Four hundred and eighty paired recordings obtained from 16 patients (55 +/- 10 yrs; 10 men, 6 women) were analyzed to determine the reproducibility of both acquiring and measuring myocardial velocities recorded by tissue Doppler echocardiography. To assess intraobserver variability, 1 observer recorded and measured the data twice, from the same patients, and to assess interobserver variability, patients were examined by 2 independent observers. For the left ventricle, intraobserver reproducibility was higher when assessing long-axis velocities (+/- 10% to 16%) than short-axis velocities (+/- 14% to 24%). For the right ventricle, intraobserver reproducibility was high for the tricuspid annulus (+/- 9% to 15%), but unsatisfactory for the right ventricular anterior wall (+/- 21% to 25%). The highest interobserver reproducibilities were obtained for systolic and diastolic velocities of the lateral mitral annulus (+/- 9% to 17%) and systolic velocity of the tricuspid annulus (+/-13%). Interobserver reproducibility of the ratio of early-to-late peak diastolic velocities was very low for all investigated sites (+/- 20% to 52%). With the use of current techniques and software, reproducibility of acquiring and measuring tissue Doppler echocardiography is suboptimal for both systolic and diastolic myocardial velocities.