This study examined the validity of self-reported delivery weight among 3,518 respondents to the 1988 National Maternal and Infant Health Survey. Self-reported delivery weight was ascertained from a mail survey administered during the postpartum period. Measured delivery weight was obtained by abstraction of medical records from the hospital of delivery. On average, a woman's reported delivery weight was 2.82 pounds (1 pound = 0.45 kg) less than her measured delivery weight (p < 0.001). The level of underreporting increased significantly with increases in prepregnancy body mass index, current body mass index, pregnancy weight gain, and weight change from delivery to recall. Reporting error also increased among women who were non-White, less educated, and unmarried; whose pregnancy was unintended; and who initiated prenatal care late, responded late to the survey questionnaire, became pregnant again before responding, and reported a delivery weight ending in zero. When reported delivery weight was used to calculate weight gain and was categorized into typical weight gain categories, 30-40% of women were classified incorrectly. An empirical evaluation of how this misclassification might impact epidemiologic analyses indicated that associations between weight gain and birth weight were attenuated when weight gain was based on reported delivery weight rather than on measured delivery weight.