The objectives of this review are to discuss the diversity of mechanisms that may explain the association between heart rate (HR) variability and mortality, to appraise the clinical applicability of traditional and new measures of HR variability and to propose future directions in this field of research. There is a large body of data demonstrating that abnormal HR variability measured over a 24-h period provides information on the risk of subsequent death in subjects with and without structural heart disease. However, the mechanisms responsible for this association are not completely established. Therefore, no specific therapy is currently available to improve the prognosis for patients with abnormal HR variability. Reduced HR variability has been most commonly associated with a risk of arrhythmic death, but recent data suggest that abnormal variability also predicts vascular causes of death, progression of coronary atherosclerosis and death due to heart failure. A consensus is also lacking on the best HR variability measure for clinical purposes. Time and frequency domain measures of HR variability have been most commonly used, but recent studies show that new analysis methods based on nonlinear dynamics may be more powerful in terms of risk stratification. Before the measurement of HR variability can be applied to clinical practice and used to direct therapy, more precise insight into the pathophysiological link between HR variability and mortality are needed. Further studies should also address the issue of which of the HR variability indexes, including the new nonlinear measures, is best for clinical purposes in various patient populations.