In a study of 105 anorexia nervosa patients, large hospital differences were found in weight gain which corresponded exactly to the amount of experience that each hospital had had in treating this disorder, and also corresponded to the degree of milieu structure imposed by each hospital. A further hypothesis to account for the hospital differences was in terms of the prognostic quality of the patients recruited to each hospital. To test the latter hypothesis, a number of prognostic indicators which had already been shown to be related to weight gain were statistically controlled by means of partial correlation. When this was done, the large hospital differences vanished, indicating that they were indeed a function of the qualities the patients brought with them to the hospital rather than what the hospital and its staff did to the patients. The results also indicate that the more experienced investigators seem to be able to recruit patients with better prognoses, whereas the less experienced ones are referred other clinicians’ treatment failures.