In 1989, 674 schoolchildren aged 12-14 years in nine elementary schools in a municipality in Copenhagen, Denmark, answered a questionnaire about their dietary habits and knowledge. The majority of the pupils had fruit (87%), vegetables (72%), rye bread (81%), and drank fat-reduced milk (73%) every day. A diet score (reliability = 0.58) was calculated on the basis of the intake of 8 food items relevant to current dietary recommendations. There were no age and sex differences as to dietary habits, but immigrant children had a lower diet score than native children. Dietary knowledge was measured by the ability to state correctly whether 11 different food items had a high content or not of fat, sugar or dietary fibres. Dietary knowledge was highest for questions about fat and sugar. A knowledge score measured the number of correct answers to all 33 questions (reliability = 0.90). Knowledge was highest among older children, native children, and children with the most healthy dietary habits. In the multivariate regression analysis, knowledge, health attitudes and ethnicity were the only significant predictors of dietary behaviour. It is concluded that both social and personal factors are important for dietary behaviour, and health promotion in children should include other methods than educational programmes.