Changes in left ventricular size may be of importance in the development of angina and in its amelioration by glyceryl trinitrate, but left ventricular dimensions have not been measured during exercise in the upright position, the circumstance in which angina most often occurs. To assess changes in left ventricular end-systolic and end-diastolic dimensions, echocardiograms were obtained from 5 normal subjects and 6 patients with angina during rest and exercise upright, both before and after glyceryl trinitrate. The end-systolic dimension was considered an index of regional performance rather than an estimate of overall left ventricular behaviour. During exercise the end-diastolic dimension rose both in the normal subjects and in the patients. The end-systolic dimension fell progressively with exercise in the normal subjects but changed inconsistently in the patients. Glyceryl trinitrate lowered the end-diastolic dimension during exercise below comparable pretreatment values in both groups; the glyceryl trinitrate-induced decrease in end-diastolic dimension in the patients diminished at higher exercise levels. In all normal subjects and in 4 of the 6 patients with normal systolic shortening (= end-diastolic--end-systolic dimension) glyceryl trinitrate caused relatively small changes in the end-systolic dimension and systolic shortening on exercise. In contrast, in the 2 patients with hypokinetic ventricles glyceryl trinitrate conspicuously decreased the end-systolic dimension and increased systolic shortening. Thus, during exercise, in the normal subjects and in the patients with normally functioning left ventricles on echocardiography, glyceryl trinitrate caused a modest fall in the end-diastolic dimension and little change in the end-systolic dimension, but in patients with ventricular hypokinesia it greatly improved regional contractile performance. Alleviation of ischaemic hypokinesia during exercise may be an important part of the benefit of glyceryl trinitrate.