The objective of this study was to assess the reliability of lactose consumption measured with a food-frequency questionnaire and to compare food-frequency data to prospectively recorded data. A self-administered food-frequency questionnaire (FF1) was completed by 295 women ages 38 to 49. Two months later, 75 received the same food-frequency questionnaire (FF2), and 75 received a 1-week diet diary listing 23 commonly eaten lactose-containing foods. Estimated mean intake of lactose was 16.1 g/d with FF1. Intake was slightly (but not significantly) higher at follow-up: Mean difference (95% confidence interval) was 1.4 (-0.8, 3.7) g/d for FF2 versus FF1, and 0.5 g (1.3, 2.3) g/d for the diet diary versus FF1. A moderate level of correlation was found: r = 0.57 between FF1 and FF2, and r = 0.65 between FF1 and the diet diary. Smaller subsets of items (n = 15 and n = 7) could be used to estimate lactose, with 89% and 82% agreement within tertiles, respectively. Since relatively few foods contain substantial amounts of lactose, a limited food-frequency or structured diary may be useful in research studies on the association between lactose consumption and ovarian cancer or other diseases.