In place of the existing institutionally based clinical assessment model for identifying inmates for Community Correction Center (CCC) placement, the PDC wanted to implement a risk management classification system that maximizes community releases in a safe and effective manner, affords a consistent assessment process, and provides a vehicle for policy level population control. The core component of the community release classification system that was developed for the PDC is a RAI designed to identify those prisoners who constitute a good risk for placement in community corrections programs. A central task in designing an effective risk classification system is perceived to be the development of an objective point-scale classification or prediction instrument. The RAI that was developed used a presumptive, quantitative point-scale to identify low-risk prisoners for CCC release and high-risk prisoners for denial. Issues related to the external validity that often are neglected or glossed over in social science research are examined. The article addressed two validity issues that surfaced during the course of the study. The first related to sampling bias, as examined through predictive validity. The second is a conceptual validity problem involving the choice among alternative outcome measures. In the concluding discussion it is argued that the experiences in this study reinforce the need to be rigorous about sampling and the need to recognize and value different ways of knowing.