The paper by Dare et al (Reference Dare, Eisler and Russell2001), on a trial of psychological treatments for anorexia nervosa, has two major shortcomings. The investigators planned for a year of weekly sessions of 50 minutes of psychoanalytic therapy; a year of weekly to 3-weekly sessions (60 to 75 minutes) of family therapy; 23 sessions (50 minutes) of cognitive—analytic therapy (CAT), and an unstated frequency of 30 minute sessions for 1 year for the ‘routine treatment’ group. The patients in the psychoanalytic arm ended up receiving a mean of 24.9 sessions as opposed to 12.9 for the CAT, 13.6 for the family therapy and 10.9 for the ‘routine’ arm. The differences in the numbers of sessions planned and those actually taking place has not been taken into account in evaluating the results. A summarised by Bergin & Garfield (Reference Bergin and Garfield1994), a large number of different studies show that more sessions are associated with greater improvements. However, the relationship is not linear and begins to taper off after 26 sessions: a figure almost reached by the patients in the psychoanalytic arm but far removed from that of the other three groups.

Not only did the ‘control’ group receive the fewest number of sessions, with each session lasting only 30 minutes, but as noted and implied by the authors: therapists assigned to this group had the least commitment to and experience in treating anorexia nervosa. The paper does not state how many therapists each patient ‘went through’ during the course of the study. All these factors would predispose to the formation of poor working alliances compared with the other groups. Thus, the poor results obtained by the ‘control’ group could be accounted for by a combination of fewer sessions of shorter duration and weak therapeutic alliances, rather than the superiority of specific psychological treatment models.