In order to compare variability in M-mode echography and MRI in the assessment of left ventricular mass, 20 echogenic patients without evidence of coronary artery disease were investigated. Two MR and two M-echo examinations were performed within 4 days by different trained operators, each unaware of the other's results. M-mode echo was carried out according to Devereux's method, using the 'Penn-Cube' formula. MR protocol included multislice (8 to 12) true, short-axis spin-echo imaging (10 mm thick with a 1 to 3 mm gap) encompassing the entire left ventricle. Planimetry was manually traced with standardized window settings. Correlations between both echographic and both MR measurements showed r = 0.89, SEE = 22.7 g and r = 0.96, SEE = 11.2 g, respectively. Mean inter-study variability was 11 +/- 6.4% and 6.75 +/- 3.8% (P = 0.0021). The threshold value corresponding to the 95th percentile of the variability data was 21.5% for echography and 13.5% for MR. In conclusion, MR appeared to be a significantly more reproducible examination tool, when compared with M-mode echo, for the evaluation of left ventricular mass (variability, 63% higher with echo than with MR). The main practical consequence of this result lies in the reduced number of patients required to demonstrate a significant change in the LVM with MR as compared with echography.