The object of this study was to develop a method to assess the accuracy of an image-free total knee replacement navigation system in legs with normal or abnormal mechanical axes. A phantom leg was constructed with simulated hip and knee joints and provided a means to locate the centre of the ankle joint. Additional joints located at the midshaft of the tibia and femur allowed deformation in the flexion/extension, varus/valgus and rotational planes. Using a digital caliper unit to measure the coordinates precisely, a software program was developed to convert these local coordinates into a determination of actual leg alignment. At specific points in the procedure, information was compared between the digital caliper measurements and the image-free navigation system. Repeated serial measurements were undertaken. In the setting of normal alignment the mean error of the system was within 0.5°. In the setting of abnormal plane alignment in both the femur and the tibia, the error was within 1°. This is the first study designed to assess the accuracy of a clinically-validated navigation system. It demonstrates in vitro accuracy of the image-free navigation system in both normal and abnormal leg alignment settings.