Archeomagnetically determined values of inclination and declination are commonly reduced to a base location, the regional geomagnetic field being approximated either by the axial dipole or by a virtual dipole. We investigate the error in these approximations by using the spatial variations of the 1965 IGRF to simulate the temporal secular variation. The virtual dipole approximation is found to be significantly more accurate than the axial dipole approximation, particularly when the archeomagnetic site is at the same latitude as the base. Inclination error and declination error each have a normal distribution, but the directional dispersion is not circular-normal (Fisherian). Tables are pre-sented by which the statistical error in reduction of archeomagnetic directions to a base site can be readily estimated.The secular variation of geomagnetic inclination and declination over the past few milenia can be reconstructed by measuring the TRM of archeological baked clay. This has already been accomplished to varying degrees in several parts of the world, notably Is the "axial dipole" method of geographical correction used by Aitken and Weaver (1965), as good as the "virtual dipole" method used by Watanabe and DuBois (1965)? Our approach is statistical.We consider the geomagnetic field direction predicted for a site A from the field direction at a site B to be statistically distributed about the actual field direction at site A. The nature of the distribution will depend on the relative separation of the two sites and on the geographical correction method used to reduce the field at B to the site A. Time is not considered as a parameter, so we ignore the possibility of