The motivation for this paper is the authors' recent work on several randomized control trials in which they found the primary result, which averaged across subgroups or sites, to be moderated by demographic or site characteristics. They are led to examine a distinction that the Institute of Education Sciences (IES) makes between "confirmatory" and "exploratory" questions that are being addressed in an RCT. IES correctly wants to encourage a disciplined methodology in requiring that researchers "call their shots" by naming the small number of outcomes considered most important. While they embrace the need for this kind of discipline, the authors want to guard against taking the notion of confirmation too far. Thus, the authors ask how much faith researchers, working in school-settings, should put on the confirmatory results of RCTs? Is all rigorous school-based research exploratory at heart? The authors suggest a positive answer to this based on concerns about generalizability and about the limits of external validity. They will cite two recent examples of their RCTs where the value of the confirmatory result is brought into question. In each case, they find moderator effects, which challenge the value for practitioners, developers, and policy makers of the overall confirmatory result. (Contains 1 figure and 1 footnote.)