Regional abnormalities of left ventricular function are frequently determined from superimposed cavity outlines at end-diastole and end-systole. The timing of end-diastole is standardised with respect to the electrocardiogram, but that of end-systole is undefined, and frequently determined by simple inspection. To examine the effect of variation in the timing of the end-systolic frame on apparent regional wall motion, 35 left ventricular angiograms from patients wit ischaemic heart disease were digitised frame by frame, and the timing of aortic valve closure and mitral valve opening superimposed on contour displays. In 15 patients, the pattern in the two frame display with end-systole taken as aortic valve closure was similar to that when it was taken as mitral valve opening. In the remaining 20, however, there were major discrepancies between the two, because of changes in cavity shape during isovolumic relaxation. These shape changes were the result of asynchronous termination of systole in different regions of the ventricle. In addition, there was a significant increase in left ventricular cavity area of 3.9 +/- 1.7 per cent during isovolumic relaxation, corresponding to an increase in calculated volume of 8.1 +/- 4.5 per cent. Estimates of ejection fraction derived from end-systole taken at end-ejection were thus consistently greater than those at mitral valve opening. These differences in the apparent pattern of regional wall motion resulting from lack of definition of end-systole are large compared with those reported after intervention angiography, or surgery. Thus, if two frame methods are to be used, the timing of end-systole must be standardised. We suggest that end-ejection be used. Changes of shape during isovolumic relaxation should be considered separately.