To date, no study has directly and simultaneously measured the discrepancy between what people actually eat and what they report eating under observation in the context of energy balance (EB). The present study aimed to objectively measure the ‘extent’ and ‘nature’ of misreporting of dietary intakes under conditions in which EB and feeding behaviour were continuously monitored. For this purpose, a total of fifty-nine adults were recruited for 12 d, involving two 3 d overt phases and two 3 d covert phases of food intake measurement in a randomised cross-over design. Subjects had ad libitum access to a variety of familiar foods. Food intake was covertly measured using a feeding behaviour suite to establish actual energy and nutrient intakes. During the overt phases, subjects were instructed to self-report food intake using widely accepted methods. Misreporting comprised two separate and synchronous phenomena. Subjects decreased energy intake (EI) when asked to record their food intake (observation effect). The effect was significant in women ( − 8 %, P< 0·001) but not in men ( − 3 %, P< 0·277). The reported EI was 5 to 21 % lower (reporting effect) than the actual intake, depending on the reporting method used. Semi-quantitative techniques gave larger discrepancies. These discrepancies were identical in men and women and non-macronutrient specific. The ‘observation’ and ‘reporting’ effects combined to constitute total misreporting, which ranged from 10 to 25 %, depending on the intake measurement assessed. When studied in a laboratory environment and EB was closely monitored, subjects under-reported their food intake and decreased the actual intake when they were aware that their intake was being monitored.