In this work the hypothesis “All quenching of steel components should be interrupted at ~50% martensite formed in order to achieve better fatigue performance” has been tested. The results are very positive. The effect of slower quenching seems to be generic, i.e. valid for different steels and different heat treating processes. It has now been studied on three steels, three heat treatment methods and three quenching method (in this and previous projects) and it can be concluded that in all cases slower quenching gave better fatigue resistance when no HTTPstructures were formed in the surface. HTTP-structures are believed to act as a low strength skin neutralising the increased fatigue resistance from slow cooling. One important fact that brings us to the conclusion that the effect is generic was that in all cases dilatometer measurements with slower cooling gave less expansion during the martensite transformation, indicating more retained austenite, which would explain the better fatigue resistance by crack arrest from expansion of transforming retained austenite during fatigue. EBSD measurements have shown that the martensite laths were larger in the slower cooled samples and indicate more retained austenite in slower cooled samples. No difference in retained austenite content could be measured by x-ray, not on the fatigue test bars or in the dilatometer test bars. The measurement error in the x-ray method is likely to be too large to resolve the smalldifferences between different coolingrats. The effect of slow cooling on residualstresses remains unexplained. Smallerdilation during phase transformation inthe surface is thought to yield lowercompressive stresses, not higher asmeasured on the test bars.