This study was undertaken to determine the reliability of self-reported sexual behavior using the test and retest technique when used with self-reported sexual behavior. The subjects were 116 asymptomatic homosexual men who participated in another study (an examination of behavioral and demographic determinants of HIV antibody status). The subjects were asked to complete two questionnaires. The first contained demographic and sexual behavior questions. The second, administered an average of 6 weeks later, used a subset of the questions in the first questionnaire. The reliability of the test-retest procedure was measured by the Kappa statistic, which assesses the proportion of agreement between two data items, accounting for the amount of agreement expected by chance. The highest degree of reliability as measured by Kappa was found with demographic information, smoking history, and sexual orientation. Self-reported sexual behaviors for the previous 6 months generally had the next highest degree of reliability as measured by Kappa. Questions examining change over the previous 5 years had the lowest reliability. Behavior changes during the time between questionnaires, subjectivity of the answer categories, and social desirability of the answers are three factors that may result in a lack of reliability in this self-reported sexual behavior questionnaire. This raises methodological concerns about the measurement of behavioral risk factors for AIDS and the ability to assess meaningfully subjective reports of behavioral change.