We critically analyzed published data on proposed sonographic criteria for intrauterine growth retardation (IUGR) to determine the positive and negative predictive values of the criteria. The data were drawn from 21 studies and encompassed nine distinct criteria for which the sensitivity and specificity could be determined. Predictive values were computed using Bayes theorem, based on an IUGR prevalence rate of 10%. Seven of the nine criteria had positive predictive values of less than 50%, indicating that a fetus with an abnormal test result according to any of these criteria is more likely to be normal than growth retarded. The best criterion involves the ratio of head circumference to abdominal circumference, but even with this criterion, fully 38% of fetuses with abnormal test result will not be growth retarded. We conclude that none of the proposed criteria allows a confident antenatal diagnosis of IUGR to be made. Future prospective studies may reveal new criteria, or combinations of existing ones, that are highly predictive for this condition.