While most geoscientists are well aware that past changes in the geomagnetic field are useful as a means of timing sea-floor spreading and stratigraphic correlation, and that records of the direction of palaeomagnetism are keys to ancient plate movements. Most, however, understand only vaguely why Earth has a magnetic field that flips polarity from time to time: there is some kind of self-sustaining dynamo due to motion in the liquid-metal outer core. That aspect of geomagnetism involves tough theory and maths. So for Scientific American to present an up-to-date review of how that dynamo might work is both surprising and welcome (Glatzmaier, G.A. & Olson, P. 2005. Probing the geodynamo. Scientific American April 2005, p. 33-39). The review covers what is currently known about convective motion in the outer core, both laminar and turbulent, and how the simpler laminar convection has been used in computer modelling that simulates how the geodynamo works. It is complex even at that level of simplification, because thermal convection is affected by the Coriolis effect: much like that in the atmosphere. Even though the idea of a dynamo inducing magnetic flux is a basic principle of physics, one based on fluid circulation is in constant motion and change. Surface monitoring of shifts in the magnetic field help chart that aspect. The issue of reversal is, literally, the knottiest problem for geomagnetists, and they have to resort to the old idea of lines of flux and the effect of contortions by motion at the core-mantle boundary to grapple with how polarity flips might occur. Computer simulations show the development of what can only be described as chaos in the geomagnetic field at the core-mantle boundary, and much smoothed, but nonetheless odd variability at the surface, as the poles prepare to reverse. For a period of around 6 000 years the field wobbles like a massive jelly as it lurches across the planet, sometimes splitting into several “blobs” of different polarity. Eventually it settles down into its new configuration. To some extent this strange behaviour is matched by what little is known in detail about the progress of reversals from the geological record (see Magnetic polarity reversals in May 2004 issue of EPN).