A failure rate analysis provided estimates of probabilities of recidivism (reincarceration) for any offense at 0.35 for non-Aborigines and 0.80 for Aborigines. Estimates of probabilities of committing further offenses of violence, including further sex offenses, were 0.21 and 0.62 respectively, but these figures are not reliable for predicting offending behavior. A method of covariate analysis for recidivist data demonstrated that prior record, age, and race were crucial determining factors, but little evidence of specialization was found. The authors argue that the evaluation of interventions should proceed on the basis of estimates of violent rather than repeat offending. The authors also discuss the utility of incapacitation strategies, problems of crime prevention, research implications, and policy implications. Tables, figures, references