Interpreting the record of the Earth's magnetic field preserved in rocks - palaeomagnetism - is a complicated business, but at the heart of it is one very simple assumption: except when it is reversing, if you average over a few thousand years or so, the geomagnetic field resembles a dipole aligned with the Earth's geographic poles.
This relatively uncomplicated shape means that there is a very simple relationship between latitude and the magnetic inclination (the angle magnetic field lines make with the horizontal); it is zero at the equator, and gradually increases to 90 degrees at the poles. If you measure the direction of the fossil field direction carried by rocks at a particular site, a simple formula converts the inclination of this ancient magnetisation into the palaeolatitude of that particular chunk of crust at the time the rocks formed. Because the field is symmetric, a reversal changes the polarity, but not the shape, of the field; for example, an inclination value of 50 and -50 degrees both always correspond to a latitude of 30 degrees.
But what if our simple assumption is wrong, and the Earth's magnetic field has not always been a dipole? There are more complicated quadropole and octopole components in the present geomagnetic field, but they are fairly minor and, except during a magnetic reversal, seem to average out to zero over a few thousand years. But what if at some point in the geological past these components were not only a more significant part of the geomagnetic field, but also did not average to zero over geological time? This would produce an asymmetric long-term field geometry, as in the figure below, where 15% of the earth's magnetic field energy is in the quadropole component. For a point at mid-to low northern latitudes, rocks forming in a normal polarity field would have a shallow magnetic inclination, whilst rocks forming in a reversed polarity field would have a much steeper inclination. The warped field geometry means that there is no longer a one-to-one relationship between inclination and latitude, which makes working out the plate motions recorded by all of those ancient magnetic directions much more difficult.
