Results from analysis of 24 h urine collections, verified for completeness with para-amino benzoic acid, and blood samples collected over 1 year were compared with 16 d weighed records of all food consumed collected over the year, and with results from 24 h recalls, food-frequency questionnaires and estimated food records in 160 women. Using the weighed records, individuals were sorted into quintiles of the distribution of the urine N excretion: dietary N intake ratio (UN:DN). UN exceeded DN in the top quintile of this ratio; mean ratio UN:DN = 1·13 Individuals in this top quintile were heavier, had significantly greater body mass indices, were reportedly more restrained eaters, had significantly lower energy intake:basal metabolic rate ratios (EI:BMR), and had correlated ratios of UN:DN and EI:BMR (r - 0·62). Those in the top quintile reported lower intakes of energy and energy-yielding nutrients, Ca, fats, cakes, breakfast cereals, milk and sugars than individuals in the other quintiles but not lower intakes of non-starch polysaccharides, vitamin C, vegetables, potatoes or meat. Correlations between dietary intake from weighed records and 24 h urine K were 0·74 and 0·82, and between dietary vitamin C and β-carotene and plasma vitamin C and β-carotene 0·86 and 0·48. Correlations between dietary N intake from weighed records and 24 h urine excretion were high (0·78–0·87). Those between N from estimated food records and urine N were r 0·60–0·70. Correlations between urine N and 24 h recalls and food-frequency questionnaires were in the order of 0·01 to 0·5. Despite problems of underreporting in overweight individuals in 20% of this sample, weighed records remained the most accurate method of dietary assessment, and only an estimated 7 d diary was able to approach this accuracy.