Recent experimental and clinical studies suggest that torsade de pointes may be precipitated by early afterdepolarizations in the Purkinje or ventricular muscle fibers. This hypothesis offers an alternative to the earlier one that attributes torsade to the underlying dispersion of repolarization. This review lists the clinical conditions associated with torsade de pointes and examines the experimental background of the two proposed electrophysiologic substrates of torsade, namely, the dispersion of repolarization and the early afterdepolarizations. The strengths and weaknesses of the two hypotheses are compared in relation to the following characteristics of torsade de pointes: facilitation by slow heart rate, suppression by pacing, R on T phenomenon, difficulty of induction by programmed stimulation, aggravation by hypokalemia, manifestation of an idiosyncratic reaction to class IA antiarrhythmic drugs, spontaneous termination, suppression by magnesium salts and isoproterenol and induction by such drugs as sotalol, bepridil and prenylamine. It appears that most clinical observations can be explained by either mechanism, but in some cases difficulties are encountered for the afterdepolarization hypothesis.