The increased use of criterion-referenced statewide testing programs is an outgrowth of the need for more diagnostic information for planning and decision making than is provided by norm-referenced programs. There remains, however, a need for state agencies to compare the results of local districts to a variety of comparison groups for the purpose of identifying where the greatest needs lie. This paper deals with nonparametric techniques for the comparison of matrices of criterion-referenced scores (rather than the comparison of means). Specific examples include chi square, the median test, rank correlation, the Wilcoxon tests, Kendall's W, and others. (Author)