Thank you for the interest in our paper; we would like to clarify some points.

First, we conducted a 6-hour training intervention in reattribution because, on the basis of a series of studies of training in primary care, this is the length of training that most general practitioners (GPs) are prepared to attend in the UK and also in many other healthcare systems in the world. The 6-hour training produced the changes in communication that have been reported with 20-hour training in reattribution. ^{Reference Blankenstein, van der Horst, Schilte, de Vries, Zaat, Andre Knotternus, van Eijk and de Haan1} Moreover, more extensive training in reattribution for more than 20 hours by GPs does not necessarily improve patient outcome. ^{Reference Rosendal, Olesen, Fink, Toft, Sokolowski and Bro2} We used nurses and a psychologist because in practice these trainers would carry out this training in the work place if the intervention was ever implemented in routine practice in the UK. We received systematic feedback from the GPs about training via feedback forms at the time of training, a survey carried out later, and via in-depth qualitative interviews carried out in a sample of the GPs. The issue that the trainers might not understand the consultation was not raised as a concern by the GPs in the study.

Second, the paper describing the reattribution model, ^{Reference Goldberg, Gask and O'Dowd3} which was written by one of our team (L.G.) and subsequent descriptions of reattribution written by members of our team, have always promoted a model in which doctors provide the ‘making the link’ explanation although they should do this through negotiation with the patient. In our trial, the intervention group of GPs gave the ‘making the link’ explanation in a negotiatory manner much more frequently than the treatment as usual group. We agree that reattribution may be more effective on patient outcome if patients made the link themselves between their physical symptoms and a psychosocial cause. However, GPs may need to spend much longer with patients to achieve this.

Third, we agree that an instrumental task-oriented consultation such as reattribution might be perceived as less empathic by patients with medically unexplained symptoms than treatment as usual. However, in our trial the data from the patient satisfaction questionnaire suggests that compared with treatment as usual, after reattribution training twice as many patients were very satisfied with how well the GP understood the nature of their problems and their worries (reattribution training (n=57) v. treatment as usual (n=68): nature of the problem 34 (60%) v. 23 (34%); worry 34 (60%) v. 20 (29%); P<0.10 for both items, intention-to-treat analysis allowing for missing data, clustering at practice and GP level, age and gender of patient using generalised linear latent and mixed models). The data suggest that patients perceived GPs trained in reattribution to be no less empathic than GPs delivering treatment as usual. Therefore, there may be other features of the reattribution intervention delivered by GPs in this way that may explain its lack of effectiveness. We have explored this in a qualitative interview study with patients in the trial that will be submitted for publication.

Finally, we agree that certain subgroups of patients with medically unexplained symptoms may benefit from reattribution. However, our trial was not powered to examine this issue.