To assess the accuracy of maternally reported birth weights, we compared birth weights reported by mothers in the Tennessee Women, Infants, and Children Supplemental Feeding Program (WIC) from 1975 to 1984 with the birth weights recorded on the corresponding Tennessee birth certificate file. Differences in birth weights between these two sources were compared for the total group and were also stratified by sociodemographic and medical variables that might influence the accuracy of birth weight recall. An accurate birth weight was defined as one reported within 1 ounce of the birth certificate birth weight. We also calculated the proportion of birth weights that would be incorrectly classified as low or normal by maternal reporting. A total of 72,245 WIC records were matched with their corresponding birth certificates. Of these, 46,637 had WIC birth weights recorded within the specified birth weight range. Eighty-nine percent of birth weights were reported within 1 ounce of birth certificate birth weights. Lower accuracy of birth weight reporting was associated with the infant's low birth weight, preterm delivery, and low Apgar scores, and with the mother's grand multiparity, less than a high school education, black race, single marital status, and young age. Only 1.1 percent of birth weights would have been incorrectly classified into low or normal birth weight categories based on maternal reporting. Overall, our results suggest that maternally reported birth weights are sufficiently accurate for research and programmatic purposes when birth certificate information is not readily available.