Population studies often estimate mammography use using women's self-reports. In one North Carolina county, we compared self-report surveys with a second method--counting mammograms per population--for 1987 and 1989. Estimates from self-reports (35% in 1987, 55% in 1989) were considerably higher than those from mammogram counts (20% in 1987, 36% in 1989). We then confirmed 66% of self-reports in the past year. Self-reported use is more accurate regarding whether a woman has had a mammogram than when she had it, but self-reports accurately measure change over time.