An important problem for the automatic implantable cardioverter-defibrillator is to discriminate ventricular tachycardia (VT) from superventricular tachycardia (SVT), and these arrhythmias from normal sinus rhythm (NSR). The authors present a new multiway sequential hypothesis testing (MSHT) algorithm for simultaneous discrimination of VT, SVT, and NSR. The MSHT algorithm allows explicit specification of false positive and false negative error rates, and a tradeoff between detection time and tolerable error levels. Endocardial atrial and ventricular recordings were obtained from 30 patients, and 17, 8, and 12 episodes of VT, SVT, and NSR, respectively, were evaluated. The error rates were selected so that all the rhythms were accurately discriminated and detection tests never took longer than 8 s for SVT-VT discrimination with average discrimination detection times of 2.9, 4.3, and 1.7 s for NSR, SVT, and VT, respectively. The algorithm discriminated the three rhythms with a sensitivity of 0.95 and specificity of 0.97.