Objective : The present study was conducted to report upon the reproducibility of a 98-item food frequency questionnaire among Koreans. Subjects : The study subjects were recruited from among those who visited for a regular health check-up at the health centers from Samsung Hospital and Hallym University Hospital. Setting : The FFQ was administered first in April to June of 2002 to 145 Korean adults aged 40 and over residing in Seoul and its vicinity and was then re-administered to 126 three months later between July and September of 2002 (FFQ 1 and FFQ 2). Methods : Reproducibility was evaluated using the Pearson correlation coefficients of log-e and the calorieadjusted nutrient score. Weighted kappa (k) statistics with 95% confidence limits were calculated to assess the chance adjusted level of agreement between the FFQ 1 and the FFQ 2. The proportions of correctly categorized subjects in the same or adjacent quintiles were calculated. Results : The average intake in FFQ 1 was no more than 12 percent different from the average intake in FFQ 2. Correlations varied between 0.47 for sodium and 0.72 for vitamin C. All k values exceeded 0.5 except that of fiber. The average k for all nutrients was 0.67. The percentage agreement varied from 62% for energy and potassium to 82% for vitamin B2 and cholesterol. The average of the agreement was 72%. Conclusion : The results of this study verify that it is possible to use tailored, relatively simple, but comprehensive, self-administered food frequency questionnaires to study nutrient consumption in large-scale epidemiological studies.