A Monte Carlo approach is proposed, using the Statistical Analysis System (SAS) programming language, for estimating reliability coefficients in generalizability theory studies. Test scores are generated by a probabilistic model that considers the probability for a person with a given ability score to answer an item with a given difficulty parameter correctly. Three types of reliability-like coefficients available in generalizability theory are considered: (1) the generalizability coefficient that is the analog of the classical reliability coefficient; (2) the index of dependability, appropriate for criterion-referenced testing; and (3) the classification reliability index, for the reliability of classification decisions based on a cutting score. The proposed approach is illustrated for a single-facet crossed design but it works for higher levels of crossed or nested generalizability designs. The SAS program allows flexibility and control on factors such as the type of probabilistic model, the type of ability score and difficulty parameter distributions, the location of the cutting score, and the amount of information provided by each item. An appendix presents the SAS program for Monte Carlo reliability estimations. (Contains one table and eight references.) (SLD)