In case-control studies of breast cancer, cardiovascular disease, and osteoporosis, a woman's ovarian function may be considered as a main exposure, confounding factor, or effect modifier. To evaluate the agreement between self-reported ovarian number following ovarian surgery for nonmalignant disease and medical record reports, we analyzed data from a population-based case-control study, the Cancer and Steroid Hormone Study. Reports on ovarian number after the most recent surgery for a who had been diagnosed with breast cancer and 496 control subjects. We then calculated agreement rates by comparing the number of ovaries indicated by the medical record with the number reported by the woman. Agreement rates on the presence or absence of ovaries exceeded 90% for both case and control subjects. Agreement rates on exact ovarian number exceeded 84% for both groups. Women who had been diagnosed with breast cancer had slightly higher agreement rates than did control women for both presence or absence of ovaries and exact ovarian number. Our data suggest that investigators can rely on self-reported ovarian number as an accurate measure of actual ovarian number in women who have had surgery for nonmalignant conditions.