To assess the variability of oestrogen receptor (ER) testing using immunocytochemistry, centrally stained and unstained slides from breast cancers were circulated to the members of the European Working Group for Breast Screening Pathology, who were asked to report on both slides. The results showed that there was almost complete concordance among readers (kappa=0.95) in ER-negative tumours on the stained slide and excellent concordance among readers (kappa=0.82) on the slides stained in each individual laboratory. Tumours showing strong positivity were reasonably well assessed (kappa=0.57 and 0.4, respectively), but there was less concordance in tumours with moderate and low levels of ER, especially when these were heterogeneous in their staining. Because of the variation, the Working Group recommends that laboratories performing these stains should take part in a external quality assurance scheme for immunocytochemistry, should include a tumour with low ER levels as a weak positive control and should audit the percentage positive tumours in their laboratory against the accepted norms annually. The Quick score method of receptor assessment may also have too many categories for good concordance, and grouping of these into fewer categories may remove some of the variation among laboratories.