A paper referred to me for statistical review has drawn my attention to the issue of appropriate evaluation of the effectiveness of treatments used in psychiatry. It has been suggested that the classical randomised controlled trial (RCT) is inappropriate to many aspects of psychiatry. I feel it is important that this issue should be looked into closely. I believe that the principle that treatment evaluation should be by RCT wherever this is feasible is certainly no less important in psychiatry than in other medical specialties. Discussion of this issue is needed, so that guidelines can be set out which will both be acceptable to researchers in psychiatry and meet scientific standards accepted in other areas of medicine.