Recent advances in the study of geomagnetic field reversals are reviewed. These include studies of the transitional field during the last geomagnetic reversal and the last geomagnetic excursion based on paleomagnetic observations, and analysis of reversals in self-consistent 3D numerical dynamo simulations. Field models inferred from observations estimate reversal duration in the range of 1-10 kyr (depending on site location). The transitional fields during both the Matuyama/Brunhes reversal and the Laschamp excursion are characterized by low-latitude reversed flux formation and subsequent poleward migration. During both events the dipole as well as the non-dipole field energies decrease. However, while the non-dipole energy dominates the dipole energy for a period of 2 kyr in the reversal, the non-dipole energy merely exceeds the dipole energy for a very brief period during the excursion. Numerical dynamo simulations show that stronger convection, slower rotation, and lower electrical conductivity provide more favorable conditions for reversals. A nondimensional number that depends on the typical length scale of the flow and represents the relative importance of inertial effects, termed the local Rossby number, seems to determine whether a dynamo will reverse or not. Stable polarity periods in numerical dynamos may last about 1 Myr, whereas reversals may last about 10 kyr. Numerical dynamo reversals often involve prolonged dipole collapse followed by shorter directional instability of the dipole axis, with advective processes governing the field variation. Magnetic upwellings from the equatorial inner-core boundary that produce reversed flux patches at low-latitudes of the core-mantle boundary could be significant in triggering reversals. Inferences from the observational and modeling sides are compared. We summarize with an outlook on some open questions and future prospects.H. Amit ( )