Most of the existing fatigue design tools that account for mean stress effects were originally based on limited experimental data with relatively low mean stresses. However, residual stresses resulting from manufacturing processes are frequently at or near the yield stress level of the material. Therefore, to improve the fatigue design of engineering components which have residual stresses, more experimental as well as analytical studies are required in this regime. In this paper data from several constant maximum stress and constant minimum stress fatigue test sets are presented. These tests are designed so that material constants required to implement a closure-based fatigue damage method can be established. It is then shown that mean stress effects on fatigue are more complex than what can be modeled by several of the most commonly used methods. By adopting a more mechanism-oriented crack closure concept in an initiation-based fatigue approach the life prediction capability can be significantly improved.