A food frequency questionnaire (FFQ) was developed to assess long-term habitual dietary intake in a cohort of approximately 100,000 40 to 79 year-old men and women living in the Southeastern US. Using the NHANES-III database for the southem region for specific race and sex subgroups, a list of 262 food categories was developed, coded and reduced to 102 food items that could discriminate between racial groups and account for large portions of cancer-relevant nutrients. The developed FFQ was tested in a pilot study in three southeastern states involving 239 African Americans and Whites, aged 56.9 +/- 12.2 years. The frequencies of consumption and portion sizes of the 102 foods were determined and intakes of various nutrients were estimated and compared with the NHANES-III data. African Americans reported higher total energy intakes and higher consumption of macronutrients and several micronutrients, compared to Whites. Estimated nutrient indices were higher among pilot study than among NHANES-III participants, although adjustment for total energy essentially eliminated the differences. Analysis of the frequency distribution of individual foods shows that the questionnaire includes commonly eaten foods that can discriminate between African Americans and Whites. The FFQ is currently being calibrated within the cohort study population.