An ultrasound phantom was constructed simulating fetal femurs in amniotic fluid. Bones of 26, 36, 50, 58, and 70 mm, representative of gestational ages ranging from 17 to 36 weeks, were scanned with mechanical sector, phased sector, and linear array systems (ATL, Diasonics, Acuson, and GE). Measurements were made with the bone in both a horizontal and nearly vertical orientation at 5, 10, and 15 cm from the transducer. The ultrasound measurements were compared with the true bone length. With bones in a nearly vertical orientation (parallel to the ultrasound beam) the ultrasound measurements corresponded more closely to the true bone length regardless of the type of equipment or distance from the transducer. The wide aperture linear system was most accurate with no measurable difference from the actual bone length and a mechanical sector scanner had the largest error which was 6 mm. In the horizontal position (perpendicular to the beam) the smallest errors occurred when the bone was in the focal zone. This ranged from no error for the wide aperture linear array to 8 mm for the mechanical sector scanner. When the bone was not in the focal zone the error ranged from 8 to 26 mm for the mechanical sector scanner. Errors in ultrasound-measured femur lengths can be shown to result from the focal characteristics of the equipment as well as the orientation and distance of the bone from the transducer. These differences can produce errors in estimation of gestational age as large as ten weeks.