To predict fatigue life under variable loading accurately, it is important to understand the damage factors determining fatigue life. In this study, the damage factors of high-cycle fatigue were investigated for Type 316 stainless steel and a procedure for predicting the fatigue life of two-step tests was discussed. Fully-reversed axial fatigue tests were performed in ambient air at room temperature. In order to evaluate applicability of the linear damage accumulation rule, two-step tests were performed. The crack initiation and propagation were investigated by replica observations. It was found that the linear damage accumulation rule did not always evaluate the fatigue life conservatively due to stress dependence of crack size against normalized fatigue life. Furthermore, the fatigue limit decreased in the two-step test. In order to evaluate the role of cracking on the fatigue life, the surface layer of specimens was removed after the first step of the two-step tests. The fatigue lives of the surface-removed specimens were almost equivalent to that of undamaged specimens, although the fatigue limit was less than that of the undamaged specimens. It was revealed that the fatigue limit of surface-removed specimens decreased due to cyclic softening caused by the cyclic loading in the first step. The fatigue lives correlated with the strain amplitude rather than the stress amplitude and fatigue limit was almost the same for the surface-removed, undamaged, and undamaged specimens subjected to constant strain tests. It was concluded that the fatigue life of Type 316 stainless steel should be predicted based on the strain amplitude even for the high-cycle regime. Accuracy of the fatigue life prediction could be improved by considering the stress dependency of the crack size in addition to the cyclic softening and hardening material behavior.