Background. Mammography self-report is used to monitor screening and evaluate intervention trends; however, few studies have examined reliability. Methods. Reliability of self-reported lifetime number of mammograms, most recent mammogram date, and predictors of reliability were assessed using data from Project H.O.M.E. The study population was 2494 women 52 years and over, listed in the U.S. National Registry of Women Veterans, with no history of breast cancer, who completed both baseline (2000–2002) and year 1 (2002–2003) surveys. Results. Reliability of lifetime number of mammograms was 60.9% for exact consistency and 79.9% for consistency within one mammogram. Thirty-five percent was exactly consistent in reporting mammogram date; 55.6% was consistent within 3 months. Completing both surveys by mail and reporting fewer lifetime mammograms at baseline were positively associated with consistency of reporting lifetime number. White race/ethnicity, having a Bachelor's degree, reporting a healthcare provider's recommendation for a mammogram, having a screening mammogram, completing both surveys by mail, and being in the maintenance or action stages of change were associated with consistency in reporting date. Conclusions. Reliability varies with the measure of self-reported mammography. Likewise, predictors show different patterns of association with different definitions. Our findings call attention to the need for explicit definitions and measures of mammography use.