Two hypotheses have been suggested to explain the far-sided nature of virtual geomagnetic poles from low latitude sites. One is the off-centered axial dipole model, and the other is the zonal nondipole field model. At high latitudes the two models predict different inclination anomalies. Two sets of data from Iceland (which is at an ideal latitude to distinguish between the two models) were therefore studied to resolve the relative merit of the two hypotheses. Three tests were done on each set of data and also on the combined data. The weight of evidence lies toward favoring the axial off-centered dipole hypothesis, at least for the period between 13 and 7 m.y.