Objective: To evaluate the implementation of a controlled, 6-week, environmental and educational intervention to improve dietary intake and body composition, and to study the association of implementation fidelity with diet and body composition outcomes. Design: A process evaluation documented participation, dose of nutrition education delivered, participant satisfaction, fidelity and completeness of the food environment intervention implementation, and context through observations and interviews with staff and residents. Intervention sites were scored and categorized as high or low participation and implementation and compared on essential elements of the food environment and on diet and body composition outcomes. Setting: Six urban residential drug-treatment facilities in Upstate New York. Subjects: Fifty-five primarily black and white men in residential drug-treatment programmes. Results: Participants were exposed to 94 % and 69 % of the educational and environmental elements, respectively. High implementation sites were significantly more likely to provide water and 100 % juice, offer fruit or vegetable salad, offer choices of fruits and vegetables, and limit fried foods. Mixed-model analysis of covariance revealed that participants in the high participation and implementation sites reported greater reductions in total energy, percentage of energy from sweets, daily servings of fats, oils and sweets, and BMI over the intervention period. Participants in low participation and implementation sites reported greater reductions in percentage of energy from fat. Differential implementation of environmental elements limited the intervention impact. Conclusions: These findings document the contribution of changes in eating environments to facilitate dietary behaviour change in community residential substance-abuse settings.