Animal studies have, in general, been supportive of a protective effect of fish and fish (n-3) PUFA against breast cancer risk; but the epidemiologic evidence of such a relationship is limited. Case-control and cohort studies have rarely shown significant associations. The association between total fish intake and the effect of fat content and preparation method of the fish, in relation to the incidence rate ratios of breast cancer, were investigated among postmenopausal women. We also investigated the effect of fish intake with respect to estrogen receptor expression of breast cancer tumors. A total of 23,693 postmenopausal women from the prospective study "Diet, Cancer and Health" were included in the study. During follow-up, 424 women were diagnosed with breast cancer. The incidence rate ratio (IRR) and 95% CI per each additional 25 g of mean daily intake of fish were 1.13 (CI, 1.03-1.23). Analysis of fatty fish gave IRR of 1.11 (CI, 0.91-1.34), and the result for lean fish was 1.13 (CI, 0.99-1.29). When fish intake was stratified into three types of preparation methods, the IRR for fried fish was 1.09 (CI, 0.95-1.25), for boiled fish 1.09 (CI, 0.85-1.42), and for processed fish 1.12 (CI, 0.93-1.34). The IRR per additional 25 g of mean daily intake of fish was 1.14 (CI, 1.03-1.26) for estrogen receptor-positive (ER+) and 1.00 (CI, 0.81-1.24) for estrogen receptor-negative (ER-) breast cancer. In conclusion, this study showed that higher intakes of fish were significantly associated with higher incidence rates of breast cancer. The association was present only for development of ER+ breast cancer.