The use of different measures of similarity, clusterization criteria, and theories in cluster analysis makes it difficult to compare interpretations of a single data set. The authors present some methods as well as software tools for comparison of cluster partitions. The package CLUSTER can solve the following basic problems of cluster analysis: (1) How many clusters are present in a given set of objects X if its cluster structure is unknown? (2) If different partitions of X are obtained for a given value of c (the number of clusters), using some clustering methods, then which of these partitions is the most valid? and (3) Are there good clusters for multiple values of c? It can also be used to study the cluster structure of X using different metrics for the calculation of the distance between the objects as well as the dependence of different final partitions on the functional's parameters. >