Background: Illness Management and Recovery (IMR) is intended to provide a structured psychosocial programme that helps to manage the disabling effects of severe mental illnesses such as schizophrenia and bipolar disorders. It is curriculum based and aims to improve different aspects of illness management and recovery through interventions such as goal-setting, psycho-education, coping and social skills training. Its overall aim is to improve illness outcomes and support subjective and objective recovery. To date there have been four RCTs on IMR; as these yielded mixed results, further research is needed. Our hypotheses aim to test the interrelatedness assumed in Mueser's Conceptual Framework for IMR for the many aspects of illness management, illness management outcomes and recovery. Methods/design: This randomised multi-centre, single-blinded clinical trial is intended to compare IMR with treatment as usual for 200 outpatient clients with a severe and persistent mental illness (SMI). We will investigate whether IMR leads to better illness management, fewer symptoms and fewer relapses, and also to better subjective and objective recovery. The primary outcome measure is the score on the client version of the Illness Management and Recovery Scale. Secondary outcome measures are the clinician version of the Illness Management and Recovery scale, measures of illness management, coping, symptoms, the number of relapses, and measures of recovery. Measurement will take place before randomisation, and 12 and 18 months after randomisation. Discussion: Overall, our study has the following strengths: 1.) our use of an RCT design in a country where the earlier RCTs on IMR were not conducted; 2.) the fact that participants will consist not only of people with a diagnosis of schizophrenia, but also of those with various types of SMI; 3.) our inclusion of 200 participants; and 4.) the fact that we will explore the working mechanisms described in Mueser's Conceptual Framework for IMR. Finally, 5.) because the RCT will be conducted in everyday clinical practice, we believe that the generalisability of our results will be good.