Future antitachycardia devices must be able to deliver a variety of therapies according to the requirements of the underlying arrhythmia. To ensure that appropriate treatment is prescribed the device must use a detection algorithm that is able to discriminate between multiple arrhythmias. Current criteria such as rate, change of rate, duration at high rate, and high rate stability are inadequate for this purpose. Many algorithms that evaluate the morphology of the endocardial electrogram are of too great a complexity to be incorporated in implantable devices that require real-time analysis without undue power consumption. In this study the sensitivity of a simple morphological technique (temporal electrogram analysis) is examined. The method sets threshold 'rails' above and below the isoelectric line and classifies ECG complexes according to the sequence and duration of rail excursions. A total of 27 ventricular tachyarrhythmias were induced in 25 patients (17 with a history of recurrent arrhythmias and eight undergoing risk stratification postmyocardial infarction). Temporal electrogram analysis (TEA), initially detected the onset of the ventricular arrhythmia in all patients whose surface ECG showed polymorphic or right bundle branch block pattern tachycardia, in 5/8 of cases with left bundle branch block pattern and in 4/5 of patients with concordant complexes across the precordial leads. After minor modifications the overall sensitivity of the method was improved to 95% (26/27 arrhythmias detected). TEA is a technique of low computational demands, which in this study, reliably discriminated between resting sinus rhythm and ventricular arrhythmias.