To determine the minimum number of days of dietary intake interviews required to reduce the effects of random error (day-to-day variability in dietary intake) when using the multiple-pass, multiple-day, 24-h recall method. Cross-sectional study. University research department. A total of 50 healthy non-smoking overweight and obese (body mass index=26-40 kg/m2) adult men and women aged 39-45 years completed the study. Participants were randomly selected from volunteers for a larger unrelated study. Each participant completed 10, multiple-pass, 24-h recall interviews on randomly chosen days over 4 weeks. The minimum number of record days was determined for each macronutrient (carbohydrate, fat, protein) and energy, for each gender, to obtain a 'true' (unobservable) representative intake from reported (observed) dietary intakes. The greatest number of days required to obtain a 'true' representative intake was 8 days. Carbohydrate intakes required the greatest number of days of dietary record among males (7 days), whereas protein required the greatest number of days among females (8 days) in this cohort. Sunday was the day of the week that showed greatest variability in macronutrient intakes. Protein (P<0.05) and fat (P<0.001) intakes were significantly more variable than carbohydrate on Sundays compared with weekdays, for both men and women. A logistically achievable 8 days of dietary intake interviews was sufficient to minimize the effect of random error when using the multiple-pass, 24-h recall dietary intake method. Sunday should be included among the dietary interview days to ensure a 'true' representation of macronutrient intakes. This method can be confidently applied to small cohort studies in which dietary intakes from different groups are to be compared or to investigations of associations between nutrient intakes and disease.