The assumed three-dimensional structure of the 12-item version of the Eating Attitudes Test was examined by confirmatory factor analysis in a non-clinical sample of 224 female and 261 male adolescents in grade 8 aged 14-15 years. Alternative models were examined. Data did neither support a three-dimensional structure, nor a one-dimensional structure using twelve items. Three dimensions were supported when one and three items were omitted from the girls' and boys' analyses, respectively. A second order factor structure was supported in both genders. Conceptual and psychometric issues are discussed. A discussion of the measurement domain of the eating problem construct in non-clinicals is warranted.