Criteria for the fatigue assessment of multiaxial, non-proportional stress cycles have been prescribed by engineering design codes and implemented in fatigue post-processors. Cases in point are the normal-stress (API RP 17G), Findley, Mises (BPVC-VIII-2), Mises–Sines, Tresca (BPVC-III-1), and Tresca–Sines (BPVC-VIII-3) criteria (design codes). In the present work, these criteria have been applied to the tension–torsion fatigue limit.Theoretical predictions have been compared with fatigue test data from 220 different tension–torsion fatigue test series on wrought steel, cast iron and an Al alloy. The normal-stress criterion yields excellent predictions for cast iron, whereas those for wrought alloys tend to be unsafe. For wrought steel and the Al alloy, the shear-stress criteria yield predictions that are closer to the observed data. For all criteria, mean values of the predictions are within 12% of the observed fatigue limits, whereas standard deviations are typically <20% of the mean values.In addition to the traditional formulation of critical-plane criteria, based on extreme values of normal and shear stress components during a cycle, a ‘simultaneous stress component’ (SSC) formulation according to the ASME code, based on the stress components for a critical pair of points in time during the cycle, has been considered. For the stress cycles investigated, the SSC formulation makes a significant difference only for the Findley criterion.