Objective: To evaluate the repeatability of a children's food frequency questionnaire (FFQ) by gender, ethnicity, and age group. Design: A 117-item FFQ asking about food intake patterns over the past 4 weeks was developed using food records from 428 children (204 boys and 224 girls) and the reproducibility on average 13 days apart was tested in 130 children (78 boys and 52 girls). Children were recruited using clustered probability sampling (n ¼ 103), and a convenience sample of 25 Maori children. Setting: Children aged 1-14 y from Auckland, Feilding and Shannon, New Zealand. Subjects: There were 71 Maori, 20 Pacific, and 39 Other children. Results: Spearman correlations between the two FFQs ranged from 0.50 for bread to 0.82 for fruit, with a median of 0.76 for spreads and nonmilk drinks, and Cronbach's coefficient a's ranged from 0.59 for bread to 0.92 for nonmilk drinks, with a median of 0.85 for mixed meat dishes. There were no significant differences between the two administrations, apart from reporting higher intakes of vegetables and snacks & sweets in the first FFQ. Correlation coefficients tended to be slightly higher in boys than in girls, and in Other ethnic groups compared to Maori and Pacific children. Correlations were slightly higher for the 1-4 y age group, intermediate in the 10-14 y age group, and lowest in the 5-9 y-old age group. Conclusions: Overall, the FFQ described here shows similar or better repeatability in New Zealand children of all major ethnic groups compared to other child or adolescent FFQs.