Recent research suggests that left ventricular torsion is an important indicator of cardiac function. We used two theoretical models to study the mechanics of this phenomenon: a compressible cylinder and an incompressible ellipsoid of revolution. The analyses of both models account for large- strain passive and active material behavior, with a muscle fiber angle that varies linearly from endocardium to epicardium. Relative to the end- diastolic configuration, the predicted torsion exhibits several experimentally observed features, including a peak near end systole, rapid untwisting during isovolumic relaxation, and increased twist near the apex. The magnitude of the twist is sensitive to the fiber architecture, the ventricular geometry, and the compressibility and contractility of the myocardium. In particular, the model predicts that the systolic twist increases with increasing compressibility, contractility, and wall thickness, while it decreases with increasing cavity volume. The peak twist approximately doubles (from about 0.02 to 0.04 rad cm(-1)) with a doubling of myocardial compressibility or with a change in the endocardial/epicardial muscle fiber angles from 90/ -90 degrees to 60/ -60 degrees. The twist is less sensitive to changes in contractility and ventricular geometry. These findings provide a basis for interpreting measurements of ventricular torsion in the clinical setting.