This study assessed the validity and reliability of a 12-item soy questionnaire designed for use in cancer prevention research. The questionnaire measures soy intake over the past year. Subjects were 199 healthy 35- to 46-yr-old premenopausal women participating in an ongoing soy intervention study. Soy questionnaire estimates of isoflavone intake over 1 yr were compared with individual and combined estimates from two reference measures covering the same period: three or four repeated 24-h recalls and one or two repeated urine analyses. The sensitivity and specificity of the soy questionnaire in classifying high vs. low exposure (high exposure = soy intervention group membership) were comparable with 24-h recalls and were superior to urine analyses (sensitivity = 94.8%, 97.9%, and 71.1%; specificity = 95.1%, 97.1%, and 90.3% for the soy questionnaire, 24-h recalls, and urine analyses, respectively). Soy questionnaire isoflavone estimates were highly correlated with the combination of the two reference measures for the entire study population. Its brevity, ease of administration, and good measurement properties over a 1-yr period make the soy questionnaire well suited to the needs of researchers who wish to identify high and low consumers of soy foods, especially in populations who consume traditional Asian soy foods.