The extent to which the Nuffield Risk Prediction System yields accurate and operationally useful release risk predictions for aboriginals is assessed using a range of traditional measures, such as the Index of Predictive Efficiency and the Relative Improvement Over Chance, and new methods, such as the Dynamic Parole Success Rate and the Dynamic Relative Improvement Over Chance. The results of efforts to develop and test alternative risk prediction systems are presented for systems based on simple Burgess scoring techniques and linear regression models. Alternative risk prediction systems are compared with the Nuffield system in terms of factors each system finds useful in predicting release risk and in terms of their predictive accuracy and usefulness. Statistical prediction systems are also evaluated in relation to the system used by Canada's Parole Board to grant or deny parole. Each chapter of the report concludes with a brief summary of major findings and promising directions for further investigation. Appendixes contain supporting statistical tables. Footnotes, tables, and figures