Described is a sequential hypothesis testing algorithm capable of discriminating supraventricular arrhythmias from ventricular arrhythmias. The algorithm uses simultaneous recordings of pure atrial and pure ventricular electrograms obtained using dual chamber endocardial electrodes on a single catheter. Adaptive impulse filtering is used to subtract the ventricular component from the atrial lead signal to get a pure atrial electrogram. Probability distributions of the atrial and ventricular inter-beat interval for various arrhythmias are calculated. A sequential hypothesis testing (SHT) algorithm is used to determine which distribution the rhythm belongs to. The advantage of this method is that it allows preselection of the false positive and false negative error probabilities. The tradeoff for selecting lower error rates is a longer time required to reach a decision. The SHT algorithm is used to discriminate between normal and SVT rhythms in 20 data segments. The 9 normal rhythms and 10 SVT rhythms were classified correctly within 3 seconds.