Energy intake is the most difficult component of energy balance to measure. All known methods are subject to considerable error, both random error and bias. High variability within and between subjects means that sampling is important to increase reliability. Further, the fact that true intakes are unknown means that some outside measure of validity is desirable as a check on the accuracy of recording of habitual intake. This report examines food intake for populations where expenditure has been measured by doubly labeled water. For 21 populations where aggregate data have been reported, intake came within one megajoule (MJ) of expenditure in about one‐half of the cases. In seven populations where individual intake and expenditure were noted, reported intakes did not differ from measured expenditure for four groups, but differed significantly for two groups of African farm women and one group of obese volunteers. Individuals also differed in ranking of intake and expenditure, with intake able to predict the rank of expenditure only for one group of German students. Energy intake data are important for both ecological and epidemiological studies but are poor measures to use in estimating energy balance. © 1996 Wiley‐Liss, Inc.