The relationship of women's sociodemographic characteristics, knowledge, attitudes, and beliefs to breast self-examination (BSE) practice is not clear. We therefore studied these variables among older women at risk for developing breast cancer to determine which might be associated with the sensitivity, specificity, and frequency of BSE practice. We interviewed 300 women 40 to 68 years of age and measured BSE sensitivity and specificity using manufactured silicone breast models containing lumps. Of 54 variables and 10 scales examined univariately, six were associated with BSE sensitivity, one was negatively associated with specificity, and 10 were associated with frequency. No variable was associated with more than one component of BSE practice, and BSE frequency was not associated with BSE sensitivity or specificity. Using multivariate analysis, BSE sensitivity was best explained by type of employment, health interest, and perceived vulnerability to breast cancer, which accounted for approximately 16% of the variance. BSE frequency was best explained by intention to perform BSE, knowing how to perform BSE, using the correct method of BSE, self-confidence in the ability to perform BSE monthly, and self-confidence in the ability to find small lumps. These variables accounted for 27% of the variance. Sociodemographic characteristics, knowledge, attitudes, and beliefs poorly predicted how accurately women practiced BSE but somewhat better predicted how often women practiced BSE.