To determine the accuracy of our ultrasonographically predicted birth weights we studied 1301 women delivered of infants within a week of an obstetric ultrasonogram to compare the ultrasonographically predicted birth weights with the actual birth weights. The fetuses varied from 700 to 5800 gm and were consecutive singleton fetuses in vertex presentations delivered at a single institution. Overall 74% of the infants had birth weights within 10% of the ultrasonographic estimates and 42% had birth weights within 5% of the ultrasonographic estimates. The presence of oligohydramnios or polyhydramnios made no difference in the percent errors. The sensitivity for identifying a fetus with macrosomia (birth weight greater than 4000 gm) with an estimated weight of greater than or equal to 4000 gm was 65%. The specificity or percent of fetuses correctly identified ultrasonographically as not macrosomic was 90%. If fetuses predicted by ultrasonography to be greater than 3800 gm were included, the sensitivity for the prediction of macrosomia rises to 82% but the specificity would be 79%. There appears to be a fixed limitation to obtaining estimated fetal weights by ultrasonography, even in large series, because these data reaffirm the success and limitations of other methods used to estimate fetal weight previously reported in the literature.