To determine specificity of rodent models of arrhythmia for different Vaughan Williams classes of antiarrhythmic drugs, we tested 17 drugs from the four classes in one in vitro and four in vivo models. In the mouse chloroform-induced ventricular fibrillation model and in the guinea pig ouabain-induced arrhythmia model, drugs of classes I (amefalone, aprindine, lidocaine, mexiletine, phenytoin, procainamide, or quinidine), II (metoprolol or propranolol), and IV (bepridil) were active. Class III drugs (bretylium, clofilium, or melperone did not suppress ouabain arrhythmias, but were active in the mouse chloroform model. In the rat coronary artery ligation model, disopyramide (class I), amefalone and melperone significantly (P less than 0.05) reduced the number of extrasystoles. Propranolol, sotalol, and verapamil (class IV) were less effective. In the rat coronary artery ligation/reperfusion model, all four classes of antiarrhythmic agents were active in vitro (isolated heart) and in vivo (anesthetized rat). Thus, one model of automaticity, the guinea pig ouabain model, detected class I, II, and IV drugs, whereas another automaticity model, the mouse chloroform model, also detected class III agents. The model of reentry induced by ischemia plus reperfusion (rat coronary artery ligation reperfusion) can be recommended as a screen for new antiarrhythmic agents based on its sensitivity to all four classes of antiarrhythmic drugs. The Vaughan Williams class of an antiarrhythmic agent must be determined, however, by additional mechanism studies.