A new, powerful and inexpensive test for the stability of reference materials characterised by the results of proficiency tests is demonstrated. Participant-blind re-issue of the same materials provides the scheme organiser with two independent estimates of the location and variance of the results. Because many participant laboratories are involved, the location estimates will not be confounded with changes in the execution of analytical procedures. The method is illustrated by its application to foodstuff test materials almost certain to be stable, with periods between the re-issues of between seven and 52 months. Apart from one discrepant result (out of 41), no statistically significant instability was demonstrated.