Over how much time should the geomagnetic field be averaged to obtain the mean-palaeomagnetic field?

Average geomagnetic field models are now available over six orders of magnitude in time. Models based on Present and historical, archaeomagnetic and palaeomagnetic data illustrate how the amplitudes of secular variation of nonaxial dipole terms strongly decrease as a function of averaging window, mostly when jumping from the 1-yr to the 1700-yr average. The mean archaeomagnetic field bridges the gap between the complex historical field and the far simpler palaeomagnetic field, and appears to be much closer to the latter. Confidences in average field coefficients based on statistical field models give a quantitative basis to the practice of averaging data over a few millenia in order to 'suppress secular variation'. In particular, the axial quadrupole emerges from noise when the available time series exceed about 6 kyr. This happens to be close to the timescales of excursions and reversals. It is now a priority to obtain detailed data spanning the last 50 kyr.