We examined the effect of a training programme to reduce interobserver variation in interpretation of electrocardiography in suspected myocardial infarction. Sixteen doctors with 6-24 months of clinical training in internal medicine read serial electrocardiographic recordings in 107 patients and assessed whether signs indicative of acute myocardial infarction were present. There was disagreement in approximately 70% of cases. Eight of the doctors were randomly allocated to attend an 8-h intensive course on interpretation of electrocardiography in myocardial infarction. The remaining eight participants were allocated to a control group, received no training, and were not told about the subject of the study. All the doctors then reviewed another series of electrocardiographic recordings. No difference was found in the level of agreement within the two groups before and after the training programme, or between the two groups before and after the training. The raters' ability to discriminate between electrocardiograms with a high and low indication of infarction remained unaffected. We conclude that the training programme did not increase agreement regarding the interpretation of electrocardiographic data in suspected myocardial infarction. Our results suggest that the diagnostic approach of physicians is established at a very early stage in their clinical training. The effect of training programmes should be evaluated by the use of randomized clinical studies.