Freehand, three-dimensional (3-D) ultrasound (US) systems, which incorporate an electromagnetic tracking device to register the 3-D spatial location of images acquired using a standard linear array scan-probe, are a flexible and cost-effective solution for many clinical applications. The reconstruction accuracy of one such system was investigated by using a precision-made phantom. The error in 3-D distance measurements, under conditions appropriate to US investigations of the carotid arteries, was found to be -0.45 +/- 1.30 mm, equivalent to -0.53 +/- 3.39% (mean +/- SD). The results are relevant to data acquired using a single sweep scan and for distances in the range 25.00 to 79.06 mm. Both the overall accuracy and precision in point-target location were found to be relatively unaffected by scan depth, and the precision of point-target location was found to be poorest in the elevational direction. In conclusion, the system tested in our laboratory performed with high accuracy, adopting a setup and scan-sweep identical to that used for imaging of the carotid arteries in 3-D.