Past weight or patterns of weight change may be more important to chronic disease risk than current weight. Self-reports, however, are often the only source of information about past body weight. To date, very few studies have examined factors affecting the validity of self-reported past body weight. We examined the validity of self-reported past body weights of 1,931 U.S. adults who were participants in the First National Health and Nutrition Examination Survey (1971-1975) and were interviewed again in the National Health and Nutrition Examination Survey I Epidemiologic Follow-up Study (1982-1984). We compared the body weight measured during the initial examination (1971-1975) with the recalled 1971-1975 body weight reported during the follow-up interview (1982-1984). Recalled past weight was strongly correlated with previously measured weight (r = 0.73 for men, and r = 0.74 for women). Men overestimated their past body weight, whereas women underestimated their past weight. Although 39% of men and 41% of women estimated their past weight within 5 pounds, approximately 17% of women and 10% of men underestimated their past weight more than 15 pounds. Accuracy of reporting was influenced by sex, race, current body mass index, and the amount of weight gained over the 10 years following the initial examination. These factors should be considered when using recalled weight in epidemiologic studies.