This single-case study describes the importance of presenting relevant baseline conditions in planning and evaluating treatment for aggression in a severely brain injured man 1 year after injury. An artificially low natural baseline resulted from staff reluctance to deliver instructions and corrective feedback needed for rehabilitation of self-care skill because these were frequently followed by aggression. A subsequent structured baseline presented these antecedents at the higher rates that were necessary for progress in the patient's rehabilitation. This resulted in an increase in aggressive behaviour, but also gave a more accurate representation of what his behaviour would be like under effective rehabilitation conditions. Intervention was based on data from the structured baseline, and included providing clear expectations, social reinforcement, and decelerative procedures. After reduction of aggression to zero, regular staff were reintroduced and presented instruction and corrective feedback as required. No further aggressive behaviour was noted, and self-care improved so that only minimal assistance was needed. Six months following reintroduction of regular staff, both baseline conditions were replicated. No aggressive behaviour was observed during either, suggesting that maintenance of gains could not be attributed to an artificially reduced rate caused by staff avoiding trigger antecedents.