After discussing the uses of prediction in bail, sentencing, and parole decisions and in treatment-offender matching, the book reviews current research on predictive statistical methods and presents a typology of predictive approaches. This typology is the framework for discussing various predictive factors, including sex, race and ethnicity, age, personality, intelligence, socioeconomic status, criminal history, institutional adjustment drug and alcohol use, the family, peer influences, and situational factors. Each of these factors is assessed according to its validity in predicting criminality and the ethical propriety of using it to predict criminality. The way in which the prediction variables are statistically combined is also analyzed. Following a discussion of variable measurement and sampling, the book reviews some of the statistical methods commonly used to predict criminality, including the Burgess Method, predictive attributive analysis, multiple regression, multidiscriminant analysis, and log-linear techniques. The book concludes with an assessment of the potential value of statistical predictions in the development of criminal justice theories, policies, and practices. 261-item bibliography and subject and author indexes. (Publisher abstract modified)