In the videodistortion method subjects are asked to reconstruct the exact sizes of their boby image as shown in a distorted form on a video screen. The reliability of this method (test-retest) is studied in anorexia nervosa patients and female students. Anorexics showed a clearly better test stability than control subjects. Significant differences were found in estimations departing from a widened (`fat') versus narrowed (`thin') image distortion. The findings will be discussed especially with regard to methodological problems.