Cluster analysis has often been proposed as an alternative to factor analysis for the reduction of binary data; however, a review of the clustering literature fails to reveal a method well-suited to clustering binary variables. An ideal clustering method for binary variables would be an efficient, robust method that does not impose a hierarchy of disjointedness on the solution. The method should also allow the researcher a choice of several different measures of association. BINCLUS is a clustering procedure that incorporates all of these features. BINCLUS clusters variables using matrices of Goodman-Kruskal gammas, Pearson r's (phi's), or the quality index q (Cliff, 1979). The clustering solution is presented in the form of an easily interpreted binary cluster membership matrix. In addition, BINCLUS provides a second-order solution that is especially useful when the first-order clusters are not clear-cut. The method has been applied to extensive artificial data and several sets of empirical responses. It is successful in identifying cluster in both artificial and real data. Tabular data and 32 references are provided. (Author abstract modified)