A self-administered semi-quantitative food frequency questionnaire (FFQ) was developed for the Dutch cohort of the European Prospective Investigation into Cancer and Nutrition (EPIC). The reproducibility and relative validity of nutrient intake as assessed by this questionnaire were investigated in a population of 121 men and women. To assess the relative validity, 12 monthly 24-hour recalls served as reference method, together with four determinations of 24-hour urinary nitrogen excretion, predicted basal metabolic rate, and serum beta-carotene and alpha-tocopherol levels. Protein and among women, energy intake were underestimated by the questionnaire compared to urinary nitrogen excretion and the basal metabolic rate, respectively. The underestimation for protein decreased with increasing protein intake. Pearson correlation coefficients between nutrient intakes assessed by repeated questionnaires ranged from 0.70 to 0.94 among men and from 0.59 to 0.94 among women. Correlation coefficients between nutrient intakes assessed by the questionnaire and 24-hour recalls ranged from 0.26 to 0.83 for men and from 0.35 to 0.90 for women, with medians of 0.59 and 0.58, respectively. Correlation coefficients between 0.2 and 0.5 were observed for beta-carotene and vitamin C for men and for beta-carotene and vitamin E for women. Associations with serum beta-carotene (r = -0.16 for men; 0.13 for women) and alpha-tocopherol (0.23 and 0.15, respectively) were much poorer than those obtained with 24-hour recalls. Correlations between protein intake and 24-hour urinary nitrogen excretion were 0.47 and 0.53, respectively. The FFQ seems adequate for ranking subjects according to intake of energy, macronutrients, dietary fibre and retinol, but it does not yield such good results for beta-carotene, vitamin C for men, vitamin E for women.