Reentrant ventricular tachycardia (VT) is the most common sustained arrhythmia leading to ventricular fibrillation (VF). However, despite more than a century of research, the mechanism(s) of the conversion from reentrant VT to VF have not been elucidated. Based on their different electrocardiographic appearance, reentrant VT and VF have traditionally been thought of as resulting from two widely different mechanisms. Whereas VT is seen as a rapid but well organized process whereby the excitation wave rotates about a single well-defined circuit, fibrillation has been described as turbulent cardiac electrical activity, resulting from the random and aperiodic propagation of multiple independent wavelets throughout the cardiac muscle. Recently, the application of concepts derived from the theory of non-linear dynamics to the problem of wave propagation in the heart and the advent of modern high-resolution mapping techniques, have led some investigators to view VT and VF in terms of a single mechanism, whereby the self-organization of electrical waves forms 'rotors' that give rise to rapidly rotating spiral waves and results in either VT or VF, depending on the frequency of rotation and on the interaction of wave fronts with the cardiac muscle. As such, monomorphic VT is thought to result from a stationary rotor, whose frequency of rotation is within a range that allows 1:1 excitation of both ventricles. On the other hand, VF is thought to result from either a single rapidly drifting rotor, or a stationary rotor whose frequency of excitation is exceedingly high, thus resulting in multiple areas of intermittent block and giving rise to complex patterns of propagation with both deterministic and stochastic components. This article reviews the prevailing theories for the maintenance of VF, and discusses recently proposed mechanisms underlying transitions between VT and VF.