The study objective was to compare three methods for teaching breast self-examination. The study design was a randomized controlled trial with factorial design. The patients included continuing-care patients from 40 to 68 years of age from a general medicine group practice in a university hospital. Of 456 eligible women, 156 refused participation, 300 were randomly assigned, 269 completed assigned interventions, and 260 completed the post-test 1 year later. One third of patients received nurse instruction stressing tactile skills (Mammacare group) (Mammatech Corporation, Gainesville, Florida); one third, traditional nurse instruction emphasizing technique (traditional group); and one third, no nurse instruction (control group). Half of each group received physician encouragement. One year later, women in the Mammacare group found more lumps (mean, 57%, 95% CI, 54% to 60%) in manufactured breast models than did those in the traditional (mean, 47%; CI, 44% to 51%) and control (mean, 45%, CI, 42% to 48%) groups. Lump detection specificity was unaffected. Self-reported examination frequency rose in all groups, to 5.1 times per 6 months in the traditional group, 4.2 in the Mammacare group, and 3.9 in the control group. Physician encouragement did not improve sensitivity, specificity, or overall frequency. Women in the Mammacare group used more of seven examination techniques (4.9) than did those in the traditional (3.9) or control (3.2) groups (P < 0.001). No group increased health care use or reported more overall worry about breast cancer. In conclusions, mammacare instruction resulted in more long-term improved lump detection and examination technique use than did traditional instruction or physician encouragement. Breast self-examination instruction should emphasize lump detection skills.