This study examined whether direct, interval-by-interval measures of treatment integrity would make it possible to distinguish whether equivocal intervention results could be attributed to the intervention itself, or to poor implementation. Josh, an eight-year-old 3rd grader, performed at or slightly above his peers' academically, yet engaged in problem behaviors (yelling, throwing objects, slamming his desk into a peer's desk) on a daily basis. A functional behavioral assessment (FBA) identified these behaviors were maintained by gaining attention (positive reinforcement) and escaping from certain assignments (negative reinforcement). A function-based intervention was then developed, tested, and implemented during ongoing activities in the classroom. On-task behavior occurred throughout more than 91% of the intervals when the intervention was implemented correctly, compared to only 9% when it was implemented incorrectly. Positive treatment acceptability ratings were obtained from both Josh and his teacher, even though she continued to implement inconsistently throughout the study. Implications for both research and practice are presented. (Contains 1 table and 3 figures.)