The model was applied both to the entire sample of offenders and separately to rapists and child molesters to determine whether different causal paths exist within these broadly defined types. The study subjects were 125 convicted male sexual offenders committed to the Massachusetts Treatment Center. The primary data source for each subject was his clinical file, which included all information gathered during the subject's evaluation and commitment periods at the Center. The prediction of aspects of criminal career from life history variables included the following steps: (1) conceptual delineation of a preliminary structural model, with its attendant assumptions and specification of subsets of criterion and predictor variables; (2) reduction through principal components analysis of each of these initial variable subsets into a manageable number of cohesive, stable constructs; and (3) application of a series of simultaneous multiple regression analyses using item groups derived from the reduction phase. In the path analysis for the combined sample, the strongest association found was between adolescent social and academic incompetence and severe psychopathology/sexual pathology in adulthood. The path analytic solution for rapists was reasonably similar to the combined model, while the path analytic solution for the subsample of child molesters differed in a number of ways from the combined sample. The single most important and distinctive feature was the network of paths leading to frequency of criminal offense. Longitudinal relationships were present in family/parental pathology (a low incidence of parental substance and child physical abuse); sexual deviation in the family, childhood, and adolescence; and antisocial behavior and severe psychopathology and sexual pathology. Eighty-three references are provided, along with illustrations of the path analysis.