An updated analysis of the global paleomagnetic database shows that the frequency distributions of paleomagnetic inclinations for the Cenozoic and Mesozoic eras (younger than 250 Ma) are compatible with a random geographical sampling of a time-averaged geomagnetic field that closely resembles that of a geocentric axial dipole. In contrast, the frequency distributions of paleomagnetic inclinations for the Paleozoic and Precambrian eras (prior to 250 Ma) are over-represented by shallow inclinations. After discounting obvious secondary causes for the bias, such as from data averaging, sedimentary inclination error, inhomogeneous lithological distributions, and tropical remagnetization, we show that the anomalous inclination distributions for the Paleozoic and Precambrian can be explained by a geomagnetic field source model which includes a relatively modest (∼25%) contribution to the axial dipole from a zonal octupole field and an arbitrary zonal quadrupolar contribution. The apparent change by around 250 Ma to a much more axial dipolar field geometry might be due to the stabilization of the geodynamo from growth of the inner core to some critical threshold size, a gross speculation which would imply that either the threshold size was rather large or the inner core nucleated rather late in Earth history. Alternatively, if a geocentric axial dipole model is assumed or can eventually be demonstrated independently, the anomalous inclination distributions for the Paleozoic and Precambrian may reflect a tendency of continental lithosphere to be cycled into the equatorial belt, perhaps because geoid highs associated with long-term continental aggregates influence true polar wander.