Rehabilitation using cognitive and behavioural treatment methods (i.e., experience-based interventions) faces particular challenges in improving its evidence base through rigorous studies such as randomised controlled trials (RCTs). Experience-based treatments are often complex, with multiple "active ingredients" that may be difficult to characterise. In addition to the difficulty in specifying treatment ingredients, experience-based rehabilitation researchers face challenges in designing or selecting appropriate control or comparison conditions to test the efficacy of complex treatments. Based on lessons learned in designing a cognitive-behavioural intervention for anger self-management for people with traumatic brain injury (TBI) for the National Institutes of Health (NIH)-funded TBI Clinical Trials Network, we review the advantages, disadvantages and applications of a variety of control conditions for experience-based interventions. We discuss controls in which active treatments are withheld (no-treatment controls, waitlist controls, and placebo-analogue designs); controls that involve comparison to naturally occurring or devised usual care treatments; and conditions that compare active treatments (dismantling designs, dose controls, and equivalence trials). Recommendations for selecting and developing control groups that maximise both equipoise and participant enrolment/retention are discussed. © 2007 Psychology Press.