Almost everyone would like to see the enactment of sound, practical measures to help disadvantaged people get off welfare and find jobs at decent wages, and over the past quarter-century federal and state governments have struggled to develop just such programmes. How do we know whether these vast outlays of money are helping the people they are designed to reach? All welfare and training programmes have been subject to demonstrations designed to test new ideas. This book reviews what we have discovered from past assessments and suggests how welfare and training programmes should be planned for the 1990s. The authors of this volume, each a recognized specialist in the evaluation of social programmes, do more than summarize what we have learned so far. They clarify why the issue of the proper conduct and interpretation of evaluations has itself been a subject of continuing controversy. In part, the problem is organizational, requiring the integrated efforts of social scientists, public officials, and the professionals who execute evaluations. In addition, there is a dispute about scientific method: should evaluators try to understand the complex social processes that make programmes succeed (or fail), or should they focus on imputs and outputs, treating the programmes themselves as "black boxes" whose machinery remains hidden? This book should be important for policy researchers and evaluation professionals, social scientists concerned with evaluation methods, public officials working in social policy, and students of public policy, social work.