Precision and accuracy in accelerator mass spectrometry (AMS) dating relies on the systematic reduction of errors at all stages of the dating process, from sampling to AMS measurement. With new AMS systems providing much better precision and accuracy for the final stage of the process, we need to review the process as a whole to test the accuracy of reported results. A new High Voltage Engineering Europa (HVEE) AMS system was accepted at Oxford in September 2002. Since then, the system has been in routine use for AMS dating and here we report on our experiences during the first year. The AMS system itself is known to be capable of making measurements on single targets to a precision of better than 0.2% for the14C/13C ratio and better than 0.1% for the13C/12C ratio. In routine operation, we measure known-age wood to a precision of just above 0.3%, including uncertainties in background and pretreatment. At these levels, the scatter in results is no higher than reported errors, suggesting that uncertainties of ±25 to ±3014C yr can be reliably reported on single target measurements. This provides a test of all parts of the process for a particular material in a particular state of preservation. More generally, sample pretreatment should remove as much contamination as feasible from the sample while adding as little laboratory contamination as possible. For more complex materials, such as bone, there is clearly more work needed to prove good reproducibility and insignificant offsets in all circumstances. Strategies for testing accuracy and precision on unknown material are discussed here, as well as the possibilities of one day reaching precisions equivalent to errors of <±2014C yr.
