This study investigated the relative validity of a self-administered 24-h recall questionnaire in a dietary survey on 3,653 men and women 7 years of age and older. The validation was carried out in a group of 41 men. An estimated dietary record kept over 3 days served as reference method. Comparison of the questionnaire and the estimated 3-day record showed good agreement. The Wilcoxon matched-pairs ranked signs test (p < 0.05) demonstrated that the only differences were the crude energy and carbohydrate intake and the estimated nutrient density of protein. The estimated proportion of calories from carbohydrate, fat, protein, and alcohol differed by no more than 2.4%. The median percentage differences in crude nutrient intakes and nutrient densities between the two assessment techniques ranged from -9% to 22%. The daily food intake differed significantly in only three of ten food groups. Spearman's correlation coefficients were higher than 0.35 for all density measurements. The highest correlation coefficients of about 0.60 were observed for alcohol and dietary fiber intake. It is concluded that the self-administered 24-hour recall questionnaire is a valid method for estimating the median and mean dietary intake of large groups of subjects.