In designing a system for measuring the performance of corrections agencies, evaluators should take into account what performance dimensions to include, whose needs are to be served, how the system affects as well as describes performance, and how different measures can be combined to summarize performance. Criteria for potential performance measures include reliability, cost and ease of data collection, comparability, clarity, and timeliness. When interpreting performance measurements, users should keep in mind distorting factors such as program dropout rates, learning curves, self-correcting cases, and age of the data. If no single performance model meets the criteria, an eclectic model should be developed; the report provides examples of models measuring the extent of postrelease criminal activity and legitimate wage attainment. The report also describes how to select statistical techniques for evaluating performance models, and suggests directions for further research. Charts, tables, and a bibliography are included. Appendixes provide organizational data for corrections institutions by State (1979), a description of major corrections positions, definitions of performance measures, and the study models.