An alcohol-server training program to prevent fetal alcohol syndrome was developed, implemented, and evaluated in a comparison study of public drinking establishments in New Mexico and Oregon. Method: The management and serving staffs of 148 establishments licensed for on-premise alcohol sales in the two states studied were trained to discourage alcohol consumption by pregnant customers. Pre-and post-tests of server responses to pregnant-appearing "pseudo-patron" actors ordering alcohol in experimental (n = 148) and comparison (n = 183) establishments were a key method of evaluating the effi cacy of this intervention. Results: Within-group chi-square analyses compared rates of service refusal at baseline with 1-month, 6-month, and 12-month follow-up points for both the trained (experimental) and the comparison establishments. No differences were found between experimental and comparison establishments at baseline at either site, but signifi cant differences were found for New Mexico at each posttraining measurement point. In Oregon, the refusal rate at baseline increased from 1.5% at baseline to 8.3% at 1 month, which only approached signifi cance. In New Mexico, at baseline the refusal rate was 8.6%, and it rose to 39.2% at 6 months (p < .0001, odds ratio = 6.83) and remained high at 28.2% at 12 months (p < .001, odds ratio = 4.15). No similarly signifi cant gains were recorded at control establishments. Conclusions: Supplemental responsible beverage service training for alcohol servers to aid in the prevention of fetal alcohol exposure can be effective in reducing the serving of alcohol to visibly pregnant women, with robust effects continuing over the subsequent year in the New Mexico establishments.