To describe birth-weight-for-gestational-age patterns by race, sex, and parity in the United States population, and to discuss the measurements of gestational age by different methods, the pitfalls of each method, and the potential effects of the errors on birth-weight-for-gestational-age curves. We used the computerized certificates of live births from the entire population in 1989, consisting of more than four million infants born to residents of the United States. Gestational age was based on the date of the last menstrual period (LMP) modified by the clinical estimate in those situations in which normal distribution of birth weight does not apply. Birth weights for the tenth, 25th, 50th, 75th, and 90th percentiles were calculated by each gestational age and by race, sex, and parity. Eight sets of smoothed birth-weight-for-gestational-age curves were created for black-white, male-female, and primipara-multipara comparisons in sequence. Compared with previous major curves, our curves were closer to those in which the gestational age was derived from the LMP. There were marked differences observed between our curves and those in which the gestational age was based on ultrasound estimation. In the measurement of gestational age, the LMP may produce misclassification of gestational age, thereby elevating birth weight percentiles in preterm births and lowering birth weight percentiles in postterm births. However, ultrasound estimation is likely to create a differential misclassification of gestational age, which exerts the opposite effect of lowering birth weight percentiles early in gestation and increasing the percentiles late in gestation.