The authors investigated whether soy intake is associated with sex steroid levels in Japanese children. This cross-sectional study was conducted in autumn 2006. Subjects were substantially healthy preschoolers, 230 boys and 198 girls, aged 3-6 years. Dietary data, including soy intake, were assessed using 3-day dietary records. Each child's dietary intake was controlled for total energy intake using the Willett method (Nutritional Epidemiology. Oxford, United Kingdom: Oxford University Press; 1990:245-271). Urinary estrone, estradiol, testosterone, and 5-androstene-3β,17α diol levels measured using liquid chromatography-electrospray ionization tandem mass spectrometry, and urinary dehydroepiandrosterone level measured with a radioimmunoassay, were adjusted for urinary creatinine levels. In the analysis of covariance for sex steroids after adjustments for age and body mass index, soy intake was significantly negatively related to estrone and estradiol in boys and positively related to testosterone and 5-androstene-3β,17α diol in girls. Isoflavone had a significant tendency to be negatively associated with estradiol in boys and to be positively associated with testosterone in girls. Total energy intake was not associated with any sex steroids in boys or girls. These results suggest that soy intake might affect the secretion or metabolism of sex steroids in childhood and that the effects might differ by sex.