We studied the influence of age, sex, education and time between interviews on the reproducibility of a semi-quantitative food frequency questionnaire (FFQ) developed in Italy for a case-control study on cancers of the breast and digestive tract. The questionnaire had been administered twice to 452 Italian men and women and included the weekly consumption of 77 food items or groups of foods, seven summary questions and three questions on some general dietary habits. Spearman correlation coefficients for the 77 dietary items plus the seven summary questions did not differ between males (median 0.61) and females (median 0.58), volunteers younger than 50 years (median 0.58) and aged 50 or more (median 0.59), volunteers with fewer than 10 years of education (median 0.58) and with 10 or more (median 0.59). A slightly higher Spearman correlation coefficient was found when the two interviews were conducted 5-6 months apart (median 0.60) or more than 6 (median 0.59) than when less than 5 (median 0.55). Similarly, Pearson correlation coefficients for the intake of 27 nutrients or micronutrients, plus caloric intake, computed from the FFQ showed no differences between males (median 0.65) and females (median 0.64), two age groups (median 0.67 for subjects aged < 50 years and 0.65 for those aged > or = 50), and two educational levels (median 0.68 for < 10 years and 0.65 for > or = 10 years). The median Pearson correlation coefficient for nutrients was slightly higher for those subjects who were interviewed with the planned interval (5-6 months, median 0.71) than when the interval was shorter (median 0.57) or longer (median 0.64).(ABSTRACT TRUNCATED AT 250 WORDS)