This article examines past evaluations of government training programs for the economically disadvantaged and offers an agenda for future research. It is found that government training programs are producing modest increases in earnings for adult men and women, but are probably not producing positive effects for youth. Future research must better document links between program-provided training and acquisition of valuable skills and must explore potential returns from increased scale. The recent adoption of random assignment has improved the accuracy of field evaluations but would benefit from an economic theory of evaluation to guide research into increasing training effectiveness.