A self-administered semiquantitative food frequency questionnaire including 75 food items and providing information on the habitual intake of 31 nutritional parameters, based on the intake of protein, fat, carbohydrate, fiber and 11 vitamins and minerals, was developed for use in epidemiologic research on chronic disease among the elderly, such as diabetes and cardiovascular disease. By means of detailed frequency and quantity questions, specifications of types of food, preparation methods and seasonal variation, the questionnaire was expected to be an improvement on existing instruments. The relative validity of the questionnaire was examined in 74 men and women, aged 50-75, by comparison with a modified dietary history. Systematic differences were absent or negligible for all nutrients, except vitamin C. Bias depending on the level of intake could be ruled out for all but seven nutrients. Pearson correlation coefficients for estimates from the questionnaire and dietary history were on average 0.71 (range: 0.65-0.78) and 0.66 (range: 0.36-0.81) for macronutrients, and vitamins and minerals, respectively. Classifying individual intake estimates into tertiles of the distribution for both methods, on average 62.4 and 54.7% of the intakes were categorized into the same tertile and 3.9 and 5.9% into the opposite tertile for macronutrients, vitamins and minerals, respectively. These results demonstrate an acceptable relative validity for this newly developed questionnaire, as compared to the dietary history method.