We report on the reproducibility and validity of a food frequency questionnaire (FFQ) developed for southern India. One hundred and two adult subjects aged > or =20 years drawn from the Chennai Urban Rural Epidemiology Study participated. The FFQ was developed based on local foods and habits, and was administered three times at 0, 6 and 12 month periods (called FFQ1, FFQ2 and FFQ3) to assess the reproducibility. To test the validity of the FFQ, multiple 24-h recalls collected at 2-monthly intervals for a period of 1 year and the estimated energy intake/basal metabolic rate (EI/BMR) ratio were used. The 'EpiNu' in-house food and nutrient database was used to compile dietary intakes, which were analyzed for statistical analysis. The intraclass correlation coefficient for all three FFQs ranged from 0.72 for carbohydrates to 0.45 for folate. The de-attenuated Pearson correlation for the energy adjusted nutrients between FFQ3 and 24-h recalls ranged from 0.73 for carbohydrates to 0.35 for calcium. Bland and Altman plots for energy intake between the two methods showed the limits of agreement ranged from 768 to -1358 calories/day (+/-2 standard deviations). Misclassification was low for most of the nutrients. Under-reporting of the energy intake (EI/BMR ratio<1.2) was higher in females than males. This FFQ appears to be a robust tool to measure dietary intakes in southern India.