Subkoviak suggested a technique for estimating both group reliability and the reliability associated with assigning a given individual to a mastery or non-mastery category based on a single test administration. Two assumptions underlie this model. First, it is assumed that had successive test administrations occurred, scores for each individual would be independent over repeated testings. Second, it is assumed that scores for each individual are binomially distributed. According to Subkoviak, these assumptions are frequently violated. The purpose of this investigation was to compare the results obtained from his model, for both a simple and compound binomial distribution of test scores, with classical estimates of reliability: Guttman, and KR-20 and KR-21. An inverse relationship existed between the two types of estimates. For students who exhibited no learning on successive test administrations, a comparison of the reliability estimates indicates that the Subkoviak estimates are less sensitive to a small observed score variance. However, the model provides useful reliability estimates in situations where the classical measures are clearly inappropriate. The model gave reasonable reliability estimates even though its assumptions were not met. (Author/CP)