To test the accuracy of self-reported energy intake, reported intake was compared with measured energy expenditure. Results from nine studies were reviewed in which intake data were obtained by recall or weighed record for at least 7 days. Expenditure was measured for 7 days or more by the doubly labelled water method. Individual differences between reported intake and expenditure were large (range +25 to -76%). Group mean differences were smaller. Lean, nonathletic groups living in industrialized countries demonstrated the smallest mean difference between self-reported energy intakes and expenditure (0 to -20%). Obese populations demonstrated the largest mean differences (-35 and -50%), but women living in the Gambia and elite athletes also demonstrated large mean differences. Most of the difference appears to be due to under-reporting, but some subjects lost weight during the reporting period indicating that some of the difference was due to undereating. Because the greatest bias was observed in obese subjects, current methods for self-reported energy intake are not recommended for use in obesity research.