Dr Mitchell professes both confusion and irritation at our report of the follow-up results of the SoCRATES study. I will attempt to clarify. When implemented with standard care, CBT has consistently been found to result in clinical benefits in terms of symptom relief compared with standard care alone. This has been found in post-acute illness, as with the reported study, and chronic phases (Pilling et al, 2002). Cognitive–behavioural therapy is a structured psychological treatment usually implemented from a manual, which makes it relatively amenable to ‘roll-out’; CBT has less impact on relapse rates. Family interventions have been shown to have the benefit of significantly reducing relapse rates (Pilling et al, 2002). Thus, combining both CBT and family interventions would appear to be the most parsimonious way of capitalising on these developments to improve patient care by reducing symptoms of psychosis and reducing risk of subsequent relapse. In addition, both service users and carers have been increasingly vocal in wishing a greater range of interventions to be made available, including both psychological treatments and assistance for families. It is regrettable that Dr Mitchell's comments implicitly appear to wish to deny them these further options. With respect to his comment on our inclusion criteria for the trial, a first episode of psychosis resulting in treatment by mental health services is an event that can be identified with reasonable accuracy (at least, much more accurately than emergence of symptoms prior to this). As 80% of study participants had first-episode illness using this criterion, the use of the term ‘recent onset’ is not unreasonable. The inclusion of those suffering from schizophreniform psychosis, delusional disorders and unspecified psychosis reflects clinical practice and conforms to convention on large pragmatic clinical trials in having broad inclusion criteria to aid generalisation (see Johnson, 1998).