The accuracy (reproducibility and relative validity) of a self-administered quantitative food frequency questionnaire including 180 food items was evaluated. A total of 38 elderly women kept multiple weighed diet records for a total of 14 days over a 6-week period and filled in the questionnaire both before and after this period. Spearman rank correlations between the nutrient intakes from the two questionnaires varied from 0.43 for carbohydrate to 0.88 for energy percentage from alcohol. The median correlation was 0.70. Seven women whose recorded average energy intake was less than 1.17 times their measured basal metabolic rate were excluded from the analysis comparing the questionnaire and the diet records. The first questionnaire gave on the average 10% higher nutrient median values than the records, while the second questionnaire did not in general produce higher values. Unadjusted correlation coefficients comparing intakes measured by the two methods ranged from 0.31 for vitamin C to 0.79 for energy percentage from carbohydrate, the median coefficient being 0.61. On the average 77% of the subjects were classified in the same (+/- 1) quintile in the first food frequency questionnaire and the diet records. The present study indicates that the self-administered quantitative food frequency questionnaire is useful for measuring individual or group intakes for a variety of nutrients.