Morriss et al ^{Reference Morriss, Dowrick, Salmon, Peters, Dunn, Rogers, Lewis, Charles-Jones, Hogg, Clifford, Rigby and Gask1} performed a high-quality cluster randomised controlled trial in which reattribution for medically unexplained symptoms was taught to general practitioners (GPs). We compliment the authors on this trial. Strong points of their trial are the avoidance of selection bias by using an independent GP for inclusion, and the inclusion of patients for whom unexplained symptoms of sufficient duration were the reason for the encounter. However, we have some critical comments as well.

First, the training of GPs took only 6 hours and was performed by non-expert trainers. Reattribution is not an easy technique to learn. Other researchers have used training programmes of at least 20 hours. ^{Reference Blankenstein2,Reference Rosendal3} The trainers in this study were three nurses and a psychologist. Although they were prepared intensively, they might not have been familiar enough with GP consultations. Consequently, we have doubts about the thoroughness and effectiveness of the training for GPs.

Second, the effect of reattribution training on doctor–patient communication has been evaluated in only one consultation. Reattribution usually takes more than one consultation. ^{Reference Goldberg, Gask and O'Dowd4} Making an inventory of the problems and broadening the agenda can lead to quite a disturbance of the normal flow of the consultation. Patients often need more time to make a link between their psychosocial and physical problems. In the article it seems like it was mostly the doctor who made the link. This does not fit into the original reattribution model. A negotiating style is needed in order to let the patient raise the possibility of a link him- or herself. ^{Reference Goldberg, Gask and O'Dowd4} For the purpose of effective reattribution, the patient has to come up with the link and not the doctor. ^{Reference Blankenstein2,Reference Rosendal3}

Third, we know that the effectiveness of psychological treatments consists of specific and non-specific effects. Non-specific effects are effects caused by mutual trust, empathy and shared understanding. ^{Reference Stewart5} The training in reattribution and applying it might have influenced the physicians' relation with the patient negatively because of the physicians being absorbed by the application of the new intervention. Less attention for empathy and other non-specific effects might have been an additional cause for the absence of treatment effects.

Finally, it is a pity that the authors did not differentiate the outcome effects for subgroups. Patients with medically unexplained symptoms form a heterogeneous group. ‘Treatment effects are always moderate’ due to the differences in levels of emotional and physical stress. ^{Reference Schweickhart, Larisch and Fritzsche6} The subgroup of patients with low emotional stress before treatment might have experienced deterioration in outcome measures after reattribution because of the consequent opening up and admittance of their problems. Although this is a clinically valuable change process, by reporting the overall treatment effects, this profit might be concealed.

In short, we think that some of the questions surrounding the treatment of patients with medically unexplained symptoms has been clarified by this high-quality trial, but there remain many others.