The reproducibility and relative validity of a semi-quantitative food frequency questionnaire (FFQ) were evaluated. Subjects were 72 (36 men and 36 women)residents of N island, Shimane Prefecture in Japan, aged 50-76 years. The questionnaire was administered twice at an interval of one week(FFQ1, FFQ2), and 7-day weighing diet records were collected during that period. This protocol was conducted in summer and winter. The mean intakes for 11 nutrients calculated from the FFQ1 and FFQ2 were similar and, those from the FFQ1 were smaller than those from the diet records all but fat and vitamin C. Pearson correlation coefficients assessing reproducibility between nutrient intakes from FFQ1 and FFQ2 ranged from 0.59 for vitamin C to 0.90 for carbohydrate. Validity was measured by comparing nutrient intakes from FFQ1 with those from the 7-day diet record, pearson correlation coefficients ranged from 0.15 for iron to more than 0.35 for total energy and macronutrients. These correlation coefficients were slightly differ in gender or age group, but did not vary materially between season. These data indicated that this questionnaire is reproducible but has some problems in validity.