This article describes the use of an observation system to measure middle school staff practices, environment characteristics, and student behavior in the school common areas. Data were collected at baseline from 18 middle schools participating in a randomized controlled trial of school-wide Positive Behavior Support. The observations were reliable and showed sensitivity to differences between school settings and between schools. Multilevel models with students nested in schools were used to examine the associations of staff practices and the school environment with student behavior. Less effective behavior management and more staff criticism, graffiti, and percentage of low-income students were associated with student problem behaviors. Greater use of effective behavior management and positive attention, and fewer low-income students were associated with positive student behavior. The use of data-based feedback to schools for intervention planning and monitoring is illustrated. Implications for school-wide efforts to improve student behavior in middle schools are discussed. (Contains 6 tables and 1 figure.)