This study has been performed to test the hypothesis that different oestrogen receptor beta (ER ) splice variants may be important determinants of clinical parameters, including outcome, in post-menopausal women with breast cancer receiving adjuvant endocrine treatment but no chemotherapy. Splice variants ER 1, ER 2 and ER 5 have been analysed by semi-quantitative RT-PCR in a cohort of 105 patients with primary breast cancer. Clinical correlates included age, grade, size, nodal status, ER , progesterone receptor, Ki67, relapse-free survival (RFS) and overall survival (OS). Seventy per cent of cases were ER 1 positive, 69% ER 2 positive and 70% ER 5 positive. Within the cohort, 47% were positive for all three variants while 10% were negative for all three. ER 1 exhibited no discernible relationship with disease outcome. ER 2 and ER 5 expression was significantly associated with better RFS (P,0·005), and ER 2 with better OS (P=0·0002). In multivariate analysis, ER 2 (P=0·006), nodal status and the level of Ki67 expression were independent predictors for RFS while ER 2 (P=0·0008) and Ki67 status were independent predictors for OS. In the ER -positive cases, or in the subset of those receiving adjuvant tamoxifen, ER 2 was significantly associated with good RFS (P,0·0005) and was the only independent marker of OS. We conclude that precise identification of splice variants of ER are more important assessors than is ER 1 alone of the biological status of individual breast cancers, and hence in predicting their response to endocrine therapy.