Birth weight (BW) and log10 birth weight (LBW) are expressed as linear, multilinear, parabolic, and polynomial regression functions of the abdominal circumference (AC), and combinations of fetal index measurements, namely, AC and head circumference (HC); biparietal diameter (BPD) and AC; and BPD, HC, and AC. The relationship of somatic weight and log10 somatic weight to AC was similarly determined. The analysis generated 20 equations whose accuracy in providing BW estimates is compared. Several equations provide BW estimates whose mean percentage deviations from measured BWs do not differ significantly. The deviations, however, have the smallest variance when BW is estimated from the HC and AC either by using a multilinear regression equation or by deriving separate estimates of brain and somatic weights from these index measurements. The role of individual and population differences in fetal anthropometry as causes of error in estimating BW by different methods is discussed. Because HC is a function of both BPD and occipito-frontal diameter (OFD), it is a better brain-size modulus than the commonly used BPD and avoids the errors of underestimation which occur when the BPD is unusually small in cases of dolichocephaly. A comparison of ponderal growth in American and South African fetal populations shows differences throughout most of the gestational range (26-40 wk) analyzed in this study.