A Confirmation of the Eight Factor Structure of the Eating Disorders Inventory in a Non-Clinical Sample, with New Zealand Norms

The Eating Disorders Inventory (EDI) is a widely used questionnaire for clinical assessment and for research with people who have eating disorders. It has eight subscales, providing information on eight separate dimensions of cognitive and behavioural aspects of eating disorders. While the existence of eight factors underlying these subscales is generally accepted for clinical samples, this has not always been the case when the EDI has been administered to nonclinical groups, a situation that has clear implications for its use as a screening device. In this study, the EDI was administered to a sample of 260 New Zealand female university students. An initial confirmatory factor analysis with five previously identified subscales supported a five factor structure, while successive analyses with each of the three remaining subscales included in turn, confirmed the eight factor structure. A single factor confirmatory analysis also provided strong support for the presence of an underlying general factor. As the EDI may therefore be a much more promising instrument for use with community samples than some critics have argued, summary statistics and norms were developed for all eight subscales and for the full EDI. ********** The Eating Disorders Inventory (EDI) is a 64 item, self-report questionnaire, designed to provide information on eight separate dimensions of cognitive and behavioural aspects of anorexia nervosa and bulimia (Garner, Olmsted, & Polivy, 1983, 1983a). The development and validation of this measure was first described by Garner, Olmsted, and Polivy (1983), who reported data on the internal consistency of its eight subscales, and also provided evidence for their convergent and discriminant validity. The eight subscales that comprise the EDI are called: Drive for Thinness (DT), Bulimia (B), Body Dissatisfaction (BD), Ineffectiveness (I), Perfectionism (P), Interpersonal Distrust (ID), Interoceptive Awareness (IA), and Maturity Fears (MF). While the EDI has gained considerable popularity recently with both clinicians and researchers, controversy has surrounded the nature of its underlying factor structure. In particular, the failure to demonstrate an eight factor structure consistently among non-clinical samples, has been cited as evidence that the EDI is not a suitable instrument for screening for eating disorders in community samples (Schoemaker, van Strien, & van der Staak, 1994; Bennett & Stevens, 1997). By "non-clinical" we refer to samples, such as university students, where it is assumed that most of the participants have never been diagnosed with an eating disorder. That the eight factors are reliably observed with clinical samples of people with an eating disorder does not appear to be disputed. Evidence for an eight factor structure among non-clinical groups Klemchuk, Hutchinson and Frank (1990) examined the factor structure of the EDI for a sample of 1,506 North American, female undergraduates. They compared five, six, and eight factor solutions using both oblique (promax) and orthogonal (varimax) methods of rotation. While the six factor solution was considered to be the most easily interpretable these authors were quite positive about the eight factor model. They reported five robust factors which corresponded closely with the putative subscales of the EDI. The five factors were: Body Dissatisfaction, Ineffectiveness, Interpersonal Distrust, Maturity Fears and Perfectionism. The sixth factor reported by Klemchuk et al., loaded highly on items from three EDI subscales, which were Drive for Thinness, Bulimia and Interoceptive Awareness. Klemchuk et al. (1990, p.299) described this sixth factor as "the only significant departure from the clinically derived eight-scale structure of the EDI". Lee, Lee, Leung and Yu (1997), examined the factor structure of the EDI in a sample of 606 female Chinese undergraduates in Hong Kong. Using a principal axis extraction with oblique rotation, they reported a robust eight factor solution, corresponding closely to the eight subscales originally proposed by Garner et al. …