In the past two decades, normal endothelial function has been identified as integral to vascular health. The endothelium produces numerous vasodilator and vasoconstrictor compounds that regulate vascular tone; the vasodilator, nitric oxide (NO), has additional antiatherogenic properties, is probably the most important and best characterised mediator, and its intrinsic vasodilator function is commonly used as a surrogate index of endothelial function. Many conditions, including atherosclerosis, diabetes mellitus and even vascular risk factors, are associated with endothelial dysfunction, which, in turn, correlates with cardiovascular mortality. Furthermore, clinical benefit and improved endothelial function tend to be associated in response to interventions. Shear stress on endothelial cells is a potent stimulus for NO production. Although the role of endothelium-derived NO in acute exercise has not been fully resolved, exercise training involving repetitive bouts of exercise over weeks or months up-regulates endothelial NO bioactivity. Animal studies have found improved endothelium-dependent vasodilation after as few as 7 days of exercise. Consequent changes in vasodilator function appear to persist for several weeks but may regress with long-term training, perhaps reflecting progression to structural adaptation which may, however, have been partly endothelium-dependent. The increase in blood flow, and change in haemodynamics that occur during acute exercise may, therefore, provide a stimulus for both acute and chronic changes in vascular function. Substantial differences within species and within the vasculature appear to exist. In humans, exercise training improves endothelium-dependent vasodilator function, not only as a localised phenomenon in the active muscle group, but also as a systemic response when a relatively large mass of muscle is activated regularly during an exercise training programme. Individuals with initially impaired endothelial function at baseline appear to be more responsive to exercise training than healthy individuals; that is, it is more difficult to improve already normal vascular function. While improvement is reflected in increased NO bioactivity, the detail of mechanisms, for example the relative importance of up-regulation of mediators and antioxidant effects, is unclear. Optimum training schedules, possible sequential changes and the duration of benefit under various conditions also remain largely unresolved. In summary, epidemiological evidence strongly suggests that regular exercise confers beneficial effects on cardiovascular health. Shear stress-mediated improvement in endothelial function provides one plausible explanation for the cardioprotective benefits of exercise training.