Nutrient indexes derived from food-frequency questionnaires have generally been regarded as acceptably valid for epidemiologic purposes. Evaluations of these indexes, however, have considered only their convergent validity. We suggest that discriminant validity, or the ability to distinguish among exposures to different nutrients, is also important. Using baseline data from a large clinical trial, we tested the discriminant validity of indexes of intake of vitamin E, alpha-carotene, and beta-carotene. Our results suggest that the vitamin E index possesses neither convergent not discriminant validity, the alpha-carotene index adequate convergent and discriminant validity, and the beta-carotene index adequate convergent but no discriminant validity.