Routine exercise electrocardiography has been criticized for yielding too many so-called "false-positive" results. Recent studies in our institution indicate that evaluation of the time course behavior of ST segment and T wave (ST/T) changes after cessation of exercise differentiates ischemic from non-ischemic ("false-positive") stress electrocardiograms (SEs). Our method of assessing time course behavior is clarified. The principal aim of this study was to determine the accuracy of experienced observers in making this differentiation. Records of consecutive patients undergoing coronary arteriography over a 2 year period were reviewed and 30 with SEs judged positive for ischemia by the widely accepted ST/T configurational criteria alone were selected at random for the investigation. Sixteen subjects had normal coronary angiograms and had therefore previously been regarded as having false-positive SEs. Fourteen patients had at least one significantly (> 70%) stenosed coronary artery which was our yardstick for ruling that true myocardial ischemia, due to epicardial coronary artery disease (CAD), was present. Five observers, familiar with post-exercise ST/T time course patterns, independently and "blindly" analyzed all 30 configurationally abnormal SEs. Observers were informed only of the patient's age and sex; they were thus unaware of symptoms, exercise variables, coronary anatomy and the number of patients in the 2 groups. The observer consensus for ischemia of SEs using time course analysis was: total test accuracy 87%, sensitivity 79%, specificity 94%, positive predictive value 92% and negative predictive value 83%. Because all 30 patients had ST/T abnormalities, results of the sample for ischemia based on configurational criteria alone were sensitivity 100%, specificity 0% and positive predictive value 47%. We concluded that time course analysis plays a crucial role in evaluating SEs and that exercise electrocardiography remains a safe, cost-effective and reliable method of screening many asymptomatic, as well as symptomatic, subjects for CAD.