Under-reporting of food intake is one of the fundamental obstacles preventing the collection of accurate habitual dietary intake data. The prevalence of under-reporting in large nutritional surveys ranges from 18 to 54% of the whole sample, but can be as high as 70% in particular subgroups. This wide variation between studies is partly due to different criteria used to identify under-reporters and also to non-uniformity of under-reporting across populations. The most consistent differences found are between men and women and between groups differing in body mass index. Women are more likely to under-report than men, and under-reporting is more common among overweight and obese individuals. Other associated characteristics, for which there is less consistent evidence, include age, smoking habits, level of education, social class, physical activity and dietary restraint. Determining whether under-reporting is specific to macronutrients or food is problematic, as most methods identify only low energy intakes. Studies that have attempted to measure under-reporting specific to macronutrients express nutrients as percentage of energy and have tended to find carbohydrate under-reported and protein over-reported. However, care must be taken when interpreting these results, especially when data are expressed as percentages. A logical conclusion is that food items with a negative health image (e.g. cakes, sweets, confectionery) are more likely to be under-reported, whereas those with a positive health image are more likely to be over-reported (e.g. fruits and vegetables). This also suggests that dietary fat is likely to be under-reported. However, it is necessary to distinguish between under-reporting and genuine under-eating for the duration of data collection. The key to understanding this problem, but one that has been widely neglected, concerns the processes that cause people to under-report their food intakes. The little work that has been done has simply confirmed the complexity of this issue. The importance of obtaining accurate estimates of habitual dietary intakes so as to assess health correlates of food consumption can be contrasted with the poor quality of data collected. This phenomenon should be considered a priority research area. Moreover, misreporting is not simply a nutritionist's problem, but requires a multidisciplinary approach (including psychology, sociology and physiology) to advance the understanding of under-reporting in dietary intake studies.