According to the introduction, classification methods have two goals: to build new classifications and to assign unknown cases to existing classifications. Confusion concerning purpose and roles has considerably weakened the value of many traditional offender classification systems. The link between purpose and method has become more important because of the emergence of a vast range of quantitative methods, although their multidisciplinary origins mean that some concepts and logic may be inconsistent with criminology data or theory. The author contends that while no single method is best for all purposes, new approaches may dramatically upgrade the quality of classifications. The article emphasizes the urgent need for a framework for creating and validating classifications. Interacting stages for choosing a content domain, constructing the classification, checking internal validity, and establishing external validity should be included in this framework. Over 100 references. (Author abstract modified)