The ability of magnetic resonance to determine regional left ventricular function was investigated in 18 patients--13 with coronary artery disease (nine with previous infarction), one with congestive cardiomyopathy, one with mitral stenosis, one with an atrial septal defect, and two without detectable cardiac abnormality. Coronal magnetic resonance images were acquired through the aortic valve and sagittal images were acquired in the plane of widest diameter of the left ventricle seen in the coronal image, both at end diastole and end systole. Regional wall motion assessed by magnetic resonance was compared with the results of anteroposterior and left lateral x ray ventriculograms by two independent observers. The left ventricular wall was divided into three segments in each plane and the motion of the segments was classified as normal, hypokinetic, akinetic, or dyskinetic. Muscle thickness was measured in each segment of the magnetic resonance images and was considered to be abnormal if in the systolic images it was less than 75% of that in neighbouring segments or if it failed to increase by at least 25% between diastole and systole. Wall motion assessments by the two methods agreed in 68 of 105 segments analysed, but differed by one class in 32 segments and by two classes in five segments. The differences can be explained by the conditions under which the investigations were performed and by the disparity between a tomographic section and an x ray projection. Magnetic resonance showed 25 segments to have abnormal wall thickness. Only one patient with infarction did not have an area of wall thinning and no patient without infarction had an area of thinning. It is concluded that magnetic resonance allows an accurate non-invasive assessment of left ventricular wall motion and thickness.