The seemingly unpredictable occurrence of paroxysmal supraventricular tachycardia has impeded understanding of its clinical course. We documented recurring tachycardia by telephone transmission of the electrocardiogram in 14 patients and measured the "tachycardia-free periods" between the start of therapy and the first recurrence of tachycardia and between successive recurrences. We showed that attacks of tachycardia were clinically independent by calculating the autocorrelation function for 12 to 14 consecutive tachycardia-free periods. We also found that the periods between attacks closely fit an exponential probability distribution, indicating that the occurrence of tachycardia in time was a type of random event called a Poisson process. Furthermore, occurrence of tachycardia was uniformly distributed throughout a dosing interval; there was no predilection for tachycardia to occur late in a dosing interval when plasma drug levels were presumed to be lowest. Measuring the tachycardia-free period is a simple and useful method for objectively studying paroxysmal supraventricular tachycardia and may be applied to the study of other paroxysmal arrhythmias.