Using beat-to-beat analysis we studied the annotation of 44 data-bases of the Massachusetts Institute of Technology (MIT) in comparison to the classification performed by a new microprocessor of a 24-h-ambulatory electrocardiographic device, which is based on real-time analysis and a solid-state ECG documentation. QRS detection was performed with an accuracy of 99%. Sensitivity and positive predictive accuracy were 70% and 87% for supraventricular ectopy. Using a fixed prematurity index of 80%, ventricular ectopy with a total of 7,845 beats was identified with a sensitivity of 64% and a positive predictive accuracy of 97%. With the additional consideration of late premature ventricular beats (PVB) sensitivity increased to 91% with a positive predictive accuracy of 96%. A sensitivity of more than 80% for singular PVB was achieved in 28/33 databases (85%), a similar positive accuracy was achieved in 27/33 databases (82%). Altogether, ventricular pairs and ventricular tachycardia resulted in a sensitivity of 83% and 76%, respectively, and in a positive predictive accuracy of 87% and 73%, respectively. Sensitivity exceeded 80% for ventricular pairs in 11/15 databases (77%) and for ventricular tachycardia in 10/14 databases (71%); similar results were observed for the positive predictive accuracy with 11/15 (73%) and 9/14 (64%) databases. In 42/44 databases and in all databases for arrhythmias of Lown class IVA and IVB, Lown classification was determined correctly. The Sirecust-Holter-ECG-system, a new device with real-time analysis and solid-state memory results in an accuracy for singular and complex ventricular arrhythmias, comparable to some of the presently available Holter systems.