This study examines the effectiveness of alcohol server training legislation to prevent DUI, comparing state-mandated training in two states, training encouraged by laws providing license protection incentives for participating establishments in two states, and two states with no formal statewide system. Dependent variables measuring both implementation and effectiveness include: percent of alcohol servers trained; knowledge, attitudes, and self-reported practices of servers; observed server adherence to responsible practices; estimates of exiting customer intoxication; and time series analyses of DUI crash data. Survey and field data comparing mandatory, incentive, and free market states are presented. Four years of a five-year study has been completed. Mail survey data collection has been completed in six states. Observational, pseudo-patron and interview data collection is completed in Texas and Washington, and is in progress in Alabama, South Carolina, Oregon and New Mexico. Results include: management attitudes are more positive among those who have experienced server training; significantly more servers are trained under mandatory systems than either incentive or free market systems; fewer visibly intoxicated patrons were observed in states with mandatory training; appropriate responses to pseudo-patrons simulating intoxication was somewhat superior in mandatory states but was disappointingly infrequent in all states irrespective of server training systems; and patrons interviewed displayed fewer visible signs of intoxication in mandatory and incentive states than a free market state (Washington).