In an attempt to improve the diagnostic value of measuring antibodies to islet cell cytoplasmic antigen, coded sera were distributed to 38 laboratories and results were returned for analysis. Comparison between laboratories revealed that results for some individual sera differed by up to nine doubling dilutions and even within laboratories duplicate samples could differ by six doubling dilutions. By including dilutions of sera it was possible to draw a standard curve for each laboratory and this revealed major variations in shape, slope and intercept. However, using each laboratory's standard curve and converting results to units, a substantial improvement was obtained. The approach described improves standardisation and will permit laboratories to identify poor precision and/or any systematic change in assay performance.