OBJECTIVE To determine the extent to which incorporating fat-modified foods into a food frequency questionnaire influences the agreement of energy and nutrient estimates with estimates obtained from food records. METHODS Subjects completed four 2-day food records at 3-month intervals. At the end of the recording period, a food frequency questionnaire was administered to assess usual daily intake during the preceding year. METHODS One hundred and three subjects selected from a population-based sample of adults participating in the Beaver Dam Eye Study. METHODS Subjects were categorized into three groups on the basis of their frequency of consumption of fat-modified foods. For each group, correlations were calculated between food record estimates and estimates obtained from the original food frequency questionnaire, the original with a low-fat option, and the fat-modified questionnaire. RESULTS For persons categorized as high consumers of fat-modified foods, incorporating questions regarding the consumption of these products resulted in higher correlations with food record estimates (original vs fat-modified version) for percentage of energy from total fat (.32 vs .47), saturated fat (.20 vs .41), oleic acid (.32 vs .50), and linoleic acid (.40 vs .46). High consumers differed in several characteristics that could be associated with disease risk (eg, higher ratios of serum total cholesterol to high-density lipoprotein cholesterol). CONCLUSIONS Failure to account for the consumption of fat-modified foods in epidemiologic studies may result in misclassification of fat exposures. Because patterns of misclassification could be different for those at risk for disease, results of epidemiologic studies could be biased if these foods are excluded. Thus, incorporating fat-modified foods into food frequency questionnaires will improve the ability of researchers to correctly classify fat exposures and to evaluate potentially important relationships between fat intake and disease risk.