Objective: To assess the reproducibility and validity of a semiquantitative food frequency questionnaire (FFQ) to classify children and adolescents in terms of daily servings of fruits and vegetables and intake of calories, protein, fat, carbohydrate, dietary fibre, vitamin C, phosphorous, calcium and iron. Design: FFQs were collected in the autumn of 1993 and 1994. Four 24-hour diet recalls were collected during the same 1-year period and their mean was compared to the FFQ diet estimates. Setting: Low income, inner-city state schools. Subjects: A sample of 109 inner-city fourth to seventh grade students. Results: The 1-year reproducibility of the FFQ, assessed with Spearman correlations, was lower among the fourth and fifth (range: r ¼ −0.26 to 0.40) than the sixth and seventh grade students (range: r ¼ 0.18-0.47). After adjusting for day-to-day variation in dietary intake, for most nutrients and foods the correlations between the FFQ and the 24-hour recalls remained greater among the junior high school students (fourth to fifth grade range: r ¼ 0.0-0.42; sixth to seventh grade range: r ¼ 0.07-0.76). Conclusions: Inner-city sixth and seventh grade students demonstrated the ability to provide valid estimates of intake of calories, carbohydrate, calcium, phosphorous, iron and vitamin C over the past year. However, children in the fourth and fifth grades experienced some difficulty in completing the FFQ. Our results suggest that, before using this instrument with fourth and fifth grade children, investigators should assess whether study participants can think abstractly and are familiar with the concept of 'average intake'.