BACKGROUND Epidemiological evidence which suggests that prolonged breastfeeding protects against breast cancer has accumulated in recent years. Issues with regard to the timing of breastfeeding and effect modification by correlates of breastfeeding and other risk factors of breast cancer remain unresolved. METHODS A population-based case control family study of breast cancer among women diagnosed by the age of 50, conducted in two geographic areas in Germany, was used to evaluate the effect of breastfeeding on risk of breast cancer. RESULTS Among parous women in this study (553 cases, 1094 age-matched population controls), having ever breastfed a child for at least 1 month did not confer protection (odds ratio of 0.9 and 95% confidence interval (CI) 0.8-1.2). However, risk of breast cancer significantly decreased with increasing duration of breastfeeding (p for trend = 0.01) and the estimated relative risk was 0.6 (95% CI 0.4-0.9) for 13-24 months of cumulative breastfeeding and 0.5 (95% CI 0.3-1.1) for 25 months or more. Risk was less related to number of children breastfed than to increasing average length of breastfeeding per child (p for trend = 0.03). CONCLUSIONS The reduction in risk associated with duration of breastfeeding was not primarily due to breastfeeding the firstborn and more evident in women who were older (> 25 years) when they first breastfed and among women who experienced a recent full-term pregnancy. Risks were modified somewhat by a first-degree family history of breast cancer whereby a greater reduction in risk per additional month of breastfeeding was observed among women with a family history than those without (0.9 vs. 1.0). The study results support a protective role of prolonged breastfeeding against the development of breast cancer in predominantly premenopausal women in Germany.