The use of computers in clinical electrocardiography is increasing rapidly; however, the role of computers with respect to the electrocardiographer has not been established. At present all electrocardiograms (ECGs) processed by computer are also interpreted by electrocardiographers; hense effort is duplicated. In an investigation of whether conditions can be defined under which the electrocardiographer can use the computer more profitably by eliminating some of the duplication, ECGs recorded in a university teaching hospital were processed by a computer program and subsequently reviewed by 1 of 10 electrocardiographers. For ECGs interpreted as showing normal sinus rhythm the rate of agreement between computer and human reviewer was 99%. For those showing a normal ECG pattern (contour) the rate of direct agreement was only 88%. However, the rate of occurrence of clinically significant differences was only 1.64%; hence the rate of essential agreement for this classification was 98.36%. Other classifications with good agreement were myocardial infarction, sinus bradycardia and sinus tachycardia. Therefore, in circumstances comparable to those of this investigation it is feasible for electrocardiographers to use computers to reduce greatly their workload without compromising the quality of the service provided.