In this prospective study conducted from 1984 through 1987, the ability to correctly predict growth discordancy in twin gestations by ultrasonic estimated fetal weights is examined. Discordancy was defined as an intertwin birth weight difference of 25% or greater. This method resulted in a sensitivity rate of 77% and a specificity rate of 92%. The positive predictive value of an abnormal test (i.e. discordant growth) was 67% and the negative predictive value of a normal test (i.e. concordant growth) was 95%. The perinatal mortality rate of 217/1,000 in discordant twin fetuses was significantly higher than 29/1,000 in the concordant twins in this study population (p < 0.01) and even more of a contrast to the rate of 10/1,000 in our singleton population. Accurate prediction of discordant twin pregnancies which are at high risk for poor outcome opens the opportunity for potential in utero treatment modalities.