A generalizability study of the Eating Attitudes Test (EAT-12) in non-clinical adolescents.

One of the most frequently used measures of eating and weight problems is the Eating Attitudes Test (EAT). The paper discusses conceptual heterogeneity in the EAT in the light of eating disorder continuum theory. Psychometric qualities of the Norwegian 12-item version of the EAT was based on a sample of 224 female and 261 male adolescents aged 14-15 years. Generalizability theory enabled us to study the impact of multiple sources of measurement errors of the EAT-12. The analyses suggest that acceptable generalizability is achieved by the present design on applying 4 items within 3 categories of the EAT. Alternative designs are suggested, and the utility of G-theory is discussed. The results also provide evidence of a gender difference in the structure of the universe score variance. Non-clinical adolescents' ways of organizing information about food and eating are discussed in the light of gender differences.