Although some studies have applied multidimensional analytic methods to crime data (e.g., Short et al., 1963; Nutch and Bloombaum, 1968; Chaiken and Chaiken, 1982), few have examined the assumptions of these methods applied to arrest histories. Arrest histories are a special form of data not ideally suited to standard multidimensional analyses. An examination of the different theoretical assumptions of factor analysis (FA), multidimensional scaling (MDS), and variance centroid scaling (VCS) (a form of correspondence analysis) shows marked differences in what is revealed by the analysis. Computing correlation matrices on data from juvenile delinquent careers for FA and MDS resulted in a nonparsimonious solution in the case of FA and confusing dimensions in the case of MDS. A parsimonious and meaningful set of dimensions emerged from the VCS analysis, dimensions that can be used to organize the contents of delinquent careers. 4 tables, 4 figures, and 37 references. (Author abstract modified)