End points for trials promoting cancer screening are often based on self-reported screening behavior. This study was designed to evaluate and optimize the reliability of a computer-assisted telephone interview for collecting self-reported colorectal cancer screening behavior. Cases who had received a fecal occult blood test (FOBT), flexible sigmoidoscopy, and/or colonoscopy, and controls who had no record of colorectal screening were identified among 40-75-year-old members of the Denver Kaiser Permanente Health Care Program and were contacted by telephone. Sensitivities and specificities of self-reported screening were calculated by comparison of subjects' recall with Kaiser Permanente records. The questionnaire was revised based upon results of the pilot phase of the study. Using the revised questionnaire, the sensitivity of self-reported screening was 96.2% for the FOBT, 94.9% for flexible sigmoidoscopy, 88.7% for colonoscopy, and 96.2% for either endoscopic screening test. The specificity of self-reported screening was 85.9% for the FOBT, 92.2% for flexible sigmoidoscopy, 96.8% for colonoscopy, and 92.0% for either endoscopic screening test. No marked differences in the accuracy of the self-reports were detected as a function of gender, age, ethnicity, or family history of colorectal cancer of the participants. Self-reports of colon cancer screening behavior can be reliably used as end points for intervention trials when carefully phrased questions are used.