A common problem in nutrition assessment is determining the number of days of food intake information required to obtain accurate estimates of nutrient intakes. A standard definition of accuracy is needed for clinical practice and nutrition research. Any measure of accuracy should have two properties: 1) accuracy should improve as the number of observations increases and 2) the number of observations needed to achieve a given degree of accuracy should be unaffected by changes in the scale of measurement. Two approaches to defining accuracy are presented, one based on a proportion of the mean and the other using a proportion of the standard deviation (SD). The concept of reliability of measurement is introduced along with equations and tables that can be used in determining the number of repeated food records needed for a particular purpose. Defining accuracy in terms of the variability and reliability shows that increasing the number of days of a food record has a quickly diminishing return. For most purposes, 3-14 d of food diary records per subject will prove adequate.