The analytical approach to the double-torsion, constant-displacement-configuration technique has forced a strong oversimplification of the specimen model, and, therefore, of the strain field in classical applications. For the present study, accurate three-dimensional finite-element analysis was performed on a realistic specimen; that is, one with finite thickness, a groove, and an initial notch, The influence on the strain-energy release rate of the specimen width and length, the fracture length, the presence and shape of the side groove, the presence of an initial notch, and the curved shape of the fracture front was estimated from this analysis. Significant deviations from the classical analytical solution mere found, The present results should reduce the discrepancies found among measurements made in various laboratories and that traditionally have prevented the development of effectively applicable fracture models.