We assessed how accurately participants in dietary trials reported their free-living energy intake. We compared self-reported energy intake, calculated from 3-d food records, with actual intakes needed to maintain body weight during controlled trials lasting 6-9 wk. In 269 free-living healthy male (n = 119) and female (n = 150) adults with mean body weights close to ideal values (mean +/- SD body mass index in kg/m2, 22.1 +/- 2.4), energy intake reported in food records was 1.2 +/- 1.6 MJ/d (277 +/- 378 kcal/d) lower than actual energy requirements during the experiments. The relative bias was significantly smaller (P = 0.01) for men (-8.0 +/- 13.4%) than for women (-12.2 +/- 13.7%). Body mass index, daily energy intake, and age were not significantly related to the extent of underestimation. We conclude that food records systematically underestimate energy needs in young, nonobese well-educated adults.