A general linear approach to the analysis of nonmetric (nominal and/or ordinal) data developed for problems common to the health sciences is extended to the field of political science. After a brief description of the method originally presented by Grizzle, Starmer, and Koch (Biometrics, 1969), two examples are discussed in detail. The first example, using an ordered dependent variable, illustrates an analysis of variance without assumptions of normality. The data are from the University of Michigan 1964 Presidential Election Study. The second example, based on data about the disposition of petty criminal court cases in North Carolina, involves an application where the independent and dependent variables are nominal. The period since 1945 has brought substantial changes in the methodology of political scientists. As the profession has borrowed or -developed techniques suited to its problems, detailed and complex analyses of several variable problems have replaced relatively simple descriptive methods. Thus, analysis of variance, regression and correlation analysis, causal models, principal components, and factor analysis are examples of techniques that have become familiar and useful tools for many political scientists. In spite of the advancing expertise in the discipline, however, a number of persistent problems have repeatedly appeared in analysis that have compromised conclusions or invalidated results. The most persistent of these problems is that throughout the various