The accuracy of 2 commercial 24-hour electrocardiogram analyzers was tested for quantitation of ventricular premature complexes (VPCs). Scanner 1 was the Cardiodata Systems Mark III and scanner 2 was the Avionics Trendsetter DCG VII. Twenty four-hour electrocardiographic recordings from 19 consecutive ambulatory patients with frequent VPCs were analyzed by each device. Results were compared with those from hand counts of complete printouts of each of the 19 recordings. For total VPCs, scanner 1 had an average error of 13% (range 0 to 58%) and scanner 2 had an average error of 24% (range (1 to 80%). Scanner 1 had an error of more than 10% for 9 of the 19 recordings and scanner 2 more than 10% for 11 of the 19 recordings. For paired VPCs, scanner 1 had a mean error of 23% (range 4 to 77%), and scanner 2 of 56% (range 34 to 79%). For nonsustained ventricular tachycardia, scanner 1 had an average error of 20% (range 8 to 41%) and scanner 2 had an error of 56% (range 34 to 78%). Thus, when recordings from consecutive ambulatory patients with frequent VPCs were analyzed, neither device was consistently accurate for quantitation of total VPCs. Both analyzers had an unacceptable error for quantitation of repetitive VPCs. All currently available devices may have comparably large errors. This possibility is confirmed by recalculation of the reported data from a third scanner.