Our aim was to assess the intra- and inter-observer reliability in the establishment of the anterior pelvic plane used in imageless computer-assisted navigation. From this we determined the subsequent effects on version and inclination of the acetabular component. A cadaver model was developed with a specifically-designed rod which held the component tracker at a fixed orientation to the pelvis, leaving the anterior pelvic plane as the only variable. Eight surgeons determined the anterior pelvic plane by palpating and registering the bony landmarks as reference points. The exact anterior pelvic plane was then established by using anatomically-placed bone screws as reference points. The difference between the surgeons was found to be highly significant (p < 0.001). The variation was significantly larger for anteversion (sd 9.6 degrees ) than for inclination (sd 6.3 degrees ). The present method for registering pelvic landmarks shows significant inaccuracy, which highlights the need for improved methods of registration before this technique is considered to be safe.