistency of any laboratory relative to an appropriate baseline, we have defined two indices, the Systematic Bias and External Error Multiplier of the quoted error. We have evaluated the three indices over the three stages and have assessed the relative performances of gas counting, accelerator and liquid scintillation laboratories. The quoted errors describe adequately the variability in duplicate results, but there is evidence of systematic biases and underestimation of interlaboratory variability. We have considered the contribution of pretreatment, synthesis counting to the overall variability for each laboratory type. We found that, for liquid scintillation (LS) and gas counting (GC) laboratories, ca 66% of the total variation is due to counting and sample synthesis whereas, for accelerator mass spectrometry (AMS) laboratories, the major sources of variability are the sampling and pretreatment processes.

An investigation of inconsistent Hannover results in the International Collaborative Study (ICS) led to the conclusion that the main reason was contamination of the acetylene used as counting gas with recent and/or fossil carbon by the lithium used for its preparation. Despite the high level of purity of the lithium guaranteed by the producer and storage under argon in cans, different charges were partly covered with contemporary lithium carbonate and fossil oil sometimes was used to preserve the metal. Thorough cleaning of the surface of the lithium rods decreased the contamination but did not remove it entirely, which is evidenced in the wider scatter of the counting rates of various background gases than that of radiocarbon-free tank acetylene. As a result of the high risk of contamination with fossil and/or recent carbon from the acetylene counting gas, the high price of lithium, and the time-consuming preparation, the Hannover 14C Laboratory will use carbon dioxide instead of acetylene as counting gas in the future.

The current intercomparison of data from 14C laboratories reveals significant variability among liquid scintillation laboratories, suggesting that identical samples submitted to different laboratories may yield values that differ by much more than expected on a purely statistical basis. Erroneous dates (recently corrected) by a well-established 14C laboratory give rise to further concern for quality 14C data. Thus, it is incumbent on each laboratory to develop and implement a quality assurance and control (QA/QC) program in order to ensure accuracy of results and to alert lab personnel to problems.

I conducted a high-precision comparison at the 0.2% to 0.3% level with samples supplied by the radiocarbon laboratory of the Quaternary Research Center at the University of Washington (QRC). Four samples with ages ranging from modern to > 50,000 BP were dated in a blind test. The absence of cosmic-radiation background in AMS dating is a major advantage for dating samples > 35,000 BP. The reliability of AMS dates > 35,000 BP depends entirely on understanding the contamination processes. By comparing results with laboratories capable of sample enrichment, such as QRC, it is possible to identify and estimate the intrinsic 14C in the background samples as well as the contamination introduced by sample preparation.

Most radiocarbon ages are readily accepted by researchers in all disciplines. It is recognized, however, that discrepancies appear in the literature. These problems have been highlighted by the International Collaborative Study. The introduction of quality control and assurance techniques used in some laboratories for many years could reduce or eliminate aberrant results. I present here some of the basic considerations of this approach in the processes of conventional radiocarbon dating.

I describe here the establishment and use of statistical control graphs based on the analysis of variance for monitoring the stability of operation of radiocarbon dating counting systems.

I describe here a series of routine self-checks that the Uppsala 14C laboratory performs with all measurements. We estimate all uncertainties in the physical measurement of a sample. We study long-term stability, calculate mean values for oxalic acid and background and compare expected and real statistical distributions of uncertainties. To reduce the risk of bias, the samples from each series are almost exclusively run on the same counter. Some samples are, however, run on two or more counters to check the possible bias to achieve reliable activity comparisons with other laboratories. It is always possible to trace which counter is used, since different number series are used for different counters.

It is clear that radiocarbon researchers take a forward view towards the improvement of accuracy and precision in dating. Unfortunately, archaeologists base much of their research on the published dates produced in the past. Archaeologists and other users of radiocarbon dates should understand the limitations associated with past dates. This article addresses these limitations by looking at a large number of routine radiocarbon dates associated with a block of English tree-ring chronologies, the true ages of which are now known within close limits. My conclusion supports the idea of global multiplication factors as proposed by the International Study Group (1982).

We briefly present here the environmental isotope intercalibration programs of the International Atomic Energy Agency (IAEA). In fact, the IAEA has implemented two parallel programs during the last 20 years: for stable isotopes of light elements and for a radioactive isotope of hydrogen, tritium. This IAEA activity resulted in the preparation of a number of reference and intercomparison materials of various types, now stored in the Agency and available upon request.