The argument that recovery from psychosis (and other severe mental illnesses) should be defi ned in terms of improved functioning and establishing a rewarding and meaningful life, rather than the permanent remission of symptoms and impairments, has gained traction since it was proposed more than 20 years ago. 1 Recovery for people with a serious mental illness is now broadly accepted as the legitimate goal of mental health ser vices, and is endorsed by high-level policy-makers in the UK and the USA. 2,3 Despite the growing enthusiasm for the vision of recovery and the hope imbued by the new defi nition, researchers have been reluctant to wade into the murky waters of recovery beyond exploratory, qualitative studies of the meaning of recovery to service users with mental health disorders. Mike Slade and colleagues 4 have broken new ground in their innovative study of the REFOCUS intervention, which was designed to foster recovery through changes in staff and team values, knowledge, skills, behaviour, and recovery-promoting relationships with patients. In a cluster, randomised, controlled trial, these investigators assessed the eff ects of usual care plus REFOCUS compared with usual care alone (control) in 27 community-based adult mental health teams providing care to patients with psychotic disorders. As might be expected owing to the sheer ambitiousness of the project, not everything worked out as hoped or expected. Outcomes were assessed in 403 patients at baseline and 297 at 1 year, who received care from 14 teams in the REFOCUS group or 13 in the control group. The primary endpoint, recovery, which was measured with the Questionnaire about Processes of Recovery (QPR), did not diff er between the two groups (REFOCUS group 40·6 [SD 10·1] vs control 40·0 [10·2], adjusted diff erence 0·68, 95%CI –1·7 to 3·1, p=0·58). Slade and colleagues suggest that an important reason for the lack of diff erence between groups was diffi culty in implementing parts of the intervention due to factors such as competing demands on staff members’ time. These hindrances led to substantial variation in participation levels of diff erent teams and changes in the recovery orientation of staff members. Staff turnover also diluted the potential eff ectiveness of the intervention on team recovery orientation, and patients dropping out from treatment and research reduced the statistical power to detect diff erences between patients in the diff erent study groups. In view of these diffi culties, it should be no surprise that REFOCUS did not lead to signifi cant diff erences from usual care alone. Despite this null result, some fi ndings suggest that the REFOCUS intervention is promising. Patients who received care from teams in the REFOCUS group improved more in overall functioning and unmet needs than those receiving usual care. Furthermore, compared with REFOCUS staff in low-participation teams, staff in high-participation teams reported more recovery-promoting behaviour (adjusted diff erence –0·4, 95% CI –0·7 to –0·2, p=0·001) and patients receiving care from higher-participation teams had improved overall functioning, quality of life, and scores on the QPR interpersonal subscale (adjusted diff erence in patient-rated interpersonal scores –1·6, –2·7 to –0·5, p=0·005). These results suggest that REFOCUS did improve recovery outcomes for patients receiving care from teams that more fully implemented the intervention. The REFOCUS trial represents an important milestone in research on recovery-oriented services. 5