Thank you for the debate on CBT and schizophrenia (Reference Turkington and McKennaTurkington/McKenna, 2003). I would like to make the following points.

First, CBT is not a single treatment – it contains many complex components and skills, and therapist variables must be an essential issue for careful evaluation as with all psychological therapies.

Second, befriending fares significantly better than ‘treatment as usual’ in much CBT research. McKenna dismisses this as placebo or ‘special treatment’. The fact of such significant improvement from befriending says something very serious about treatment as usual. Why should those suffering from psychosis not receive special treatment? The finding points to the need for more consideration of the (poorly termed) ‘non-specific factors in psychotherapy’ – factors clearly not treated as sufficiently important in basic care in psychosis (Reference Paley and ShapiroPaley & Shapiro, 2002).

Third, in the Sensky et al (Reference Sensky, Turkington and Kingdon2000) trial quoted, CBT patients maintained their (significant) clinical improvement at follow-up, whereas the befriending controls fell back towards previous levels. It seems that CBT gives the patients a thinking structure to help manage some of their symptoms in the longer term.

Fourth, many people believe that you cannot treat persons with psychosis as if they were suffering from something such as diabetes, for which a single remedy like insulin might be sufficient. McKenna's pronouncement on randomised controlled trials is, therefore, open to serious questioning. The need adapted approach is the antithesis of the randomised controlled trial method. In the former, the treatment is individualised and intentionally different (qualitatively and quantitatively) from one case to another and may well change over time. A randomised controlled trial, equally intentionally, eliminates individuality in the treatment. Because the idea of relationships can be especially disturbing to patients with psychosis, psychological therapies can be seen by patients as threatening; therefore, the therapy has to be very carefully ‘administered’ – individually and flexibly.

Fifth, there are other outcome measurements at least as important as psychiatric symptoms. The experience of treatment is very important, as well as quality of life measurements. Turkington emphasises the high take-up rate of CBT, far higher than uptake of medication in psychosis.

Sixth, thank goodness for CBT, just one of several ways for practitioners to re-discover some tools that enable them to relate to patients with psychosis. McCabe et al (Reference McCabe, Heath and Burns2002) show how uncomfortable ordinary psychiatrists are without such tools when engaging with patients when the latter want to discuss symptoms.

Seventh, CBT and psychodynamic approaches overlap to a degree, at least as practised by Turkington (Reference MartindaleMartindale, 1998; Reference Turkington and SiddleTurkington & Siddle, 1998). Much has changed in psychodynamic therapy since the flawed studies of old. Modern psychodynamic approaches to psychosis have a much more flexible technique in engaging patients, and a greater and broader appreciation of mental mechanisms in psychosis.

Finally, relationship approaches in psychosis need encouragement, support and research. All psychiatrists need basic training in engaging with patients with psychosis. Research indicates that befriending might be a good place to start, but it is clearly not so easy – as the outcome of ‘treatment as usual’ indicates.