Current government mental health policy includes a clear commitment to set the agenda for research and to manage the national research and development (R&D) portfolio proactively. Although most of us regret the demise of responsive funding, which permitted research driven by personal curiosity and creativity, the wastefulness of underpowered and repetitive studies was all too obvious. Thornicroft and his colleagues (Reference Thornicroft, Bindman and Goldberg2002, this issue) do us a service by bringing a multi-disciplinary research perspective to this thinking and providing a framework to develop a research strategy. They have done a very good job and deserve our thanks.

They make 11 recommendations. Because their approach is admirably thorough and transparent without expressing personal convictions and hunches, it could appear that all 11 have equal weight. Freed from their scientific constraint, I would suggest that three of the recommendations are of the highest priority.

Their observation of the UK's weakness in social science capacity and the need to fund training and posts (recommendation 1) is spot on. For all its failing, UK mental health care has a tradition of highly integrated multi-disciplinary working (Reference Burns and PriebeBurns & Priebe, 1999). Outcomes research of sufficient quality to answer current questions (e.g. those about different team configurations) requires research teams who can construct and test sharply-focused hypotheses. The alternative is a series of mechanical head-to-head studies that get us nowhere. This links in with recommendation 8 on the development of realistic definitions of key concepts such as accessibility and continuity. The authors may be pleased to note that the National Co-ordinating Centre for Service Delivery and Organisation has just commissioned a 5-year study into a better understanding of continuity of care in mental health. Such a study would simply not be possible without highly-qualified social scientists.

There really is no alternative to large-scale randomised controlled trials (RCTs) to resolve important questions that remain ambiguous despite other attempts. Following recommendation 3 for funding such studies would go a long way to improve rigour in mental health research and force the growth of genuinely collaborative multi-centre research initiatives that have been so successful in other branches of medicine.

The one recommendation missing from the list that I would have liked to see is for a strengthening of capacity in theory building. The British tradition of pragmatism in research is likely to be further entrenched by a more centrally steered research agenda, explicitly devoted to evaluating the NHS Plan. Recommendation 1, about building social science capacity, and recommendation 8, about refining key concepts, may go some way to achieving this. If we are going to fund large-scale RCTs (which will cost millions of pounds, take several years to conduct and are rarely repeatable) then it is crucial that adequate time and status is invested in developing and refining the questions asked. A recent systematic review into home treatment for mental illness (Reference Catty, Burns and KnappCatty et al, 2002) found the two significant variables in reducing hospitalisation were integration of health and social care in the same team and regularly visiting at home. It found no effect for case-load size. Had that work been commissioned before the UK700 trial (Reference Creed, Burns and ButlerCreed et al, 1999) would we have selected case-load as the independent variable?