Mean stress effects in finite-life fatigue are studied for a number of sets of experimental data for steels, aluminium alloys and one titanium alloy. Specifically, the agreement with these data is examined for the Goodman, Morrow, SmithโWatsonโTopper and Walker equations. The Goodman relationship is found to be highly inaccurate. Reasonable accuracy is provided by the Morrow and by the SmithโWatsonโTopper equations. But the Morrow method should not be used for aluminium alloys unless the true fracture strength is employed, instead of the more usual use of the stress-life intercept constant. The Walker equation with its adjustable fitting parameter ฮณ gives superior results. For steels, ฮณ is found to correlate with the ultimate tensile strength, and a linear relationship permits ฮณ to be estimated for cases where non-zero mean stress data are not available. Relatively high-strength aluminium alloys have ฮณ โ 0.5, which corresponds with the SWT method, but higher values of ฮณ apply for relatively low-strength aluminium alloys. For both steels and aluminium alloys, there is a trend of decreasing ฮณ with increasing strength, indicating an increasing sensitivity to mean stress.