BackgroundSome realists have criticised randomised controlled trials for their inability to explain the causal relations that they identify; to take into account the influence of the social context of the interventions they evaluate; and to account for individual difference. However, among realists, there is controversy over whether it is possible to improve trials by making them realist, or whether realism and the philosophical assumptions underlying trials are incompatible. This paper contributes to the debate in Trials on this issue. The debate thus far has concentrated on the possibility of combining trial methodology with that of realist evaluation.Main bodyWe concur with the contention that it is not feasible to combine randomised controlled trial design with the realist evaluation approach. However, we argue that a different variant of realism, critical realism, provides a more appropriate theoretical grounding for realist trials.In contrast to realist evaluation, which regards social mechanisms as an amalgam of social resources and people’s reasoning, critical realism insists on their distinction. It does so on the basis of its assertion of the need to distinguish between social structures (in which resources lie) and human agency (which is at least partly guided by reasoning). From this perspective, conceiving of social mechanisms as external to participants can be seen as a valid methodological strategy for supplementing the exclusive concentration of trials on outcomes.While accepting realist evaluation’s insistence that causality in open systems involves a configuration of multiple generative mechanisms, we adopt the critical realist interpretation of the experimental method, which sees it as creating artificial closure in order to identify the effects of specific causal mechanisms. If randomised controlled trials can be regarded as epidemiological proxies that substitute probabilistic controls over extraneous factors for closed experiments, their examination of the powers of discrete mechanisms through observation of the variation of outcomes is appropriate.ConclusionWhile there are still issues to be resolved, critical realist randomised controlled trials are possible and have the potential to overcome some of the difficulties faced by traditional trial designs in accounting for the influence of social context and individual interpretation.