OBJECTIVE: Our aim was to evaluate the reproducibility of a food-frequency questionnaire (FFQ) developed in a French population, and its validity against four 24-h dietary recalls (24-HRs). METHODS: A total of 57 adults (aged 18-63), 17 adolescents (aged 14-18) and 20 children (aged 10-14) completed four 24-HRs (one per season) and two FFQs over a 1-year interval. Reproducibility of the FFQ was estimated by intraclass correlation coefficients (ICCs). The validity of FFQ was assessed by comparison with the four 24-HRs. RESULTS: For most nutrients, the first FFQ measurements gave higher mean values than the second FFQ. The ICCs for nutrients ranged from 0.39 for total protein to 0.83 for alcohol. The ICCs were higher for food items consumed daily such as milk (0.80) or sugars and confectionery (0.65), and lower for rarely eaten food such as inner organs (0.11). Nutrient intakes as assessed by FFQs were higher than those from the 24-HRs, except for alcohol. The de-attenuated Pearson's correlation coefficient for nutrients varied from 0.25 (dietary fiber) to 0.90 (alcohol), but the adjustment for energy did not improve these coefficients. When nutrient intakes were categorized into quintiles, FFQ and 24-HRs produced agreement rates (same or adjacent quintile) between 55% (for PUFA) and 95% (for alcohol), while misclassification to an extreme quintile was rare (<5%). CONCLUSION: The FFQ developed for the FLVS II Study can be used to classify adults or adolescents according to their nutrients and food intakes over a 1-year period.