All women diagnosed with breast cancer before age 36 in eleven geographical areas in Britain were included in a case-control study designed to investigate the relation between oral contraceptive (OC) use and breast cancer risk. 755 cases each with a matched control were interviewed in their homes by trained interviewers. General practitioner notes and family planning clinic records were abstracted and combined with the interview data. There was a highly significant trend in risk of breast cancer with total duration of OC use with relative risks of 1.43 (95% confidence interval 0.97-2.12) for 49-96 months use, and 1.74 (95% confidence interval 1.15-2.62) for 97 or more months use. The relative risks were similar for use before and after first full-term pregnancy. There is some evidence that OCs containing less than 50 micrograms oestrogen have a lower risk associated with their use than higher oestrogen dose OCs and that there may be some protective effect of use of progestagen-only pills. These apparent differences between OC types are only marginally significant statistically and need further investigation. Extensive investigations of possible bias using data abstracted from general practitioner notes for both interviewed and non-interviewed cases and controls demonstrate that biases cannot explain these results although some relative risks may be slightly exaggerated. There is, however, no support for these findings in national breast cancer registration rates which are not increasing.