Implanting fiducial markers for localization purposes has become an accepted practice in radiotherapy for prostate cancer. While many correction strategies correct for translations only, advanced correction protocols also require knowledge of the rotation of the prostate. For this purpose, typically, three or more markers are implanted. Elongated fiducial markers provide more information about their orientation than traditional round or cylindrical markers. Potentially, fewer markers are required. In this study, we evaluate the effect of the number of elongated markers on the localization accuracy of the prostate. To quantify the localization error, we developed a model that estimates, at arbitrary locations in the prostate, the registration error caused by translational and rotational uncertainties of the marker registration. Every combination of one, two and three markers was analysed for a group of 24 patients. The average registration errors at the prostate surface were 0.3-0.8 mm and 0.4-1 mm for registrations on, respectively, three markers and two markers located on different sides of the prostate. Substantial registration errors (2.0-2.2 mm) occurred at the prostate surface contralateral to the markers when two markers were implanted on the same side of the prostate or only one marker was used. In conclusion, there is no benefit in using three elongated markers: two markers accurately localize the prostate if they are implanted at some distance from each other.