Radiopaque markers were implanted in the left ventricular myocardial midwall in 58 patients and studied in the 30 degrees right anterior oblique projection by computer-aided fluoroscopy. Marker motion was used as a standard of segmental wall motion against which the accuracy of five methods for measuring left ventricular wall motion was assessed: two methods using hemiaxial measurements in rectangular coordinates, two using radial measurements in polar coordinates (all with frame-by-frame axial reindexing) and one using radial measurements in fixed external polar coordinates. The latter method showed significantly less error (25.9%, p less than 10(-6)) in measuring midwall marker motion than the other four methods (range 42.5--47.5%) in the group as a whole and in subgroups that had abnormalities of posterior, apical and anterior wall motion. This method also had the best correlation of marker motion and motion of adjacent endocardial border (of the overall left ventricle and the posterior, apical, and anterior walls separately) as visualized by ventriculography in 29 patients. The bulk of the reduction in error using this method was due to the use of a fixed external reference system, with a small additional increment of error removed by proper selection of the polar origin at a point 69% of the distance from the anterolateral edge of the aortic valve to the ventricular apex at end-systole.