Our validation study examined a three-dimensional ultrasonographic phantom that contained irregularly shaped volume targets ranging from 0.5 to 76.1 milliliters. Four different examiners made blinded measurements from volume datasets that were acquired by 4 and 7 MHz transducers. Birthweight predictions using abdominal and thigh volumes from 18 term fetuses also were compared with two-dimensional ultrasonographic methods. In vitro volume measurements were accurate, precise, and repeatable despite a small systematic overestimation with increasing object size. Mean systematic error and precision for birthweight predictions by three-dimensional ultrasonography (-0.03% +/- 6.1%) were not significantly different from those by two-dimensional ultrasonography (-0.60% +/- 8.8%). Conventional prediction methods yielded three birthweights with greater than 15% error. By comparison, except for one infant whose birthweight indicated an 11% error, all predictions based on fetal volume parameters were within 10% of true values. Accurate birthweight predictions by fetal volume parameters appear to be technically feasible at term gestation although their practical clinical application requires further investigation. Birthweight predictions in this manner may allow remote consultants to evaluate the fetus over wide-area computer networks despite the physical absence of the patient.