Although the accuracy of dietary intake information for children has previously been studied, methodological issues make the results of those studies difficult to interpret. In order to address one of the major methodological issues, unobtrusive observations were performed on the lunch meal of 34 children. These children ranged in age from 4.0 to 9.5 years (mean = 5.8 years, standard deviation = 1.6 years). Each child was accompanied by both parents. Dietary recalls were obtained the following day from (a) the mother alone, (b) the father alone, and (c) the mother, father, and child reporting as a group (consensus recall). Recalls were analyzed with nutrition software that yields information on energy, protein, carbohydrate, sugar, total fat, cholesterol, sodium, iron, and calcium. Strong correlations were seen between each recall and the observation (mean r = .86). However, the group accuracy in correctly reporting different types of foods varied from the fathers' under-reporting of breads (-27%) to fathers' over-reporting of fruit (+50%). When regression analyses were used, only the consensus recall resulted in a regression line not significantly different from 1.0 for the majority of the nutrients analyzed. Thus, it appears that the consensus recall produced a better estimate of the observed intake from one meal than did recalls obtained from mother or father.