We studied atrial flutter due to circus movement in chronically instrumented conscious dogs to identify the mechanism by which class I and class III antiarrhythmic drugs terminate reentrant excitation. We used a crossover experimental design administering five class I agents and one class III agent, by intravenous bolus followed by intravenous infusion. The class I agents other than lidocaine were almost uniformly effective in terminating the arrhythmia (disopyramide in six of seven dogs, propafenone in six of six, flecainide in seven of seven, and SC-40230 in seven of seven). Termination was preceded by a marked increase in cycle length (ranging from +78% with propafenone to +55% with disopyramide), but with the exception of disopyramide, class I agents did not significantly shorten the excitable gap. With disopyramide the gap decreased from 49 ± 3% to 28 ± 3% of the cycle length. With no class I agent did the wavelength of effective refractoriness increase to approach the cycle length of the arrhythmia. Lidocaine, used as a negative control, terminated the reentry in one dog with modest prolongation of the cycle length. Terminations with class I agents correlated with depression of conduction rather than prolongation of refractoriness. In contrast with class I agents, D-sotalol prolonged the cycle length minimally (+10%) and terminated the arrhythmia in six of seven dogs. It decreased the excitable gap from 42 ± 4% to 26 ± 6% of the cycle, but it still did not cause the wavelength of effective refractoriness to equal the cycle length. Terminations by D-sotalol seemed to result from either failure of the lateral boundaries of the circus path or reflection within the path.