Although the anteroposterior dimension of the left atrium is universally used in clinical practice and research, we hypothesized that it may be an inaccurate surrogate for volume because its use is based on the unlikely assumption that there is a constant relation among atrial dimensions. The following measurements of the left atrium were made at end ventricular systole: (1) M-mode-derived anteroposterior linear dimension from the parasternal long-axis view; (2) digitized planimetry of the left atrial (LA) cavity from the apical 4-chamber view; and (3) digitized planimetry of the LA cavity from the apical 2-chamber view. The following volume calculations were obtained from these digital measurements: (1) volume derived from the M-mode dimension assuming a spherical shape; (2) volume derived from the single plane area-length of apical 4-chamber view, which assumes that LA geometry can be generalized from a single 2-dimensional plane; and (3) volume derived from the biplane method of discs. The correlation coefficient between the M-mode and biplane methods of determining LA volume was r = 0.76. The mean difference (+/-2 SDs) between these methods is -25 +/- 33 ml. The correlation coefficient between the single plane apical 4-chamber and biplane methods of determining LA volume is r = 0.97. The mean difference (+/-2 SDs) between these methods was -5.0 +/- 12 ml, indicating good agreement. The M-mode measure of the left atrium is an inaccurate representation of its size. Two-dimensional-derived LA volumes provide a more accurate measure of the true size of the left atrium and are more sensitive to changes in LA size. When an echocardiographic measure of LA size is made either in an individual patient or as a variable in a research study, the M-mode measure should be avoided.