A study to evaluate the relative validity of two dietary assessment methods, to be used in a prospective cohort study on diet and cancer, was conducted in a random sample of the Malmö population during 1984-1985. The study compared (i) a long, self-administered 'food-use questionnaire' (about 250 food items), asking for frequency of consumption of food and beverages and for usual portion size estimated with the help of a booklet of food pictures, and (ii) a combined method involving a shorter food frequency questionnaire (about 130 food items) and a 2-week diet diary, with (iii) a reference method requiring 18 days of weighed food records. The results indicated that subject compliance was quite good for both methods and, in particular, that the diet diary was well accepted. Both methods produced fairly good correlations with the reference method, of the order of 0.40 to 0.60 for most nutrients after adjustment for energy. Both methods tended to overestimate intake of fat, protein and carbohydrates as well as of vitamins and minerals.