Food-frequency questionnaires have become the dominant method for assessing dietary intakes in epidemiologic studies. However, their accuracy continues to be questioned. In a recent study, volunteers consumed three diets of varying fat content over 6 weeks. Compared with diet records, the food-frequency questionnaire provided less reliable estimates of the absolute amounts of fats and cholesterol consumed. Advocates of the food-frequency approach argue that attempting to validate the instrument against standard diets represented a highly contrived situation. In their view, food-frequency questionnaires critically depend on the participants' long-term knowledge of their own dietary patterns and are intended to measure intakes averaged over at least a year. That viewpoint tacitly acknowledges that food-frequency questionnaires have less to do with memory for what was consumed than with subjective inferences about the nature of the habitual diet. Food frequencies, much like food preferences or body image, appear to be a measure of attitude. As such, they may not be subject to absolute validation procedures.