Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Donald A. Williamson


The area of eating disorders have been intensively by both the psychological and medical researchers over the last two decades. Differential diagnosis of the four most prevalent eating disorders (anorexia nervosa, bulimia nervosa, compulsive overeating, and the non-binging obese) continues to be a problem for both researchers and clinicians. Diagnosis of these disorders has been complicated by the many changes in diagnostic criteria for each of the disorders. One consequence of these frequent changes is that there is no single self-report inventory for differential diagnosis which is based on currently accepted diagnostic criteria. The purpose of this study was the construction of an assessment instrument, the Eating Disorders Diagnostic Inventory (EDDI), which could reliably differentiate anorexia nervosa, bulimia nervosa, compulsive overeating, non-binging obese, and normals. The EDDI was developed from a group of test items which included items from the Eating Attitudes Test, the Bulimia Test, and the Eating Questionnaire-Revised, totalling 91 items. In the initial test construction phase of the study, there were 397 subjects. Discriminant analysis identified 35 items, which formed the EDDI, which differentiated the five diagnostic groups included in the study (anorexia nervosa, bulimia nervosa, compulsive overeating, obese, and normals) with a correct classification rate of 85.5%. Factor analysis identified three scales of the EDDI: a Binge Eat scale, a Drive for Thinness scale, and a Purgative Behavior scale. A second phase of the study, the Validation Phase, included four groups: bulimia nervosa, compulsive overeaters, obese, and normals. There were 15 subjects per group. A group of anorexics could not be obtained, so this group was dropped from the study. Measures of temporal stability indicated that the EDDI had satisfactory test-retest reliability over a two-week period. However, discriminant validity of the instrument was found to be unsatisfactory in that only 75% of the subjects were correctly classified using the discriminant function established in the first phase. The concurrent validity of the three factors of the EDDI was supported, however. Discussion focused upon additional research steps which may better establish the discriminant validity of the EDDI.