OBJECTIVE: The aim of this study was to describe the sources of dietary variance, and determine the variance ratios and the number of days needed for estimating the habitual diet of adolescents. METHODS: Two 24 hour food recalls were used for estimating the energy, macronutrient, fatty acid, fiber and cholesterol intakes of 366 adolescents attending Public Schools in Natal, Rio Grande do Norte, Brazil. The variance ratio between the intrapersonal and interpersonal variances, determined by Analysis of Variance, was calculated. The number of days needed for estimating the habitual intake of each nutrient was given by the hypothetical correlation (r)>0.9 between the actual and observed nutrient intakes. RESULTS: Sources of interpersonal variation were higher for all nutrients and in both genders. Variance ratios were <1 for all nutrients and higher in women. Two 24 hour dietary recalls were enough to assess energy, carbohydrate, fiber and saturated and monounsaturated fatty acid intakes accurately. However, the accurate assessment of protein, lipid, polyunsaturated fatty acid and cholesterol intakes required three 24 hour recalls. CONCLUSION: Interpersonal dietary variance in adolescents was greater than intrapersonal variance for all nutrients, resulting in a variance ratio of less than 1. Two to three 24 hour recalls, depending on gender and the study nutrient, are necessary for estimating the habitual diet of this population.