We examine the validity and reproducibility of a food frequency questionnaire (FFQ) in a subsample of participants in the Japan Public Health Center-based Prospective Cohort Study using a database of polyphenol-containing foods commonly consumed in the Japanese population. Participants of the validation study were recruited from two different cohorts. In Cohort I, 215 participants completed a 28-d dietary record (DR) and the FFQ, and in Cohort II, 350 participants completed DRs and the FFQ. The total polyphenol intake estimated from the 28-d DR and FFQ were log-transformed and adjusted for energy intake by the residual method. Spearman correlation coefficients (CCs) between estimates from the FFQ and 28-d DR as well as two FFQs administered at a 1-year interval were computed. Median intakes of dietary polyphenols calculated from the DRs were 1172 mg/d for men and 1024 mg/d for women in Cohort I, and 1061 mg/d for men and 942 mg/d for women in Cohort II. The de-attenuated CCs for polyphenol intake between the DR and FFQ were 0⋅47 for men and 0⋅37 for women in Cohort I and 0⋅44 for men and 0⋅50 for women in Cohort II. Non-alcoholic beverages were the main contributor to total polyphenol intake in both men and women, accounting for 50 % of total polyphenol intake regardless of cohort and gender, followed by alcoholic beverages and seasoning and spices in men, and seasoning and spices, fruits and other vegetables in women. The present study showed that this FFQ had moderate validity and reproducibility and is suitable for use in future epidemiological studies.