We assessed the validity and reproducibility of a self-administered 68-item food frequency questionnaire completed in 1992-1993 by approximately 185,000 adults. Four hundred forty-one participants completed four 24-hour dietary recall interviews over a 1-year period and a repeat administration of the food frequency questionnaire. For 20 nutrients and 10 food groups, measured nutrient intakes, but not food group intakes, were consistently lower by food frequency questionnaire than by recall. Energy-adjusted, attenuation-corrected Pearson validity correlations ranged from 0.12 to 0.80, with a median of 0.58. Reproducibility measures were generally high, with a median of 0.69. The food frequency questionnaire performed similarly to food frequency questionnaires used in other cohort studies, indicating similar ability to examine diet-disease relations.