We determined the intra, inter-observer and temporal variability of upright bicycle exercise echocardiography in 50 consecutive patients with suspected coronary heart disease. Using significant artery stenoses as reference, the sensitivity of digital exercise echocardiography was 89% and the specificity 92%. Regarding the presence or absence of exercise-induced ischaemia, intra-observer agreement was 86% (73-94% with 95% confidence limits, kappa value (K) 0.70 (0.50-0.91)). Inter-observer agreement between two independent observers was 92 (81-98)% (K = 0.81 (0.63-0.99)). Temporal variability was examined by repeating exercise echocardiography after one week in 30 stable patients; it demonstrated 90 (78-97)% agreement (K = 0.71 (0.40-1.00)). The inter-observer agreement appeared to be lower in patients or myocardial segments with wall motion abnormalities at rest (86% and 82%, respectively) than in patients or segments with normal myocardial function at rest (96% and 94%, respectively); ns, P < 0.05, respectively). Analysis of specific regions showed a significantly higher level of agreement (P < 0.05) regarding exercise-induced ischaemia in segments supposed to be supplied by the circumflex branch of the left coronary artery (96%) than in the perfusion bed of the left anterior descending branch (87%) and right coronary artery (88%). Comparing estimated values of echocardiographic ejection fraction, analysis of intra-observer variability showed a correlation coefficient of 0.91 and 95% confidence limits of a single estimate of ejection fraction of +/- 10.3%. Corresponding analyses of inter-observer and temporal variability showed correlation coefficients of 0.93 and 0.84, respectively, and 95% confidence limits of single estimates of ejection fraction of +/- 9.6% and +/- 13.0%, respectively. Thus, the diagnostic result and the reproducibility of digital exercise echocardiography are satisfactory and comparable with those obtained by myocardial scintigraphy. However, approximately 10% of the examination results may be reversed when the test is repeated or reevaluated by the same or by another observer. The result seems to be less reproducible in patients with abnormal wall motion at rest than in patients with normal myocardial function before exercise.