Valid food-frequency questionnaires (FFQ) need to be developed to assess isoflavone intake in investigations of its possible association with the lower incidence of breast and prostate cancer in Asian countries. We investigated the validity and reproducibility of isoflavone (daidzein and genistein) intakes from self-administered semiquantitative FFQ used in the JPHC Study (Japan Public Health Center-based Prospective Study on Cancer and Cardiovascular Diseases). We also investigated the number of food items that would be sufficient to ensure validity and reproducibility. We collected FFQ, dietary records (DR), blood and urine samples from 215 subjects among JPHC Study participants, estimated isoflavone intakes from FFQ and DR, and measured serum isoflavone concentration and urine isoflavone excretion. For daidzein, mean intakes estimated from FFQ and DR, serum concentration and urine excretion were 18.3 mg/d, 14.5 mg/d, 119.9 nmol/L and 17.0 micromol/d and for genistein, 31.4 mg/d, 23.4 mg/d, 475.3 nmol/L and 14.2 micromol/d, respectively. Results were similar when analyzed by sex. Spearman correlation coefficients for daidzein of energy-adjusted intakes from FFQ with those from DR, serum concentration and creatinine-adjusted urinary excretion were 0.64, 0.31 and 0.43, respectively. Correlations between two FFQ estimates with a 1-y interval were 0.76. Results were similar for genistein. The shorter version of the FFQ with three items (natto, miso and tofu for miso soup) showed a similar correlation. The original FFQ and the shorter versions have sufficient validity and reproducibility to be used in epidemiologic studies.